Learnings

A growing log of guests' defining first bets and the frameworks they use to decide under uncertainty.

---

Framework Summary — How Guests Decide Under Uncertainty

A running master list of decision frameworks across all episodes, ordered by how many guests use them. Updated as each new episode is added. Guests so far: Alec Torelli (Ep 1), Shaun Gold (Ep 2), Simon Lancaster (Ep 3), Arkady Kulik (Ep 4), Aram Attar (Ep 5), Ihar Mahaniok (Ep 6), Brian Bell (Ep 7), Alex McNaughten (Ep 8), Mike Ma (Ep 9), Howard Lindzon (Ep 10), Sohun Sanka (Ep 11), Kirby Winfield (Ep 12), Gregory Raiz (Ep 13), Eric Jorgenson (Ep 14), Mahesh Narayanan (Ep 15), Martin Jacobson (Ep 16).

Shared by multiple guests

1. Trust your own signal over the crowd's "no" (Alec Torelli, Shaun Gold, Simon Lancaster, Arkady Kulik, Aram Attar, Ihar Mahaniok, Brian Bell, Alex McNaughten, Mike Ma, Howard Lindzon, Sohun Sanka, Kirby Winfield, Gregory Raiz, Eric Jorgenson, Mahesh Narayanan, Martin Jacobson — 16 of 16 guests) When information is thin, all sixteen treat consensus as noise rather than a veto — though each trusts a different kind of signal. Alec: fear of others' judgment is the biggest barrier; "when I trust my spirit, it ends up being the right decision." Shaun: never outsource your mind (to social media, AI, or influencer playbooks), bet on your own unique weirdness, and treat "this seems crazy" as a possible signal you're early — his inner voice ("North Star") carried the fund decision when every external signal said no. Simon is the interesting twist: his signal isn't intuition but proprietary evidence — when nearly every financially focused LP said "too niche," he kept underwriting the bet with what thousands of operator conversations were telling him up close. Specialization, in his words, exists to "see ahead of the curve and around corners where others couldn't"; the gap between consensus and firsthand evidence is the alpha. Arkady is in Simon's evidence camp: the crowd's "no" on medical devices is a reflex ("people hear medical device — immediately FDA, regulation, no no no"), so he refuses to inherit it as a gate; he wrote the first VC check into a neuromodulation startup because ten-plus hours with the founder and deep physics/neuroscience diligence turned the feared risks into variables he could actually price — and that reflex-vs-analysis gap is where his returns live. Aram pushes the evidence camp to its logical end — he distrusts the inner signal entirely (intuition is "only recognition of what you've done before"), so what he weighs against the crowd is a researched framework. The crowd's "no" on emerging managers rests on track record and social proof; persistence research says that compass is broken (a top-quartile fund repeats at ~30% — worse than a coin flip), so he published his five-trait thesis in advance, staked his reputation on it, and puts his own money into the shunned Fund I/II pocket — where every LP still running the consensus screen hands him "an unfair advantage that nobody else has." Ihar Mahaniok joins the evidence camp with a reference-class twist: in 2012 the crowd's "no" on grocery delivery was really a "no" to Webvan, so he re-derived the analogy himself — "the key thing that stood out for me for Instacart was that it wasn't that it was similar to Webvan, but that it was similar to Uber" — and wrote the seed check into a category VCs treated as a punchline, knowing full well "some engineers wouldn't invest" because it looked too operational. The famous failure had priced the category; his own analysis re-priced it, and it became one of his five unicorns. Brian Bell states the same idea as a pricing law: consensus is expensive by construction. "If you just do consensus investing, everything's gonna get bid up" — the hot round everyone wants prices at a hundred or two hundred times forward revenue, sometimes at infinite forward revenue because there's no revenue at all. So the job of an early-stage investor is to hunt "that sweet spot of kind of this sure bet that nobody knows about" — the same non-consensus-and-right posture he keeps toward his own machine, which he'll override when he thinks a founder is a 4.5 and it said 4. Alex states the information asymmetry outright and makes it his parting advice: "everyone around you is gonna have an opinion about what you should do, and at the end of the day, they have far less information than you." His friends thought he was mad for leaving a profitable business and a sea view; the venture consensus held that AI sales tech was already overfunded. He weighted both at close to zero against what a decade inside the profession was telling him — that the money was piling into top-of-funnel email volume while everything after the first conversation went untouched. Mike Ma is the mirror image, and it belongs here for that reason: his signal held against the crowd's yes. After thirty days with a climate deep-tech team he concluded he couldn't underwrite a commercialization path that amounted to "just trust us" from a corporate sponsor — and passed, in writing, while the round went on to oversubscribe. "You may be right. I can't underwrite that." His stated goal is the general form: "most people want to know more than the market to make money. I want to know more than the rest of the cap table" — and once you do, the cap table's verdict in either direction is just another opinion. Howard Lindzon's crowd wasn't lukewarm — it was hostile. In 2013 every Silicon Valley VC had committed to building a better Vanguard and, in his words, "hated the idea of trading"; he knew it firsthand from failing to help eToro raise in the Valley in 2010. He wired $100K into Robinhood off a disconnected design mockup anyway, because the crowd's "no" was a reflex about E*Trade and his "yes" was fifteen years of watching retail traders — and he treats the two as categorically different kinds of information: "I definitely had low information, but I also had the most information." Sohun Sanka's crowd was closer to home: family and mentors telling him to stay risk-averse while young, plus an industry orthodoxy that a solo founder isn't a real founder. He audited both rather than obeying them — giving up, by his own estimate, over a million dollars of total comp across the moves, and starting alone after a mentor pushed him to ask what the co-founder requirement was actually solving. Kirby Winfield had to override two crowds at once, and one of them was his own. The Seattle consensus in 2022 held that vehicular autonomy was solved and that anything smelling of defense "wasn't gonna win you any pats on the back from your local LPs"; a prominent local investor had already passed on Overland AI; and Kirby himself had told his LPs in public that he was not a hardware investor. He wrote the check anyway, into a company that had not yet incorporated, because the crowd's "no" was a reflex about on-road self-driving while his "yes" was first-principles arithmetic about hundreds of thousands of military ground vehicles. He has since hardened the idea into a rule about process rather than courage: he tries not to know who else is in the round. "If I feel like I need to know who else is investing, then I shouldn't make the investment... It's my taste, it's my gut, it's my algorithm. And the minute I start outsourcing that to someone else, I'm doing my LPs and myself a disservice." And he extends the discount to the signal most seed investors find hardest to ignore: a multi-stage firm in a pre-seed round is buying optionality, not validating the deal. Gregory Raiz went looking for the crowd on purpose and concluded it wasn't worth listening to. He joined seven angel groups precisely to surround himself with people who knew more than he did, and found that many angels "tend to not take it particularly seriously. They're not digging into the data, they're not doing the research" — the deciding input was usually "I like this idea, I like this founder." He lost money in that room. His replacement for it is the evidence camp's purest form — a survey instrument fitted against 3,000 entrepreneurs' outcomes — and the findings it returns are themselves anti-consensus: the prestigious degree that makes a founder fundable doesn't predict success, the older non-archetypal founders he almost dismissed on his first angel check outperform, and the immigrant-founder thesis some funds market may or may not survive contact with the data. His closing advice is the test stated as a habit: "do I think this is true because other investors have told me this is true? Or because I've actually done a little bit of digging to verify?" Eric Jorgenson's crowd was the conventional wisdom about his whole category: "nobody ever makes money writing books," you're over-investing in a passion project, why spend three years on something you could do in three months? He wrote the biggest check of his life anyway, because he had replaced the market's scoreboard with a reader of his own choosing — "an audience of like one or two," himself and Naval — and a definition of success (break even, be proud of it) that no outside verdict could touch. His investing version is the same move against the crowd's yes: whatever other tools you use, the only question is whether you're in the fat tail of the power law, "and everything else is noise." Mahesh Narayanan's crowd was the follow-on market. Impruvon Health was, in Martin's words, "not an easy company to raise money for even after we invested": a good seed fund liked the senior-care market but called it too early, the founders had to cut a $3M raise in half, and even the smaller round took six or seven months. Mahesh's signal was proximity evidence — a year of watching the founder work inside his own accelerator, a compliance jump from 30–40% to 80–90% that he calls "unheard of," and the first unpaid pilot converting to paid — and against it the slow market read as a timing problem, not a verdict. He stayed in, and the company found its real business on the far side of the grind. Martin Jacobson's version is loyalty to his own process. Before the 2014 World Series of Poker final table no one pushed back on his months of preparation, but "if they were, it was just noise to me": almost no one had ever been in that seat, so few were qualified to say what was right, and the right approach is individual anyway. He knows his own temperament — "more on the over-preparer side, for better or worse" — and trusted the process built for it.

2. Bet from your edge — mastery picks the game, not opportunity (Alec Torelli, Shaun Gold, Simon Lancaster, Arkady Kulik, Aram Attar, Ihar Mahaniok, Brian Bell, Alex McNaughten, Mike Ma, Howard Lindzon, Sohun Sanka, Kirby Winfield, Gregory Raiz, Eric Jorgenson, Mahesh Narayanan, Martin Jacobson — 16 of 16) All sixteen guests converge on it. Alec bets only where the risk/reward calculus favors him — his edge was 10,000+ hours of poker mastery; being a "risk trader" means being selective. Shaun: solve your internal problems first — money and a domain track record — before betting on an external problem ("don't write biotech decks if you know nothing about biotech"); your own unrepeatable core competencies are the durable advantage. Simon turned it into a formula that he applies to funds and founders alike: Alpha = Mastery × Focus × Network — deep domain expertise, staying inside your area of edge, and the relationships to earn your way in. You can run low on one, but never all three. Arkady's version is the deep-tech corollary: your edge is the ability to actually evaluate the science — read the corpus, or have the expert (his firm has a scientific partner), or at minimum know the landscape summary — and if you can't build that ability, don't play the game. His proof by counterexample: "a lot of people who understand how biology works would have never invested in Theranos" — everyone with the edge saw the snake oil long before the market did. And he insists the edge is idiosyncratic: every VC's process must be tailored to their own perception of risk; his 28% weight on science/tech and someone else's 14% "might both be right — for themselves." Aram's edge sits one level up the stack: ten years reverse-engineering how power law masters decide gave him an evaluation system that works precisely where the standard tool (track record) doesn't exist — so mastery picked his game too, emerging Fund I/IIs, the one pocket his framework can underwrite and most LPs can't. And he demands the same property of the GPs he backs: "Is this something you understood nobody else understood? Do you have another advantage here?" Ihar's edge is a braid of three strands he names outright: an engineer's ability to vet technical talent (the talk-shop test), a connector's network ("I know thousands of people"), and the immigrant experience itself — "I realized that this is my angle, this is my advantage." He formalized it into Geek Ventures' thesis because immigrant founders both outperform statistically and need his most natural value-add (connections) the most — and he insists the lens only works because it's authentic to his own path from Belarus: an inauthentic thesis neither filters well nor attracts the founders it's meant for. Brian's edge is two things his résumé handed him and almost no allocator has both of. First, the machine-learning one: he led AI at Amazon and launched the SageMaker Marketplace, so when he started investing he didn't reach for a checklist, he reached for first principles — "how would I set up feature extraction? What are the features that matter? And what are their weights and biases in the model that would output a score that would predict success?" Second, a calibrated sense of excellence from proximity to it: "when you've been in the room with Andy Jassy at AWS and he's grilling you on a project, you kind of know what excellence looks like." His self-assessment is the sharpest statement of edge in the whole series — "I don't even think I'm an A player. I'm probably like a B player. But I'm a B player who can spot A players." Alex is the clearest case of edge choosing the game: he had spent a decade training salespeople and had supported two to three hundred companies through hands-on services, and he attributes his contrarian read of the market to exactly that — "maybe it's because I just lived in this world for so long." Notably, he did not try to acquire the edge he lacked; his co-founders are "like a million times more technical than me," one of them fresh out of Cambridge with a master's in machine learning, so the AI wave became leverage on his own mastery rather than a field he had to enter from scratch. Mike Ma built the entire fund around his edge and is disarmingly explicit about its limits: "I'm a go-to-market guy. I don't know that much" about space materials, so on Melagen Labs he didn't try to evaluate the science — he evaluated the LOIs, the one artifact a lifetime of marketing and commercial roles at Bank of America, Vanguard and Betterment qualified him to read, and underwrote the founder's response to that coaching. The same edge is why Sidecut's thirty-day engagements are always customer discovery, funnel, and go-to-market rather than technology, and why his one non-negotiable is "go-to-market killers": he only claims conviction on the thing he can actually see. Howard Lindzon gives the edge its most fatalistic framing: "I'm supposed to have made that investment based on everything that I had done before." His edge was a stack — the 1999 retail era, founding StockTwits, pitching the trade button to Twitter in 2008, eToro in 2010 — and a live feed of it: "I talk to a million people a day on StockTwits, so I know what they want." That's why "if you build it, they will come" was a knowable fact for Robinhood and Alpaca and not a slogan; and, like Mike, he's explicit about the boundary — "I don't have expertise across a hundred different subjects," and he didn't see prediction markets coming. His advice to the next generation is the same rule stated as a career: build domain experience, and as an investor he's "not looking for generalist founders" but for people who'll work one problem for the rest of their lives. Sohun's edge is a seat nobody else had occupied: head of go-to-market inside TikTok Shop SaaS — an industry three years old — on top of agency operating. When his software plan collapsed he picked the business that edge could carry: "no one has done that... therefore I have a level of product manager acumen where I can demonstrate how these softwares run end to end, explain it to a dev, explain it to a designer, sell it, educate it." The same fact is his positioning against every other TikTok Shop solutions provider, who "don't have the angle of being a TikTok Shop SaaS PM plus agency operator." Kirby Winfield is the inversion that proves the rule, because he found his edge by first naming what it wasn't. An English major and "recovering marketer," he outsources technical diligence to his network — Rob Coneybeer's engineering read is part of why the Overland check happened — and he counts the cost of having pretended otherwise as his anti-portfolio: generational companies in data center energy management and automotive fleet management that he "talked myself out of because I thought that I had to be some sort of subject matter expert to invest." The edge he does claim was built on purpose, in two years as founding investor in residence at AI2 among AI research PhDs: "I just need to be an expert at understanding founder capacity," including pattern recognition for which PhD archetypes convert into company builders. That mastery picked his game as surely as anyone else's — roughly half the founders in his current fund are PhDs out of Carnegie Mellon and the University of Washington. Gregory Raiz sits beside Kirby and Aram rather than the domain experts: on his first angel check he says plainly "I wasn't necessarily a prop tech investor, I didn't have a lot of deep understanding of that space." The edge he went and built instead is an instrument. Having decided that "for me to have the best impact, I have to get really, really good at investing," he spent three years assembling the quantified founder data nobody would give him, and he deploys it exactly where the usual edges are unavailable — cold inbound, where there is no warm introduction to lean on and no domain expertise to substitute. He also thinks founder-market fit is the same principle one level down, and it is now half his thesis: a third-time founder in their own vertical is strongly predictive of another win, while the same founder in a new domain performs like the average. Eric Jorgenson's edge is the one he'd been building for free: years of reading and writing in his spare time, tweets that became a blog that became a book. "The things that you do for fun... that you do the best at" — he says it's hard to really believe, and the Almanack is his proof. He applies the same test to founders as the sensible-trajectory screen: the company should be the natural next chapter of their PhD, their first job, their failed first company — the kid who loved spaceships and grew up in his dad's metal shop, thirty years into asteroid mining — rather than a pivot from drop shipping to crypto to AI that marks someone "perpetually chasing a market." Mahesh Narayanan's edge is proximity he built on purpose. He came to the deal through the Autism Impact Fund and a deep family connection to neurodiversity, and he helped build an accelerator for neurodiversity technologies — which is what gave him a year of watching Impruvon's founder work before the check, access that a cold investor can't buy and that Martin openly "rode" on when he co-invested. He is also explicit that the edge has to change with the asset: on a software company he weights the founder about 70%, because the moat is execution; on hard life sciences his edge has to reach the science itself — roughly 30% team, 30% IP, 40% data, plus a demand that someone outside the company has replicated the result. Martin Jacobson's edge is transferable craft. He is explicit that a poker pro's habits of managing capital, risk and execution carry straight into founding, and he picked Cuppacoin where his edges overlap: poker taught him that "incentives is everything," and his chef's background in hospitality, plus his angel portfolio, told him staff churn is a problem across the whole industry. He chose the seat by fit, too — he weighed joining a VC firm full time, but the zero-to-one work suits him better.

3. Think in long horizons, not immediate results (Alec Torelli, Shaun Gold, Arkady Kulik, Aram Attar, Brian Bell, Alex McNaughten, Mike Ma, Howard Lindzon, Sohun Sanka, Eric Jorgenson, Martin Jacobson — 11 of 16) Eleven of the sixteen decouple the decision from short-term outcomes. Alec separates decision quality from results (avoid "resulting") and obsesses over process, since EV is locked in at decision time. Shaun thinks in seasons — "you gotta eat twelve to twenty-four months" — willingly trading a defined 1–2 year window of losses for a decades-long payoff, and stress-tests bets with the "What now?" question: am I ready for the long, invisible execution phase after the bold move? Arkady's version turns the horizon itself into a diligence criterion: "this is a very long game" — a pre-seed check means seven to ten years attached to this founder, so he underwrites the relationship (trust, open communication) as hard as the business, because starting a decade-long partnership with broken communication "is gonna end in disaster." Aram turns the decouple into his selection criterion: what sold him on Martin was an algorithm that revisits passed deals to separate decision quality from decision outcome, and he'd re-up a sound decision-maker who got unlucky — luck cuts both ways, as he says of his own $2B exit ("mostly right time... I haven't done it twice"). The horizon half is his LP hygiene: GPs pitching him quick liquidity get "relax, dude — I know it takes time to build a company, so I know it takes time to build a great fund." Brian supplies the cleanest image of the problem: "it's like you're playing poker but you don't find out if you win the hand for five years." Because the feedback loop is that long, he refuses to let outcomes stand in for process — he goes back through the decision on winners and shutdowns, re-runs the model as of the original date on the information he actually had, and updates from that rather than from the scoreboard. He is equally blunt that the horizon applies to the investor, not just the investment: "in VC, they say it takes ten million to make a VC. I think you got to go deploy into like a hundred startups before you really know what you're doing." Alex applies the horizon at the moment of choosing partners rather than measuring results: the question he set out to answer about his two prospective co-founders was "can we work together for 10 years and like be in the trenches for 10 years?" He also frames the whole move as a continuation of his life's work rather than a discrete venture, and gives the poker version of the same idea — you will not win every hand, but "if you lock in for a session, like you'll do all right." Mike Ma is squarely in Arkady's and Alex's camp — the horizon is the relationship. His dad joke is the underwriting logic: the average US marriage lasts 8.6 years and the average stay on a cap table is about the same, "so why doesn't someone move their toothbrush in for a month?" Thirty days of shared work before the check exists precisely because he is signing up for the better part of a decade with this person and wants to know how they handle ambiguity, conflict and indecision before, not after, the wire. Howard Lindzon's horizon is the career itself: get close to the leaders, stay humble, "and the money will just come to you after 10, 15 years because they leave so much alpha." It took StockTwits a decade, so the founder he's looking for is one who will "be trying to solve this problem for the rest of their life" — and the first thing a young trader must learn, before any strategy, is how they react to their first blow-up, because a lifetime of compounding depends on not making four big mistakes in a row. Sohun named the trade explicitly on his way out of SaaS — "the short-term cash versus long-term upside" — walking away from compensation that "only got bigger and bigger" because AI was thinning the moat under the entire category. He is the case that shows the frame is about horizon, not pace: his operating rule is aggressively short (nothing that doesn't move the constraint in the next eighteen hours), and it is the long-horizon read on where the category is going that decides which constraint he is working on at all. Eric Jorgenson worked the long horizon from both ends. The Almanack took three years of nights and weekends that he refused to compress to three months, and the founders he backs are the ones whose decades of curiosity about a problem mean they'll "continue all the way through the messy middle" rather than cash out; he also frames his own risk appetite as a slow psychological build — you have to grow yourself along with the portfolio, or you'll abandon a strategy that works just because the numbers got bigger. Martin Jacobson brings the poker player's version straight from the table: "every decision has an EV," so you judge the decision rather than the result, because outcomes include things outside your control while the decision is fully yours — and as a founder, you focus on what you can control and stick to the process. He adds the harder half of the audit: be honest when a bad decision happened to work out, which is "much, much harder" than forgiving a good decision that lost.

4. Cap the downside so you can survive to win (Alec Torelli, Shaun Gold, Aram Attar, Brian Bell, Alex McNaughten, Howard Lindzon, Sohun Sanka, Eric Jorgenson, Martin Jacobson — 9 of 16) Staying in the game is the precondition for the asymmetric payoff. Alec's version is mechanical: size the worst case first and get comfortable with it, practice bankroll management (risk 1–2% per bet), set guardrails before entering, and find the two-way door so the bet is recoverable. Shaun's version is temperamental: survivability is itself a competitive advantage — "be the mosquito" that outlasts people with more talent and resources. Aram supplies the LP variant: go into venture "with your eyes wide open" — never counting on early liquidity — and commit only money you're okay to lose. He deliberately writes small checks a zero wouldn't dent, and defends the family-office patriarch with 80% of a billion in private equity on the same arithmetic: living comfortably on the other $200M makes the risk sleeve survivable by construction (the mirror of Martin's 80% low-leverage real estate / 20% venture split). Brian's version is institutional: he encodes each loss into the screen so it can't recur. Eight gates "developed over the years from failures" sit in front of the model — if they aren't all full or partial passes, no check — and "these gates are kind of there to sort of protect the downside a bit." Alongside them runs an eleven-vector fragility score that asks where a single crack sinks the whole thesis, plus a small-check, ~100-deals-a-year cadence that means no one bet can end the fund. Alex adds the step the others assume: most people never actually map the downside, which is why it stays frightening. "I think a lot of people don't do that, and maybe it seems a lot scarier because they haven't actually mapped the downside." Once mapped, his worst case reduced to a pause in net-worth growth and a return to work he already knew how to do — and he applies an age weighting on top, since in your twenties or early thirties the same downside is "almost immaterial, really" by sheer fact of remaining time. He also kept the first outlay small by funding the ninety-day co-founder trial out of his existing company rather than out of the leap. Howard Lindzon's version is aimed at the new trader with universal access to the pipes: "Everybody's gonna make a big mistake. You may blow up your first hundred dollar account or ten thousand dollar account. The question is, what do you do next?" — because "it's making four big mistakes in a row that really gets you in trouble." His prescription is the same as Alec's, one generation on: start with money you can lose ($10 fractional shares will do), learn your risk score from the loss, then treat risk management and position sizing as the whole curriculum for the next phase. Sohun engineered his before he needed it: living at his parents' house to hold burn near zero, a total startup cost of a Claude Code subscription, Figma, Slack and about $500 for a Stripe Atlas LLC, and a re-entry job he had priced in advance — "could I reasonably get a job in two years and get paid even more than what I was making now? ... So virtually it was a zero risk decision on my part." Eric Jorgenson capped his by structure and by psychology. Structurally, the Almanack was a side project, never a leap out of his job, and the one real cash outlay was funded by a windfall (a flooded apartment and a couple of months of free rent); his portfolio rule is "relatively cheap asymmetric bets over and over," so "any one of the individual ones failing is not catastrophic," with position size growing only as results arrive. Psychologically, close relationships mean "my long-term happiness is not dependent on whether or not any individual of these bets are successful." Martin's matching structure is his barbell — 80% low-leverage real estate that throws off cash, which is what buys the freedom to take the venture risk. Martin Jacobson kept a fallback under each early risk: no college, but he could always return to cooking — "it's easy to get a job in a restaurant" — so leaving the kitchen for poker "wasn't that risky for me at the time."

5. Bet on asymmetry — upside must dwarf downside (Alec Torelli, Shaun Gold, Aram Attar, Ihar Mahaniok, Alex McNaughten, Sohun Sanka, Eric Jorgenson — 7 of 16 guests) Alec and Shaun treat it as their core model, Aram wrote it into his power-law-master profile, and Ihar prices it at the point of entry. Don't judge a bet by its probability of working; judge it by the payoff multiple. Alec: a low-probability bet with a 10x–100x payoff is worth making every time ("hunt for asymmetry"). Shaun: "I don't ask whether something is likely to work or fail — I ask whether the upside is dramatically larger than the downside. If that's the case, I'm gonna go for it." Aram's version is trait two of his five — willing to swing big — with the same lopsided math behind it: "You can just lose one X but you can miss hundred X, so why don't you go all in?" Ihar supplies the pricing corollary: the multiple is manufactured at entry, because "upside and risk-reward always takes into account the entry valuation" — valuation has been an input in every decision since. His comp work found a Silicon Valley rival with less ARR asking a higher price, PandaDoc sat roughly 5x below Instacart's valuation, and the discipline was decisive in both directions: the cheap entry returned ~150x, and the identical company at Instacart's price "I wouldn't have invested." (Three guests come close from the upside end alone and aren't counted here. Arkady's venture-scale-ambition screen exists because an outcome that can't return the fund fails him even if the company succeeds. Brian scores every deal for "power law" — "hey, if everything works out, is this like a ten billion dollar company?" — and separately scores fragility and red flags, but he never frames the two against each other as a single ratio. Kirby Winfield comes closest of the three — "our job as VCs is not to be too worried about if it doesn't work" is an explicit instruction to drop the downside term altogether — but what he weighs the upside against is a ceiling rather than a loss, so his test belongs with the promotion-focused entry below rather than here.) Alex uses the term for the decision itself — "it was a very asymmetric bet" — and both sides are concrete. Downside: net worth stops compounding for a year or two, and "I could always just come back to that," with an interesting story to tell and the option to fail upward into something else. Upside: a shot at transforming a profession with 5.7 million B2B salespeople in the US and tens of millions more globally. He is explicit that the second number "outweighed any reduction in net worth growth that would have happened otherwise." Sohun runs the same arithmetic at personal scale: a couple thousand dollars and some months of income against a business with, in his words, "this much upside," with the fallback job intact — which is why he describes going all in as "virtually a zero risk decision" rather than a brave one. Eric Jorgenson names it outright as his operating mode — "this like downside is capped but upside is not, asymmetry farming mindset" — and calls the Almanack's outcome "unbelievably asymmetric." His contribution is the discipline around it: asymmetry farming is a different headspace from Buffett-style don't-lose-money investing, and mixing the two is how you end up rationalizing bad decisions, so he batches them and reminds himself which heuristics apply to which bet.

6. Pivot, don't quit — resilience over blind grit (Shaun Gold, Aram Attar, Ihar Mahaniok, Brian Bell, Alex McNaughten, Sohun Sanka, Mahesh Narayanan — 7 of 16) Shaun's rule: change tactics as often as you need, but abandoning the game ("I'm going back to selling aluminum siding") is the only real failure mode. Aram sharpens it with Angela Duckworth's vocabulary: YC's famous finding — the one trait shared by successful founders was that they didn't quit — is grit, perseverance plus passion; but grit alone "can lead you to do the same thing always" straight into a wall. Resilience is the better trait, because it allows the pivot — and its test is whether someone can "take failure and make something out of it." Ihar underwrites the pivot before the check: tenacity was the deciding read on PandaDoc's founders ("they were creative, they were resilient, they had tenacity"), and at early stage he backs knowledgeable people over plans — "I trust founders to figure it out. If the first iteration doesn't work, you pivot and do something else in the same space" — which is why whether Cytronic built new robotic warehouses or retrofitted existing ones was, to him, "already a minutia." Brian treats the pivot as the base rate rather than the exception — somewhere between 15–20% and 50% of a YC batch ends up doing something other than what it was funded for — and derives his ranking of criteria straight from it: "that's why founders is usually the first and last thing... founders first, founders second, and founders third, because especially before you're like one or two million of ARR, you don't know if you have something." His most concrete concession to the base rate is a podcast rule: he no longer records founders below roughly $1–2M of ARR, because too many came back asking him to pull episodes describing a business they'd already pivoted away from. Alex supplies the temperament underneath the tactic, and it is close to a verbatim match for Aram's test — take failure and make something of it: "failure is inevitable... you can't win absolutely everything, but every failure you learn from," and in his life "the next thing has always been great." The poker discipline he pairs it with is the operative part — pay attention to why you are losing, to the other players and the table, and a lost hand never becomes a lost session. Sohun ran the compressed version. He had already quit when his co-founder stalled and then burned out, killing the SaaS before it existed. Instead of going back or waiting out a replacement he re-asked the question in cash terms — "what is the easiest thing I can do today that I have the best advantage on, that is highly profitable and will put cash into my pocket?" — and had a services business standing within weeks. Same thesis (TikTok Shop operators need better tooling), different vehicle. Mahesh Narayanan's Impruvon is the longest pivot chain in the series under a single founder: medication compliance for IDD clinics, then long-term and geriatric care, then — when facilities couldn't pay and the state Medicaid payer turned out to care more about the software than the box — statewide medication-compliance software that doesn't need the hardware at all. None of the final business was on the roadmap when he wrote the check; what he had underwritten was a founder who would "make this happen whether I'm there or not," and who moved within days when the market or an investor pushed. Martin supplies the test of that temperament: when one state offered to prepay a multi-year contract for something adjacent to what the company had built, "there are actually CEOs who would have said no to that deal."

7. When everyone zigs, zag to the underserved part of the workflow (Simon Lancaster, Alex McNaughten, Howard Lindzon, Sohun Sanka, Kirby Winfield — 5 of 16) Promoted from Simon's unique tier when Alex described making the same move in a different market; Howard Lindzon supplies the fintech version. All five looked at a category where capital and attention had bunched into one narrow slice of the job, and took the adjacent slice nobody was serving. Simon's zag was away from the accuracy race everyone else was running: rather than chase the hardest inference problem, he attacked the neighbouring workflow where lower accuracy already creates value, so revenue — and the answer about whether the thesis works — arrives now instead of in ten years. Alex's zag was down the funnel. The AI sales wave had concentrated almost entirely on top-of-funnel volume — "how can we send more emails? How can we do more top of funnel volume?" — and his read was that "the sales world just didn't need another email spam cannon." The hole was everything after: "from first conversation through the close through to ongoing customer relationships." A crowded category, in both readings, is not a closed one; the crowd is usually standing in one corner of it. Howard's corner was assets under management: "every VC had committed to building a better Vanguard" — Wealthfront, Betterment — because they'd been burned on trading and decided the world didn't need another E*Trade. His objection was arithmetic: even a better Vanguard isn't ten times better, and the margins on the switch are tiny. Meanwhile the trading side of the workflow was not merely unfunded but despised, which is why two broke founders with a design mockup were available at an $8–9M valuation. The Valley's collective allocation, in his reading, is a map of where not to compete. Sohun's zag was to stand between the two crowded lanes of his market. TikTok Shop has software companies on one side and agencies on the other; he built a practice straddling them — existing softwares as data layers, plus teaching brands and agencies to build their own internal tools — because his diagnosis was that the binding constraint had moved from building software to deciding which software actually returns capital. "There's actually also arbitrage on helping people make that decision." Kirby Winfield's zag was inside a problem the market had declared finished. Vehicular autonomy was "solved" in 2022 — by a decade of self-driving companies, every one of them pointed at mapped roads. The adjacent slice nobody was serving was navigation in unmapped, dynamic, off-road battlefield terrain, which needed a new class of models rather than a better version of the existing ones. His test for having found such a slice is Simon's and Alex's stated as a checklist — big enough, urgent, and greenfield, "no one else was doing it" — and he is running the play again in converged security, where the incumbents are twenty years old and not worth much, which to him is the tell that the corner is empty rather than that the market is.

8. Disrupt the old-world giant when the unlock arrives — bet the inevitability, gate on "why now" (Simon Lancaster, Ihar Mahaniok, Howard Lindzon, Sohun Sanka, Eric Jorgenson — 5 of 16) All five hunt enormous markets whose transformation they judge inevitable, then let timing make the call. Simon's version is the last-bastion test: find the major GDP sector still running on paper, Excel, and email — manufacturing, the one never digitally transformed — bet that it "cannot continue to be left behind" forever, and move when the tipping-point signals (AI-accelerated coding, a generational handover from retiring boomers) say the moment is now. Ihar names "disruption of old markets with new technology" as one of his favorite approaches and runs the same gate: e-commerce fulfillment sat at the intersection of two trillion-dollar markets he'd watched grow for years, but robotics five years earlier would have been a pass ("I would be very reluctant to invest in robotics") — he wrote the Cytronic check at pitch-deck stage in 2024 only because robotics had just crossed from research into commercial deployability, making the why-now answer "perfect." Howard Lindzon's giant was the brokerage industry, and his unlock was a latency collapse: retail got its head handed to it in 1999 working off twenty-minute-delayed quotes, and when Twitter arrived "we went from 20 minutes to zero — an exponential change in how information came." That was the why-now for StockTwits in 2008. The why-now for Robinhood was the stack of the app store, the cloud, and zero-cost customer acquisition — and the fact that regulation had frozen everyone else in place, since "if the SEC sues, you gotta shut down," so seven years after he'd pitched Twitter on a trade button, still nobody had built a mobile brokerage. He then ran the same gate one layer down: once Robinhood outgrew Apex's old plumbing, the next inevitability was "the pipes for the next 500 Robinhoods," and Alpaca was the bet. Sohun Sanka is the software-native version, aimed at the incumbents in his own category. His read: e-commerce marketing SaaS is "form to table" software — you put stuff in a form, it outputs a table — so with Claude Code and Codex anyone can mimic the logic and rebuild it, the moat is gone, and the inevitable end state is hyper-personalized internal tooling pulling from public APIs and MCPs. The why-now is the harness itself: a non-technical marketer can now build his dream software in-house, so he bet the business on teaching brands to do that rather than on selling them one more product. Eric Jorgenson supplies the curve-watcher's version. His canonical story is Steve Jobs tracking hard-drive and battery curves to know when the iPod could replace the Walkman, and his favorite pitch is the founder who says "I've been watching this threshold for ten years... we're gonna cross next year, which is why I'm starting this company now" — CRISPR tooling, uncommercialized research, or a space supply chain that turned a $100,000 in-house part into a $1,000 off-the-shelf one. He then pushes the gate one level up, like Howard did with Alpaca: once SpaceX drops launch costs by two orders of magnitude, bet on the company building the next leg of the curve — Longshot, aiming for $10–50 per kilogram with a ground-based launcher.

9. Buy the cheapest information first — run a trial marriage before the real bet (Alex McNaughten, Mike Ma, Howard Lindzon, Gregory Raiz, Mahesh Narayanan — 5 of 16) Promoted from Alex's unique tier when Mike Ma described a fund built entirely on the same move; Howard Lindzon applies it to the individual. All five refuse to make the big commitment on references and interviews when a small, cheap stretch of real work would answer the question outright. Alex, about to co-found with two people he had never worked with, put a little of his old company's money into ninety days of building for real customers — a "trial marriage" that answered "can we be in the trenches for ten years?" and whose output became the new company's seed capital; he runs the same pattern on ad channels (small budgets, read weekly, shift spend) and hires (a psychometric test before the offer). Mike institutionalized it: Sidecut works with every founder for thirty days before investing — "do you want to move your toothbrush in for a month?" — because the information that matters (how they handle ambiguity, conflict, a bad sales loop) only exists in shared work, and "information is not wine, it doesn't get better with age," so buy it before the check rather than discover it after. His own path to the fund was the same discipline one level up: a couple dozen out-of-thesis $5–10K angel checks to learn whether founders would even grant the access, before raising money around the model. And on Melagen Labs he ran it in sequence — a small angel check sized to the one variable he could observe, then the real check two years later on tracked performance. The shared logic is Martin's fifth-limper poker economics: pay a trivial price for the next card, then decide with better information on the next turn. Howard's cheapest information is a $10 brokerage account: "the only way to understand your risk profile is to lose money or make a lot of money," so open the account, play "red, green" with an amount that can't hurt you, and find out how you actually behave when you're down 30% — before you set up the financial life that depends on the answer. He's blunt that no wearable or questionnaire substitutes for it; the score "gets etched into your soul" only through a real loss. Gregory Raiz bought both kinds — the proximity and the product test. The proximity was structural: with spare space after his own exit he ran a non-equity pseudo-accelerator out of his offices, and the point of it was watching founders work, because in angel investing "the amount of information is pretty limited" while "when you see how teams are actually operating... you get true insight on how teams execute" — the same mechanism he then got at Techstars scale. The product test was cheaper still and closed the deal: unable to evaluate prop tech himself, he had the founders run their Google-form MVP against his own mortgage broker, and their rate saved him about $30,000. "And I literally said, okay, that's your angel check." Mahesh Narayanan bought his through an accelerator, like Greg: he met Impruvon's founder in April 2021 with nothing but an idea, put him in the first cohort of the neurodiversity accelerator he was helping build, and used it to push the company into unpaid pilots with health systems and long-term-care facilities — the cheapest possible test of whether the product actually changed behavior. Only after a year of watching, a compliance result nobody had seen, and the first pilot converting to paid did he write the check. Alex's ninety days, Mike's thirty, Howard's $10 account, Greg's one refinance and Mahesh's year of unpaid pilots are the same move at five price points.

10. Author your own framework — then keep versioning it (Ihar Mahaniok, Brian Bell, Kirby Winfield, Gregory Raiz — 4 of 16) Promoted from Ihar's unique tier when Brian made the same point his single closing recommendation. A framework is a personal, living instrument, not doctrine you can borrow. Ihar: "my decision framework now is different from a year ago, and obviously from fourteen years ago" — it has to evolve with the cycles (his current build: AI and robotics exclusively, growing markets, person-first underwriting). Brian's advice to anyone starting to allocate capital is to build the instrument itself rather than copy anyone's answers: work with AI on every decision because "it's going to teach you," then "start developing your own framework for things that you like and don't like and things that have been successful and unsuccessful, and you can build up your own algorithm over time." The compounding is the point — his own model is six years of overrides, gates, and retrainings deep, which is exactly why he can now say he "very, very infrequently" disagrees with it. Kirby Winfield supplies both halves — the ownership and the versioning. The ownership is why he refuses to wait and see who else is in a round: "it's my taste, it's my gut, it's my algorithm," and outsourcing it is a disservice to his LPs. The versioning is fifteen years of one specific inversion. He began by weighting market, then product, then founder — "huge market, big problem; wow, the technology solves the problem; the founder seems good, but I'm not sure I agree with how they're positioning it" — and has "completely flipped it on its head," because you are not investing in the company you would build. The current version adds a rule the early one didn't have: back technical founders so that technology is never what kills the company. Gregory Raiz is the case where authoring the framework was the bet. He went looking for someone else's: seven angel groups, then Techstars, on the reasonable theory that an organization which has seen hundreds of thousands of companies must know what predicts success. What he found was that "every managing director kind of had their own way of doing this" — all smart, none quantified — so he stopped looking and generated the data himself, surveying 3,000-plus entrepreneurs against a falsifiable hypothesis. The versioning is literal and ongoing: three years of re-running it inside Founders Edge, another batch last summer, now using AI to read hundreds of papers across tens of thousands of entrepreneurs. And the discipline underneath it is the one he offers as advice — audit each belief for provenance: "do I think this is true because other investors have told me this is true? Or because I've actually done a little bit of digging to verify?"

11. See what can go right — decide promotion-focused, not prevention-focused (Alec Torelli, Aram Attar, Kirby Winfield, Eric Jorgenson — 4 of 16) Alec practices it as a discipline: most people size the risk, freeze, and never get to the reward, so force yourself to dream the upside — enumerating everything that could go right (the travel, the heroes, the freedom) is what green-lit the poker bet. Aram found the same thing at the top of the power law and made it both trait one of his masters profile ("they tend to see what can go right — they're not focusing on what can go wrong") and step three of his decision process: when the decision finally gets made, "they don't really care about the loss, they're trying to win. They're not trying not to lose" — okay to bet the house on their own decisions rather than trying to preserve something. Kirby Winfield states it as a job description rather than a discipline: "our job as VCs is not to be too worried about if it doesn't work. If it does work, can it be the next Anduril?" The first qualifier he runs on any idea — "if it works, does it matter?" — is a ceiling question with no floor term in it at all. And he pushes the posture a step past Alec and Aram by inverting the sign on his own discomfort: he wants to feel sick to his stomach writing the check, because a bet that feels obvious has already been priced by everybody else. Eric Jorgenson's diligence starts from the same place as Kirby's: not what could go wrong, but "what could be a ceiling" — a founder who takes the $50–100 million offer, or a market too small or competitive to support more than a few-hundred-million-dollar company — and he looks for people and markets with "unlimited upside." His temperamental version is Alec's: have fun with each bet, "be optimistic and move fast and move on to the next."

12. Backtest the thesis against history — "would our screen have caught it?" (Simon Lancaster, Brian Bell, Gregory Raiz — 3 of 16) Promoted from Simon's unique tier when Brian described running the same loop continuously. Both refuse to let a framework go unaudited: you take the outcomes history already produced and ask whether your own screen would have found them. Simon regressed roughly thirty years of relevant exits — every one in his sector — and back-solved fund size and required entry ownership from the result, filtering each exit by whether his thesis would have caught it in the first place. Brian runs it as a standing retraining loop in both directions: on a winner ("we just raised this massive $650 million Series B... I got in at a five cap, so I'm sitting on a hundred X") he replays the deal as of 2022 or 2023 with only the information available then, sees the model scored it a 3.8, and asks how to catch more of those; on the shutdowns, the failures become the eight gates. He has the raw material because he keeps the whole history — YC's ~5,000 alumni plus the 40,000–60,000 companies he's personally seen in six years — and he scores himself with it: "you can actually calculate your investor F1 score over time" from true and false positives and negatives, "and see how good you personally are, and how good your AI is." Gregory Raiz ran the loop before he had a portfolio to run it on. His hypothesis was that questions exist whose answers predict founder success, and the only way to test it was against outcomes that had already happened: he surveyed more than 3,000 entrepreneurs, roughly a tenth of them carrying seven-, eight-, or nine-figure exits, and asked which answers separated that decile from the rest. Where Simon back-solved fund mechanics from thirty years of exits and Brian replays his own deals as of their original date, Greg fits the screen itself to the historical record — and keeps refitting it, which is how he learned that prestigious schooling predicts getting funded rather than succeeding, and that a serial founder outside their own domain performs like the average.

13. Underwrite venture-scale ambition as its own line item (Arkady Kulik, Kirby Winfield, Eric Jorgenson — 3 of 16) Promoted from Arkady's unique tier when Kirby described running the same test on academic founders. All three refuse to let a great entrepreneur stand in for a great venture-scale entrepreneur. Arkady's version is arithmetic: a founder content with a lifestyle outcome cannot return your fund, and you end up with a fraction of a fraction — so ambition is a graded criterion, not an assumption you inherit from the pitch. Kirby's version is the lab spin-out conversation, and it is the one he says decided Overland: building technology inside academia is very different from raising venture and scaling a multi-billion-dollar company, so he spent the diligence asking Byron Boots what he wanted in life — was this innovation for innovation's sake? Something to hand off? Or a company, potentially a generational one? Martin runs the identical gate from the other side and reports that it is where most spin-outs die: "those guys frankly just wanna be more researchers." Kirby's extension is that ambition has to be underwritten together with fundability — the founder must also project to be able to raise from tier-one venture, because a $5–10B company will need that money whether or not the founder wants the meetings. Eric Jorgenson names the same failure as a founder ceiling: the founder who "gets an offer for 50 million or a hundred million dollars and takes it rather than persevering through to try to build a billion dollar or a ten billion dollar company" has capped your return no matter how good the company is — so he screens for that trait directly, alongside the market's own ceiling.

14. "Don't tell me, show me" — weight demonstrated work over persuasion (Mike Ma, Gregory Raiz, Mahesh Narayanan — 3 of 16) Promoted from Mike's unique tier when Greg made the same demand the center of his screen; Mahesh Narayanan extends it to science. Mike and Greg start from the observation that founders are selected and trained for persuasiveness, so the words are the least reliable artifact in the room. Mike's response is temporal — watch them work with hands and feet across thirty days, because after the wire you'll see maybe one percent of what they do, and self-awareness only shows up in behavior. Greg's is artifactual, and aimed squarely at the deck: pitch decks are "kind of BS... a little bit of startup theater," so he asks founders to show him something they built. An exceptional founder answers instantly — here's the repo, here's the site, here's the company I sold — and the ones who struggle have told him something about execution that no deck would have. The asymmetry is what makes it decisive for him: "we can show you how to build a pitch deck. It's really hard to show you how to build and scale a company." His first angel check was the same test run on himself rather than on a founder — don't tell me the rate is better, beat my broker's rate — and Martin's summary of it is the entry in one line: "don't tell me, show me, and they showed you a thirty thousand dollar thing." Mahesh Narayanan makes the same demand at two scales. On Impruvon, the pitch was an idea in the founder's head; what turned it into "an investable idea" was demonstrated behavior change — compliance moving from 30–40% to 80–90% in populations that resist medication — followed by a customer converting from an unpaid pilot to a paid one. In deep tech he pushes the standard outside the company altogether, because "we've done it" isn't evidence until someone else has done it too: "No one else has done it. So I can't believe you yet. Someone else needs to validate this." Replication, publication, multiple animal models and GMP manufacturing are the science version of showing the repo.

15. Ask what it plugs into next — check the market ceiling before you fall for the founder (Kirby Winfield, Eric Jorgenson, Mahesh Narayanan — 3 of 16) Promoted from Kirby's and Eric's unique tiers when Mahesh made the same test one of his three first-meeting questions. All three run a market-ceiling check that is separate from the founder read, because a great founder in a capped market still can't return the fund. Kirby's is the first qualifier on any idea — "if it works, does it matter?" — asked before market sizing and on the assumption that you already love the founder: is there a ceiling on this thing at all? Eric names two ceilings and screens for both: the founder ceiling (taking the $50–100 million offer, covered under venture-scale ambition above) and the market ceiling, where size and competition cap the outcome at $100–500 million and the switch from growth to earnings multiples caps the value; he backs only what has no ceiling, or orders of magnitude more headroom than his alternatives. Mahesh adds the beachhead version. Impruvon's first market, IDD clinics, was real but tiny — "a great idea, we can build it for a small population and that's it" — and what made it investable was that the same product plugged into long-term and geriatric care, and later into state Medicaid programs. His rule is explicit: it doesn't have to be a platform technology, but if he can't see it plugging into another market, "oftentimes that means the market's just too niche," and he stops looking.

16. Blockers first, weights second — run a two-stage IC (Arkady Kulik, Brian Bell — 2 of 16) Promoted from Arkady's unique tier when Brian described the identical architecture. Both refuse to let a good score outvote a disqualifier, so scoring never begins until the gates are cleared. Arkady's IC1 hunts roughly eighteen blocker questions (team-heaviest); only if none fire does IC2 score the deal against fixed percentages — team 23%, science/tech/product 28%, competition 15%, strategy & round 15%, traction 11%, market 8%. Brian's model runs eight gates distilled from his own shutdowns — every one must be a full or partial pass or there's no investment — in front of a roughly twenty-feature weighted scorecard (friction, founders, moat, traction, problem/category-defining, and the rest). The shared insight is sequencing: weights answer "how good is this?", gates answer "is this disqualified?", and asking them in the wrong order is how a beautiful deal with one fatal flaw gets funded.

17. Feature or platform? Price the incumbent's bundling risk (Ihar Mahaniok, Brian Bell — 2 of 16) Promoted from Ihar's unique tier when Brian named the same variable in his model. Both ask whether the company is a product or a feature inside somebody else's product — and both think most investors point the fear in the wrong direction. Brian: "is this a feature in somebody else's platform?" is something "to really look at twice," and bundling risk of the incumbents is an explicit feature in his scorecard — the same doubt, he suspects, that made people pass on Google ("nice demo, kids... Yahoo is a billion dollar company and they'll just integrate that feature into their catalog of the internet"). Ihar sharpens where the real danger lies: the deliberate clone is mostly a phantom, because big companies don't chase small things (Adobe moved on e-signature long after DocuSign had escape velocity); the mortal risk is incidental coverage — if ChatGPT already gives away the answers your software sells, nobody buys it. Martin supplies the counterexample that keeps the test honest: DocuSign was hard to fund precisely because it looked like a feature Adobe could copy in five seconds, and the return came from a founder who did the years of unglamorous work — signing order, email workflow, integrations — to turn a feature into a platform. So the question isn't "is it a feature today?" but "is there a credible path to a control point — a system of record or system of action — before an incumbent absorbs it?"

18. Confidence compounds from closed loops — count your completed reps (Alex McNaughten, Eric Jorgenson — 2 of 16) Promoted from Alex's unique tier when Eric described the same mechanism in almost the same words. Both read their own confidence as the output of completed attempts, not as evidence about the bet in front of them. Alex's version is a before-and-after: the same leap felt enormous at company one, when he had a mortgage and about ten thousand dollars left, and comparatively small at company three — "when you just have a pattern of things going right, the confidence just compounds." Eric's is a maze: writing a book is full of "creative crossroads," but "navigating that maze all the way through one time kind of gives you the sense that you're gonna make it through and it makes you more confident entering the maze again" — which is why the Almanack made the Balaji and Elon books possible. He adds the customer version ("when you go from zero happy customers to one happy customer, there's a world of difference") and the general rule behind his risk appetite: closing the loop many times early builds the conviction that you can figure things out. The practical implication for both is to go looking for small, completable loops early, and to discount confidence that hasn't been earned by one.

19. Back the founder who'll build it with or without you (Ihar Mahaniok, Mahesh Narayanan — 2 of 16) Promoted from Ihar's unique tier when Mahesh described the identical test from the motivation side. Both refuse to be load-bearing. Ihar: "I only invest in a startup where I believe the startup doesn't need me — the startup will be successful anyway"; investor help is a 10%-faster, 10%-bigger accelerant, never a wall holding the company up. Mahesh gets there through motivation. He looks first at how close the founder is to the problem, then digs into "what drives them, what motivates them, what makes them up in the morning" — and it can't be money or fame, because "those are things that are gonna go away overnight for most founders." Impruvon's founder had lost his brother to a medication overdose, and from the first meeting Mahesh "knew he was gonna make this happen whether I'm there or not, or anyone else is there or not." That turns the investment into a simpler decision — "do I jump on for the ride or do I sit in the sidelines and watch him?" — and his screening question is the sharpest form of Ihar's rule: "how much do I believe that this founder is not going to give up on this idea, even if I do."

20. Coachability is a round trip — give the feedback, then time the response (Mike Ma, Mahesh Narayanan — 2 of 16) New shared entry: Mahesh named coachability as one of his three screens, and Mike Ma's Melagen Labs check was written on the same test. Neither wants obedience; both want to see what a founder does with feedback, and how fast. Mike told an eighteen-year-old founder that his LOIs carried no counterparty risk, expected that to be the end of it, and wrote a small check when the founder came back a month later with every LOI renegotiated. Mahesh: founders "don't have to agree with everything I say," but they have to listen — "let me think about it and then let me come back with something better." His clock is about a week: point out a problem, and if the founder doesn't return with "this is not a problem" or a couple of solutions, "that's a deal breaker," because it means the feedback was never discussed internally, just dismissed. Impruvon's founder set the high-water mark — told a role wasn't his strength, "within a day or two he'd come back with five candidates." Martin is "a little bit on the fence on coachability," since some of his best founders have ignored his advice entirely and crushed it anyway, so his version replaces obedience with curiosity: he wants "a curiosity for my comments and a respect for being challenged," because a founder who answers a hard question with "that's stupid" can't sustain a ten-year relationship. "The founders that are uncurious about things, I have a real problem with." (Arkady's gate sits at the far end of the same axis: a founder so delusional he hears no feedback is a pass.)

21. Regulatory is a variable, not a veto — price the path, not the label (Arkady Kulik, Mahesh Narayanan — 2 of 16) Promoted from Arkady's unique tier when Mahesh described pricing regulatory exposure the same way. Both refuse to let "FDA" or "regulated" function as a reflexive no; they ask what the path will actually cost and how the founder plans to walk it. Arkady treats FDA as a given and grades how the founder thinks about it — and warns that the bigger hidden risk is reimbursement codes, since out-of-pocket revenue is abysmal next to insurance. Mahesh prices the path by frequency: one 510(k) submission and done is a very different bet from a company that has to go back to the FDA every year, because "the more you interact with them, the more expensive your trials are gonna be" — meet them three times and the cost can be thirty times. Impruvon shows the practical side. Its regulatory work was state-by-state approval of the device, which took the founder eight to twelve months to crack alone before he hired someone who knew the process; its HIPAA exposure was handled by storing data with a provider that was already compliant, so the company never had to certify itself. And Arkady's reimbursement point turned out to be the hinge of Impruvon's pivot: facilities couldn't afford the product, the founder went to CMS to get it covered, and the payer wanted something bigger than the box.

So far unique to one guest

• Be a risk trader, not a risk taker — reframe risk as an exchange for reward, forcing you to weigh both sides like an EV calculation. (Alec Torelli)
• Mind the expiry — prioritize decisions with closing windows (options) over ones you can hold indefinitely. (Alec Torelli)
• Watch your linguistics — words like "should," "can't," and "I am X" program how you experience the decision. (Alec Torelli)
• "Everyone's a gangster until they gotta wire the money" — discount all soft commitments to zero; nothing is real until the wire hits, so plan as if you're alone. (Shaun Gold)
• Your word as tiebreaker — when analysis is murky, "I said I would do it" carries the decision and blocks rationalized retreat. (Shaun Gold)
• Proof of ROI over proof of concept — impatient old-industry buyers won't buy concepts, but a short-payback deployment is a must-try; target 2–3-year commercialization, never 10. (Simon Lancaster)
• Let the fundraise double as customer discovery — pitch operators and watch which personas "get it quickest"; for him, the incoming generation replacing retiring boomers proved the pain was real. (Simon Lancaster)
• Price the pivot optionality — software can pivot almost without limit (video game → Slack); deep tech barely at all (solar → nuclear = bankrupt one company, start another), so weight technology risk in inverse proportion to the escape routes. (Arkady Kulik)
• "Just talk to people" — communication as diligence and as asset — resolve interpersonal static by naming it to the founder's face; trust is what surfaces the problems founders are ashamed to raise, and it converts into secondaries and preferred liquidity later. (Arkady Kulik)
• Ask where they lie to themselves — knowing lies: never; deaf to feedback: pass; delusional about physics (single-drop blood tests, tabletop fusion): pass; delusional only about how big they can become, plus a beginner's mind: "the perfect combination." (Arkady Kulik)
• Map the science — it moves slowly, so the map compounds — unlike software, what was true of a scientific field ten years ago mostly still holds; corpus-reading and expert networks are durable assets, and competition is knowable but broader than direct rivals (for a brain-inflammation device: pharma, surgery, even diet). (Arkady Kulik)
• Never take VC advice at face value — VC is a wildly dispersed industry, not a homogeneous one; what works for Series B fintech is wrong for pre-seed pharma, so filter every rule through your stage and sector. (Arkady Kulik)
• Pause the intuition — it's a hypothesis, not a verdict — the three-step power-law-master process: have the intuition, delay it, collect data scientifically to disprove it (confirmation bias is the enemy), and only then decide. (Aram Attar)
• Trust intuition only where rules and fast feedback exist — Kahneman's conditions: poker and chess qualify; VC has deal volume but no rules and years-long feedback, so "pattern matching" is recognition masquerading as a mystical gift. (Aram Attar)
• Track record is a broken compass — persistence is unconfirmed (~30% repeat rate for top-quartile funds — worse than a coin flip) while half of each year's top-ten funds are Fund I/IIs; the metric everyone screens on locks them out of the alpha pocket. (Aram Attar)
• Evaluate emerging GPs as entrepreneurs first, investors second — are they talking to founders and testing the thesis, or paying lawyers and touring LP resorts? A deliberately modest "test myself" fund size is a groundedness tell. (Aram Attar)
• Judge character by how they decide — Aristotle's test — skip the data room; watch interviews, join the IC, ask why they picked each founder (first principles or FOMO?), and prize people who audit their own decisions. (Aram Attar)
• Don't manufacture conviction — bet the framework — chasing high conviction is usually self-persuasion; "I believe in my framework and I'm willing to put money behind it" is enough to act. (Aram Attar)
• Ask the category-switcher "why are you here?" — pause the pattern-match and hunt for the non-consensus insight; "I just prefer it" fails, a real angle marks the unfair advantage every other LP screened out. (Aram Attar)
• Test the spin-out halo: brand or person? — pedigree only counts if the success travels; the hard signal is founders following the GP out of the old firm and taking their money again. (Aram Attar)
• The five-trait screen for future power law masters — sees what can go right; swings big; okay being wrong (low loss aversion); unafraid to be contrarian; high tolerance for risk, uncertainty, and ambiguity — published in advance so his reputation rides on the calls. (Aram Attar)
• Demand the core loop proven at tiny scale — one San Francisco neighborhood and one store showed Instacart could both sign the supply side (Trader Joe's) and deliver the demand side (real customers ordering); PandaDoc's version was $50–100K of ARR someone already paid. The mechanism must spin once for real before the check. (Ihar Mahaniok)
• Know which vetting to rent and which to run — he trusted YC's founder-interviewing on Instacart ("YC is actually very good in interviewing the team"), with Khosla's lead commitment as the trigger; with no accelerator behind PandaDoc, he reference-checked through mutual friends and made his own read. (Ihar Mahaniok)
• The talk-shop test for technical moat — never reviews code pre-investment; instead verifies a proper CTO with real credentials who can talk shop, building something a cloner would need six months to a year to copy — while the MBA with landing pages, a spreadsheet, and no technologists is an auto-pass ("he will not" build a unicorn). (Ihar Mahaniok)
• One clear leader — "pitch me yourself" before the company — whatever the co-founder count, he's underwriting the one CEO who makes the company happen (no single leader, no check); he opens every meeting with "tell me about yourself before this company" and wants to be impressed enough to back anything they build — before hearing what they're building. (Ihar Mahaniok)
• The "out of advice" test — back founders who know their market so in-and-out that "I feel like I'm out of advice to you... I'm just happy to be along for the ride" — with one gate he keeps for himself: the market must be growing, never shrinking or capped. (Ihar Mahaniok)
• Pace the fund to keep the bar — about one deal a month (two only for the unbelievable); five good deals in one week get stack-ranked down to a single pick, because every "why not" is almost always "some other deal was better." (Ihar Mahaniok)
• Fish from someone else's filter — YC takes 150–200 of ~20,000 applications a batch, so the selection work is already done and free; even darts thrown inside that pool beat careful picking outside it, and the exposure recalibrates your bar everywhere else (post-YC, non-YC decks score 1.5–1.6 where YC ranges ~2.0–4.3). (Brian Bell)
• Match your capital structure to the decision clock — syndicates kept losing races: rounds closing weeks before demo day, a $20M cap becoming $30M mid-raise, founders refusing distribution to 16,000 people. The fix wasn't better picking, it was a small dedicated fund that can commit the same day. (Brian Bell)
• Underwrite as feature extraction — and score 1 to 5, not 0 to 1 — ~20 weighted features, weights fitted by the model, ratings he overrides by hand; the scale is human on purpose, because "I don't think of like a point nine startup versus a point six startup. It's better to think of like, this startup's a four and a half versus this startup's a two and a half." (Brian Bell)
• Stack-rank the whole batch to allocate attention — ingest all ~200 companies four times a year, rank best to worst, and let the ranking sequence his time; he still reviews everything manually, but ~2.2 is an automatic pass and above ~3.5 earns a meeting. (Brian Bell)
• The eleven-vector fragility score — where is the thesis a single point of failure? Founder fragility (solo founder, or no complementary business/technical pair), market fragility (the bet needs the market to turn one specific way), product and capability fragility (depends on unproven research or hard physics with no IP), GTM fragility (one motion, one partner, one platform). (Brian Bell)
• Score the deal four separate ways — a power-law score (if everything works, is this a $10B company?), the weighted scorecard, the fragility score, and a red/yellow-flag score. One blended number hides which kind of problem you have. (Brian Bell)
• Read the second-order metrics against the cohort — capital-inefficient growth, or "yeah, they're growing 100% a month, but did you notice that their logo retention is 67% and their net revenue retention is 80%?" The number only means something next to the dataset: always ask what excellence looks like in this comparison set. (Brian Bell)
• Prompt for truth, not validation — early models told him to invest in everything, so he borrowed a Marc Andreessen custom instruction built around truth and accuracy and "don't tell me what you think I want to hear." The result "almost painfully tells me what I don't want to hear" — the point of a thought partner is that it overrides you as often as you override it. (Brian Bell)
• Dump the data room in and let the machine cross-examine — it flags where the deck contradicts the call: "they said this in the call, but I checked that and that doesn't match." Mostly founders stretching, but about 1% of the time an outright lie about a credential. (Brian Bell)
• The Nasdaq-bell-ringer test for star quality — "do you envision this person being this Nasdaq bell ringer in ten years? Do you see them being interviewed on CNBC?" It's not vanity: a founder has to recruit A players out of comfortable big-tech jobs, and star quality is recruiting ability. (Brian Bell)
• Timing is its own variable — early, on time, or late — being right and early is indistinguishable from being wrong for years; Google was the seventeenth search engine to show up, and it won anyway. (Brian Bell)
• Ask for the ten-year vision and listen past the wedge — "five or ten years out, you've built this thing and it's really successful. What does the future look like?" You're listening for a vision far bigger than the initial wedge — the aspiration to "make a huge dent in the universe." (Brian Bell)
• Name the business model before you underwrite it — is there a differentiated wedge, a path to a control point, to a system of action or system of record? Which trajectory is it on — distribution-first, research-first, something else? You can't judge how a company compounds until you've said out loud what kind of company it is. (Brian Bell)
• Human + AI beats either alone — so measure both — "AI is not gonna replace venture capitalists... but a VC powered by AI is a very powerful thing," and the research on hybrids backs it; he ends most disagreements with his model in a consensus, sometimes overriding it, sometimes overridden. (Brian Bell)
• Use psychometrics as cheap pre-hire information — a test costs almost nothing and buys a whole extra dimension on a candidate. One profile flagged a red flag on an otherwise flawless finalist — trusted introduction, excellent interviews — that read as someone who would not stay through the hard stretch; digging into that area in the next conversation confirmed the wrong fit for an early-stage company, "where, let's be honest, you're shoveling shit for a fair amount of the time." (Alex McNaughten)
• Play the cards actually in front of you — the CEO version of the poker frame: what you hold is fixed, so the only question is how you play it. Raising with low revenue is the card you were dealt; a crowded market means the table is full and everyone's in the pot. "You're always gonna have limited information, but you just try and play it the best way you can... and more information's gonna come on the next turn." (Alex McNaughten)
• Take the people along the journey — making the decision is not the same as landing it. His partner's job, friends and family moved with him; his co-founders had to run the same conversation at home. The work is walking them through why, what it will look like, and what's in front of you — "it's gonna make it a lot easier." (Alex McNaughten)
• Hire salespeople who have sold without an engine behind them — a seller from Google carries Google's marketing machine, brand and support; a seed-stage seller has a half-built product and no marketing budget. "It's a very different kind of person who could be successful doing the latter." (Martin's harder-won version: only hire people whose first startup is behind them — his first big-company hire walked in on day one asking where his assistant was.) (Alex McNaughten)
• Coach first, capital second — do the value-add before the wire — the first two meetings are standard; then, instead of references and data rooms, thirty days of shared work on customer discovery, funnel, or PLG, on the fund's own time. "At the end of 30 days, you like us, we like you, we'll write you a check." (Mike Ma)
• Know more than the rest of the cap table — the selfish target for information: not "more than the market" but more than the other investors in this deal, which only shared work can produce. (Mike Ma)
• Past signal has a half-life — "I don't want to cook with frozen vegetables" — references, decks, and LinkedIn describe the last restaurant, not this one; a Michelin star gets you in the door and guarantees nothing. Sharpest form: founder muscle memory from a pre-AI world may be negative signal. (Mike Ma)
• The deck is precisely what will not happen — "it's the only thing I know" at pre-seed, so the study object is what the founder does when things go sideways. (Mike Ma)
• Honesty that lowers the number raises the conviction — a founder who cuts his own pipeline in half and explains every cut is more investable than one with a "bajillion dollars" of unqualified deals, because it proves the next eight years of reporting can be trusted. And if you really had a bajillion in pipeline, why are you asking for a quarter million? (Mike Ma)
• Killers for good — the one non-negotiable is go-to-market killers; "good" is the impact. No conviction on killer after thirty days, no check — regardless of the technology. (Mike Ma)
• "You may be right — I can't underwrite faith" — a "just trust us" commercialization path from a corporate sponsor isn't priceable, however strong the team's scientific conviction; pass in a long, coaching-style letter and let the round oversubscribe without you. (Mike Ma)
• Pick one motion — you're not funded for both — local-and-fast or Fortune 500, not both at a pre-seed raise; and hold them to what the chosen path really costs (for enterprise: live in the customer's cafeteria until people think you work there). (Mike Ma)
• Test the counterparty risk in every LOI — no acceptance criteria, no theoretical payment, no timeline means the counterparty is risking nothing; send the founder back for better paper and time how fast they return. (Mike Ma)
• Keep the bar after the sunk cost — pass about forty percent after thirty days — no metric decides it, "it just comes in," and a month of effort never argues for the check. (Mike Ma)
• Want both the moonshot and the pragmatist — asked to choose between the audacious founder and the under-promise-over-deliver one: "Yes." Build the portfolio to hold both — some bets de-risked at entry price with shorter horizons, some to the moon. (Mike Ma)
• Take the seat with the best view of the system — in a category too new to have a right answer, optimize the next decision: go where every business in the market passes through and the back end is visible. The nexus seat beats the bigger title, because it is the only place the system becomes legible. (Sohun Sanka)
• Move when the moat moves, not when the job gets worse — he left a compounding comp package because coding harnesses had made "form to table" software trivially cloneable. The trigger to change position is a change in what is defensible, which arrives long before the paycheck notices. (Sohun Sanka)
• Abstract the job, then price only the residual — podcast-interviewing founders plus being employee number one (drafting NDAs with lawyers, sitting in on hires outside his department, sourcing leads with no paid budget) meant he had already done most of the founder job, so he inverted the question: what would I have to do that I'm not already doing? The list came back short — invoicing, legal, accounting, forming an LLC — and every item had a price tag (~$500/month back office, $500 Stripe Atlas). (Sohun Sanka)
• Interrogate every "you need X" — ask which constraint it was solving — the co-founder rule exists because software is hard enough that focus splits across product, design, dev management, marketing and sales. If building gets easy and the buyer's decision becomes the bottleneck, that's a GTM constraint and the rule dissolves. Received requirements are compressed answers to old constraints; decompress before obeying. (Sohun Sanka)
• Take the shot inside the low-obligation window — "if I have a spouse in the future, if I have a mortgage, if I have other obligations to other people, those things are going to affect my likelihood to succeed." Obligations don't just change your tolerance for risk, they change the odds — so the window itself is a depreciating asset. (Sohun Sanka)
• Bank latent demand before you need it — and notice whose brand the leads come from — he brought leads to seed-stage employers off his own LinkedIn following rather than theirs ("that's why these companies are giving me equity"). Demand you generated personally is portable, and its portability is the tell that you are already carrying founder risk. (Sohun Sanka)
• Don't delay a year to resolve what only time can reveal — he couldn't judge a prospective co-founder's technical acumen at face value; it had taken ten months inside a company to see what separated one CTO from another. So he started and let reality answer it, which it did in weeks. (Sohun Sanka)
• When the plan dies, ask the cash question — not what is most ambitious, defensible or fundable, but "the easiest thing I can do today that I have the best advantage on, that is highly profitable and will put cash into my pocket." It converts a collapse into a next move fast, and cash flow buys the right to be ambitious later. (Sohun Sanka)
• Put 80% of your time on the single constraint — and do it by hand — if the constraint is sales, that's fifty-five hours a week of it, even if it's just DMing everyone Claude says fits the persona. The failure mode that looks like work: "because things can be automated, people spend too much time looking for ways to automate things, or if it's not automatable, they kick the can down the road." (Sohun Sanka)
• The eighteen-hour filter — ignore anything that doesn't move "more customers" or "happier customers" within one order of magnitude in the next eighteen hours; creator deals, referral streams and advisory roles all fail it, because "even just one to two hours away from this can completely derail momentum in one direction." (Sohun Sanka)
• Anchor to a stack of proof, not a feeling — when the "can you really do this" doubt hits, look tangibly at the pipeline and the shipped content, then ask what has materially changed: has the algorithm shifted, have five better competitors entered and targeted my exact customers? "Usually that isn't the case." Certainty as compounded reinforcement, which is also why he favors work that returns feedback in weeks. (Sohun Sanka)
• Do I feel sick to my stomach? Treat discomfort as the entry signal — you want to feel nervous; an obvious bet "just feels too easy." The specific worry to want is not "will they execute" but "could this not be backable in two or three years?" (Kirby Winfield)
• Fund enough powder to turn the cards over — then ask whether aces are enough — the round has to flip the specific cards that answer the thesis, and the half people skip is whether that result actually earns the next round. (Kirby Winfield)
• Clear milestones are easier to underwrite than blurry ones — software-enabled hardware for defense behaves more like life science than software: tranched and gated. An investor trained on blurry SaaS milestones should find the explicit kind comforting, not intimidating. (Kirby Winfield)
• Underwrite research risk, not execution risk — back technical founders so that the product almost always ships and the technology is never what kills the company; what's allowed to kill it is being too early, missing problem/solution fit, or a problem smaller or less urgent than believed. (Kirby Winfield)
• The academic-founder screen: ask what they want in life — innovation for its own sake, something to hand off, or a generational company? Then four properties: big enough vision, some secret knowledge, talent magnet, and projects to raise from tier-one venture. (Kirby Winfield)
• You're investing in the company they'll build, not the one you'd build — disagreeing with a founder's positioning or go-to-market is a category error, not a red flag; once you accept that, "really you're just investing in the founder." (Kirby Winfield)
• Multi-stage participation at pre-seed is an optionality check, not signal — "they'll tell you that the partner is gonna be on the deal, and then three months in it's the associate." The corollary is expectation management: a tier-one Series A runs the same venture math you do — twenty checks, two projected to return the fund. (Kirby Winfield)
• Decide naked — don't wait to see who else is in — get the check in on the note without knowing the rest of the round, because needing to know means you shouldn't be investing. It's hard the morning after the wire, and sometimes it's YC two days later at 3x and sometimes it's a couple of angels and a lot of work. (Kirby Winfield)
• Outside the Valley, in AI, do your own competitive diligence — if you aren't one of the first three or four funded teams in a category, it's a capital-raising problem however good the team is ("you could be better than the other teams. It doesn't matter, half the time") — and the founder will always tell you nobody else is doing it. (Kirby Winfield)
• Be smart about the founder, not about the field — "it's not my job to be smart about what you're doing, it's my job to be smart about you and your capacity to build something meaningful." His anti-portfolio is deals he talked himself out of for lack of subject-matter expertise. (Kirby Winfield)
• Want the ball on a cold streak — the inception-stage temperament test: "a jump shooter on a cold streak who wants the ball at the end of the game with the game tied" — built to get over losses fast, driven by big wins; he doubts it scales to growth stage. (Kirby Winfield)
• Become the customer — make them prove the value on you — when you can't evaluate the domain, convert the claim into a free experiment you can run yourself. He put the founders' Google-form MVP head-to-head with his own mortgage broker; their rate saved him ~$30,000, and that was the diligence. (Gregory Raiz)
• Quantify first, pattern-match second — "humans are really bad at pattern matching, and pattern matching leads to bias," so on cold inbound the founder gets scored on data before anyone forms an impression; judgment comes after the quantified read, not before it. (Gregory Raiz)
• Founder-market fit and reps, measured together — a third-time founder in the same vertical is strongly predictive of another win; the same founder in a new domain is "same as average." And attempts aren't wins: three companies with no wins is weaker signal than three companies with several. (Gregory Raiz)
• Prestige predicts funding, not returns — "school is predictive of funding, not of success." The prestigious degree makes a founder more fundable because it is a bias in the ecosystem, not because it predicts higher returns. (Gregory Raiz)
• The metrics question as an execution proxy — a graded answer to a trivial question: I don't track metrics / I have a spreadsheet / I use a third-party tool / I built my own tool. The more sophisticated the answer, the more likely the team is building rather than talking about building. (Gregory Raiz)
• The ante is build execution — the charismatic CEO not paired with the technical talent to ship is his most expensive repeated mistake: "I can probably name seven or eight companies that just went belly up because they couldn't recruit the technical talent to build." It's a disqualifier, not a weighting. (Gregory Raiz)
• Ask whether the co-founder relationship has been tested — the survey question was has your co-founder let you down?, and a yes reads positive: a strained relationship that proved resilient is stronger than an untested one ("we met at co-founder matching and built a prototype over a weekend"). Twenty to thirty percent of startups die of founder conflict. (Gregory Raiz)
• Count the co-founders — convincing smart people to work unpaid is the proof — the data runs nearly linear toward more co-founders being better, provided the team is strong. The mechanism isn't headcount but evidence: early co-founders aren't paid much, so four or five very smart people spending their limited life resources on your company says something about the company and about you. (Gregory Raiz)
• Back a solo founder only on proven reps — one solo deal in the portfolio, a founder with eight prior companies and two successes. Martin's gate is the same asked differently: why are you solo? Unless they've already proven they're a talent magnet, the recruiting problem is the scaling problem. (Gregory Raiz)
• Question whether you believe it or were told it — "do I think this is true because other investors have told me this is true? Or because I've actually done a little bit of digging to verify?" Treat received investing wisdom as a hypothesis with an owner you can't see. (Gregory Raiz)
• Curiosity is the flywheel, not the hobby — he codes his own models and tinkers with circuit boards because it's fun ("a kid again playing with Legos, only instead of Lego bricks it's LLMs"), and the return is indirect and compounding: curiosity creates deeper founder relationships, founders refer founders, deal flow follows. (Gregory Raiz)
• "Confidence" may be the wrong word — "what's next" is the engine — the successful CEOs he talks to don't perceive themselves as having cleared the bar, "they're just executing." What moves him is the next rung of thought, process, data or team, not a settled belief that he'll win. Martin's version: leaving Microsoft wasn't confidence, it was curiosity about a category he couldn't chase from the inside. (Gregory Raiz)
• Write for an audience of one — when the market's response is unknowable, pick the one judge whose opinion matters and imagine them looking over your shoulder; it sets a quality bar no forecast would. (Eric Jorgenson)
• Define success as something you control — break even and be proud of it; do the thing you'd do for its own sake, because "the process was going to be its own reward." (Eric Jorgenson)
• Put the wood behind one arrow — the value of moving from the top 1% to the top 0.1% beats "firing another lead bullet"; one video with 10 million views is easier than ten thousand with a thousand each. (Eric Jorgenson)
• The sensible-trajectory test — is this company the natural next chapter of the founder's life, or the latest stop in drop shipping → crypto → AI? (Eric Jorgenson)
• Back the company moving the curve — then ask about its next leg — once a cost curve drops, the obvious bets are downstream of it; the bigger one is the company that bends it again (Longshot: ~$500/kg → $10–50/kg). (Eric Jorgenson)
• Don't mix decision modes — asymmetry farming and don't-lose-money value investing are different headspaces; batch them and write down which heuristics belong to which bet, or you'll rationalize. (Eric Jorgenson)
• Price the full human cost of an investment — maintenance, board seats, yearly buy/sell decisions; is the payoff meaningful relative to your attention threshold? (His post "Opportunity Cost and Attention Thresholds.") (Eric Jorgenson)
• Grow your psychology with your portfolio — get desensitized to what used to be life-or-death scary, scale position size as results arrive, or you'll abandon a working strategy just because the numbers got bigger. (Eric Jorgenson)
• A secure base funds the risk — "my long-term happiness is not dependent on whether or not any individual of these bets are successful." (Eric Jorgenson)
• One rare early win is a turbocharger — a book, a PhD, an early exit, a rocket-ship résumé line changes who will meet you and how fast; it's worth making bets that could become that thing. (Eric Jorgenson)
• Re-weight the scorecard by where the risk lives — on software the moat is execution, so the founder carries about 70% of the decision; on hard life sciences it's roughly 30% team, 30% IP, 40% data — still people-first, but now the questions are how the asset is protected and whether the data holds up. (Mahesh Narayanan)
• Three questions you can answer in the first meeting — can it expand beyond the market it was built for, will this founder keep going even if I quit, and will they listen and come back with something better? "It's really not hard for most investors to figure this out." (Mahesh Narayanan)
• Size the round to the traction, not the ambition — a $3M seed at a ~$10–11M pre-money would have diluted everyone "down to nothing" on thin traction; splitting it into two $1.5M seeds let the second tranche price on the progress the first one bought. (Mahesh Narayanan)
• The first approval is the expensive one — then hire the specialist — doing state-by-state approvals himself took the founder eight to twelve months (about six a state); a hire who knew the process, plus the precedent of states that had already approved, took the company from two states to almost thirty in a year. (Mahesh Narayanan)
• Follow the payer — it may be buying something bigger than your product — facilities had no budget, so the company went to insurance; CMS cared about software compliance more than the device, and a state's request to roll it out system-wide, prepaid, became the business. (Mahesh Narayanan)
• Build once, sell to many — don't let customer pull turn you into a consulting shop — before building for the first state, ask the next ones what they'd need, build broad and customize at the edges; building per customer is "an infinite loop" that kills margins. Martin's version: if every customer wants different custom agents, "you just created a consulting company." (Mahesh Narayanan)
• A good CEO knows when to jump in and when to get out of the way — fix it fast when something's going wrong; when something's going right, "step out of the way and just let it run with whoever's running it." (Mahesh Narayanan)
• Simulate the scenarios before the real table — about thirty run-throughs of the final table with friends playing profiled opponents across every stack shape; the goal isn't prediction ("anything can happen") but that nothing on the day is new. (Martin Jacobson)
• Build a dossier on every counterparty — when the cards are random, the edge is everything else you can know: every reported hand, experience, finances. "Luck is what happens when preparation meets opportunity." (Martin Jacobson)
• Thirty opinions beat one — crowdsource the preparation plan from the strongest peers you have, then own the execution yourself. (Martin Jacobson)
• Compress the feedback loop — shorten the time between a decision and the information that says whether it had value; a CEO can engineer something far closer to poker's minutes than venture's decade. (Martin Jacobson)
• Incentives are everything — "why would the frontline staff care about your sales or your reviews if there's no incentive?" Judge an idea by whose incentives it fixes; not every incentive is financial, but there has to be one. (Martin Jacobson)
• Start from a manually proven pilot, then ask whether it generalizes — the idea arrived already working by hand in one London coffee shop; his job was to judge whether one store's problem was the industry's. (Martin Jacobson)
• Leave the capped game — once you've topped a zero-sum game with a ceiling, motivation fades; move to one where the upside is open and the problem is meaningful to you. (Martin Jacobson)
• With risk comes opportunity — "there's no opportunity if you never take any risk"; per Seth Godin, not taking risks is one of the riskiest things you can do. (Martin Jacobson)
• Watch your risk-tolerance drift — tolerance falls with age, "biologically"; be mindful of it so it doesn't quietly make your decisions. (Martin Jacobson)
• Nothing is fifty-fifty just because there are two outcomes — "it's a fight, anyone can win, it's fifty-fifty" is the amateur read; handicap the real odds ("it's eighty-twenty"). (Martin Jacobson)

---

Episode 1 — Alec Torelli

Professional poker player turned investor. "I'm not a risk taker, I'm a risk trader."

Most Compelling First Bet Stories

1. Dropping out of SMU at 18 to play poker professionally

Six weeks into his first semester at SMU, Alec hit a crossroads. He kept missing his Monday-morning economics class because the best online tournaments ran late Sunday nights, and he was about to be dropped from his classes. He had plateaued at his current level of poker and faced a binary choice: quit poker and commit fully to school, or drop out and go pro. Everyone he asked — friends, his college counselor — dismissed it as obviously stupid; he was too scared to even tell his parents. With no role models (online poker was brand new and no one had built the traveling-pro life he imagined) and no framework for a decision this big, he built one from how he plays a poker hand.

He sized the downside first and got comfortable with it: he had ~$20–30K saved, so the worst case was losing it all and coming back a year behind — painful, but recoverable. Then he forced himself to dream about the upside (which most people skip): traveling the world, playing a game he loved, controlling his own time, playing with his heroes. He decided that even just breaking even while seeing 10–20 countries was enough of a dream to justify the bet. He went all in. It worked out far beyond his expectations.

Why it's compelling: It's the origin bet of his entire career and the literal seed of the podcast's thesis — a high-stakes, low-information life decision made by an 18-year-old with no map, solved by importing a poker player's risk/reward discipline into real life.

2. Selling everything he owned at 24 to move to Italy

At 24, Alec sold everything he owned and moved to Italy — where he later met his wife and became fluent in Italian. He frames it as another big-life-decision bet carrying real risk, approached the same way: look hard at the worst-case scenario, then build in "planners" and back doors so it isn't a one-way door. The goal is to cap the downside (land at ~80% of where you were, not zero) while keeping most or all of the upside.

Why it's compelling: It shows the SMU framework wasn't a one-off teenage gamble but a repeatable operating system for life — and this bet paid off in the most personal way possible (his marriage).

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Be a risk trader, not a risk taker. "Taking" risk fixates the mind on what could go wrong. "Trading" risk reframes it as exchanging risk for reward — forcing you to weigh both sides (expected value), the way poker uses EV. Change the language and you change how you approach the decision.

• Size the downside first, then dream the upside. Most people freeze at the risk and never get to the reward. Get genuinely comfortable with the worst case ("can I accept this?"), then deliberately think about everything that could go right. The reward side is at least as important as the risk side.

• Hunt for asymmetry (risk/reward multiple). Don't judge a bet by probability of winning alone — judge it by the payoff multiple. A small pocket pair loses most of the time, but a ~100-to-1 payoff when it hits makes it worth playing. A low-probability bet with a 10x–100x payoff is a bet you should make every time.

• Mind the expiry (option vs. spot). Decisions with an expiration date deserve priority. Poker was a now-or-never option — he could always return to the "normal" path in two or three years, but he had only one window to take his shot. Treat time-sensitive, closing-window choices as more urgent than ones you can hold indefinitely.

• Most one-way doors are actually two-way doors. People over-rate permanence and trap themselves in yes/no thinking. Almost everything can be undone or returned roughly to baseline. Before deciding, look for the back door and build hedges so the downside is capped — sell action to others, keep guardrails, design an exit.

• Set guardrails before you enter. Decide your limits up front — e.g., "I'll invest three bullets max," or "I'll leave when I'm up 3x." Pre-committing to guardrails prevents getting sucked into emotional rebuys or playing too many hands once you're in the moment.

• Practice bankroll management. Never risk an amount on one bet that can ruin you. Risk only 1–2% of the bankroll per bet so any single loss is survivable, and spread across enough "at-bats" for your edge to manifest. Options (like tournaments) can go to zero and still win if they 20x one time in twenty.

• Only bet where you have an edge. He calls himself conservative: he risks money only when the risk/reward calculus is in his favor. Being a risk trader means being selective, not reckless.

• Separate decision quality from outcome (avoid "resulting"). A decision's expected value is locked in the moment you make it, given the information you had — not by how it turned out. A good decision can lose; a bad one can win on luck. Gut-check: "If this had gone the other way, would I still feel it was the right call?" — especially important to ask when you win.

• Be obsessed with process, not results. Money is the byproduct of making great decisions under pressure, not the job itself. Study your wins with the same rigor as your losses, and seek ruthless, honest feedback from people you respect (poker's no-sugarcoating ethos). Build a tight feedback loop so you can correct mistakes rather than repeat them.

• Watch your linguistics — words are programming. Avoid "should" and "have to" (they create false pressure and limiting beliefs). Notice that "I can't" usually means "I don't want to." Prefer "I feel X" over "I am X" (e.g., feel sick vs. be sick) so a temporary state doesn't become an identity. How you frame a choice shapes how you experience it.

• Trust your spirit over the crowd. The biggest barrier (Pressfield's "resistance") is fear of others' judgment overriding your own conviction. Distinguish genuine intuition from ego/delusion, but when your heart of hearts knows something is right and the only reason against it is fear of judgment, that's the signal to act.

• One-sentence summary (his closing advice): "There's going to be a conflict between your head and your heart. Most of the time when I override my intuition with logic, I pay the price; when I trust my spirit, it ends up being the right decision."

---

Episode 2 — Shaun Gold

Two decades running Miami nightlife, then a pivot into venture capital. Creator of Venture Comedy, GP of Improved Ventures. "Everyone's a gangster until they gotta wire the money."

Most Compelling First Bet Stories

1. Funding his own VC firm with his own money in the post-2021 wreckage — the same week his life fell apart

After the 2021 bubble had already burst, Shaun committed to launching Improved Ventures in 2022 with entirely his own capital. He'd had close to a dozen LPs verbally committed — including one who was going to roll their own fund into his in a GP/LP structure, paperwork and all — and one by one, every single one evaporated ("they were in the inbox, but they were not in"). The week he had to wire his money in, his apartment of eight years was sold out from under him during Miami's worst rent inflation (rents jumping from ~$1,500 to ~$3,000 overnight), and his nightlife clients offered him a secure salaried 9-to-5 — which he walked away from because it would have shut down everything else he was building. It was a perfect storm arguing against the bet: bad vintage, no outside capital, personal financial pressure, and a guaranteed paycheck on the table. He wired the money anyway, for two reasons: he'd given his word ("I'm a person of my word — that meant more to me than anything else"), and he judged it a gamble that wouldn't pay off in a week, a month, or even a year, but would eventually let him operate at a level he'd never reached before.

Why it's compelling: It's the purest form of the show's premise — real money in, every external signal saying no, commitments from others proven worthless, and the decision carried entirely by asymmetric-upside logic and personal integrity rather than validation.

2. At 17, flying to the Bahamas to throw parties — then eating two years of losses in Miami to buy two decades of a rocket ship

Before cell phones were ubiquitous, a 17-year-old Shaun was flying to the Bahamas to throw spring break parties with no idea what he was walking into. He then moved to Miami knowing nothing and no one, with only the conviction "this is what I have to do — I don't know how I'm gonna do it." It took two full years of losing money and being laughed at before he established himself — two years he now frames as the cheap price for nearly two decades of a "rocket ship" running some of Miami's highest-grossing nightclubs. That nightlife apprenticeship ("you were only as good as your last party," always working with nothing) became the exact mentality he later imported into venture: think long term, stay patient, survive.

Why it's compelling: It's the origin bet that built both his risk tolerance and his framework — the proof, at the very start of his career, that sacrificing a defined short window of pain can buy a decades-long compounding payoff.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Don't ask "will it work?" — ask "is the upside dramatically larger than the downside?" His core mental model. He doesn't try to handicap probability of success or failure at all; if the asymmetry is big enough, he goes. No hemming and hawing.

• "Everyone's a gangster until they gotta wire the money." Nothing is real until the wire hits the account. Verbal commitments, enthusiastic emails, even trips to Monaco together are worth zero. Discount all soft commitments to nothing and plan as if you're doing it alone — because you probably are.

• The "What now?" test. The bold move is the easy, braggable part. The real work is the unglamorous phase right after — actually building the structure to deliver on the intention. Before betting, accept that the "what now" period is long, invisible, and the part nobody posts about.

• Think in seasons, not days: "you gotta eat twelve to twenty-four months." Winter-summer, winter-summer. Judge bets on a years-to-decades horizon and be willing to sacrifice a defined 1–2 year window of losses for a decades-long payoff. Looking back, the sacrifice window always seems small.

• Survivability is a competitive advantage — be the mosquito. If you can survive longer than people with more talent and more resources, you win. Persistence plus staying power beats pedigree. (Echoes Martin's YC point: the single trait that predicted success was that the founder didn't quit.)

• Bet on your own weirdness; never outsource your mind. Your unique core competencies — the odd, unrepeatable parts of you — are the only durable strategic advantage. People who outsource their thinking to social media, ChatGPT/Claude, influencer courses, or copying Zuckerberg-style playbooks give up the inner strength they'll need to keep going. (His own version: betting a career on combining venture capital with comedy.)

• "If the thought of something seems crazy to you, then you weren't crazy to begin with." A measure of irrationality is required — for startups and for funds. If an idea passes the asymmetry test but still feels insane, that feeling is not disqualifying; it may be the signal you're early.

• Your word is a decision-making anchor. When the analysis is murky and the storm is overhead, "I said I would do it" carries the decision. Being a person of your word functions as a tiebreaker that prevents rationalizing your way out of hard commitments.

• Solve your internal problems first. Before chasing an external problem worth solving: fix poverty (make some money) and build a track record of core competencies in the domain — so when you do start, you can actually deliver. Don't write biotech decks if you know nothing about biotech.

• Be able to operate with and without the tools. "The person that can operate in a world without AI and with AI is gonna take all of your jobs." A hammer doesn't build the house. Tools amplify judgment, taste, selling, and marketing — they don't replace them.

• Pivoting is fine; quitting isn't. Change tactics as much as needed, but abandoning the game entirely ("I'm going back to selling aluminum siding") is the only real failure mode.

---

Episode 3 — Simon Lancaster

Fifteen years shipping hardware inside Apple, Google, BlackBerry, and Toyota, then founding partner of OmniVentures, a pre-seed "manufacturing VC" that just closed a $33M fund. Co-author of "Unlocking Alpha: The Rise of the Niche VC" and host of the Beers with VCs podcast. "If I were to say fintech, would you say building banks?"

Most Compelling First Bet Stories

1. Leaving the logos to raise a "manufacturing VC" fund before the market agreed

After fifteen years shipping hardware that ended up in billions of pockets, Simon looked at factory floors still running on paper and old ERPs and concluded that's where the money was going — before the AI-manufacturing moment, not after. OmniVentures didn't even start there: the original idea was broad deep-tech/hardware investing, but watching venture itself get democratized and commoditized, he kept asking "where is our alpha?" and double-clicked down — hardware → deep tech → manufacturing — until he hit the one major sector of GDP that has never been digitally transformed. Finance got fintech, business services got SaaS, IT got search and the web; manufacturing still runs on paper, Excel, and email — an industry so left behind that people hear "manufacturing tech" and ask if he means building factories. In 2023–24 he bet it could not stay that way, and that the tipping point was close. The market disagreed for most of three years of raising: meetings were hard to get, and the financially focused LPs almost all said "too niche" — long sales cycles, capital intensity, no flashy exits (the space's real exits are quiet unicorn-scale sales to deep-pocketed public acquirers that never make the news). Conviction came from the fundraise itself: pitching thousands of operators — many of them prospective LPs and founders — he found the people who "got it the quickest" were the next generation, the up-and-coming executives and owners' kids about to inherit businesses from retiring baby boomers, frustrated by the lack of basic automation, afraid of labor shortages, and flatly unwilling to run the companies the old way. Then, in the last year, the unlock arrived — AI-accelerated coding made specialized vertical tools buildable by tiny, domain-native teams — and the fund closed at $33M (through Cool Water Capital, with Allocator One anchoring the early close) in what he calls arguably the toughest fundraising environment of the past twenty years.

Why it's compelling: It's the show's thesis in pure form — a bet made years before it looked inevitable, against a near-unanimous "too niche," carried not by gut feel but by proprietary evidence the spreadsheet-first LPs couldn't see: a generational handover happening inside his future customers. And it's a double bet — his career and capital into the fund, and the fund into a written-off sector — that the market only marked correct at the very end.

2. The Xenode bet — doubling down on the founder who zagged

One of the only companies OmniVentures has made a follow-on investment into is Xenode, Brandon Bourne's electronic-design-AI company in San Francisco. The consensus play in the category was the ten-year holy grail: AI-powered EDA that designs PCBs itself — a product that wouldn't work for years and, worse, demands an accuracy level so high it couldn't be sold for years even once it worked. Xenode read the market with both a technical and a commercial eye and zagged: engineers spend the rest of their time outside CAD — researching parts, digesting datasheets, comparing forty open browser tabs — so it built for that workflow, where much lower accuracy is already efficient, useful, and valuable to the customer. Same giant market, but a product it could start selling almost immediately. Martin's distillation on the episode: if you're uncertain whether they'll buy it or whether you can build it, build the version that delivers value fastest — "you'll have your answer sooner than ten years."

Why it's compelling: It's Simon's fund thesis working in the wild — proof of ROI over proof of concept — and the portfolio-level mirror of his own first bet: don't out-wait the giants on the long-horizon prize; find the wedge where a modest product creates provable value now, and let a short time-to-answer de-risk the big bet. Conviction was strong enough that a concentrated pre-seed fund paid up twice.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Alpha = Mastery × Focus × Network (MFN). The framework at the heart of Unlocking Alpha. Mastery is deep domain expertise (specialization). Focus is knowing your area of edge and staying inside it — for OmniVentures, one stage (pre-seed) and one sector. Network is the standing reminder that venture is a people game: you earn your way into rounds, and people have to want to take your money. It's a product, not a sum — you can score low on one or two, but not on all three.

• Run founders through the same MFN screen. Their founder archetype — written two years before the book — maps onto it one-to-one, a correlation Simon says he noticed for the first time on this episode: technical founders (mastery) solving the biggest pain point of their careers (focus), who have spoken with 100+ customers and can call them on day one to get deployments (network).

• Keep double-clicking until you find the left-behind sector. When venture itself is commoditizing, generic positioning is worthless. Drill down — hardware → deep tech → manufacturing — until you hit ground nobody is standing on, then apply the last-bastion test: every major GDP sector has been digitally transformed except this one, and "it cannot continue to be left behind" forever. Bet on the inevitability; then the only question is timing.

• "If I were to say fintech, would you say building banks?" His check for category confusion: when the market can't even parse your sector's name (manufacturing tech ≠ building factories), the left-behind mentality is intact — which is exactly the evidence of opportunity.

• Build conviction from customers — especially the incoming generation. His conviction wasn't intuition; it accumulated from thousands of conversations with customers, LPs, and founders. The highest-signal source: the next generation about to take over from retiring baby boomers, who refuse to run the businesses on paper and Excel. He watched for three frustrations — no basic automation, labor-shortage fear, a generational handover in progress — and noted which personas "got it the quickest" when he pitched.

• Proof of ROI beats proof of concept. Industrial customers are old-industry and impatient; they don't want concepts, they want deployments that pay back fast. His winning founders showed up with near tailor-made solutions carrying very short payback — an owner shown short ROI has to try it, "otherwise you're dead in the water." Corollary: thread the needle of capital efficiency + novel technology + a two-to-three-year commercialization window, never ten.

• Model the bet — regress thirty years of exits. Credit to co-GP Sabrina Passman: an "insanely detailed" fund model took essentially every industrial-space exit of the last three decades, filtered each by "would our thesis have invested?", took the exit values, and back-solved the entry ownership required under various simulated bet counts and raisable fund sizes for a first-time manager. That arithmetic — not vibes — set the $33M fund size.

• Specialist concentration over spray-and-pray. A small emerging manager has two live options: broad exposure hoping to catch one outlier, or fewer deals at higher ownership, earned through specialization. (The generalist path can work — but only if you compensate with an insane network.) He chose concentration, because specialization is what lets you "see ahead of the curve and around corners where others couldn't."

• Steel-man your gremlin with outcome math. The fear in the back of his head was "too niche" — would there even be follow-on investors? He dispelled it quantitatively: model whether an $800M–$1B exit still returns the fund if you get in early enough, with enough ownership, on a short enough exit horizon. (He's now so far past the objection that he's personally an LP in several sub-$5M funds.)

• Ride the capital-efficiency wave — seed-strapping. Early to the view that AI-era teams can build these businesses without mega-rounds: vertical tools instead of Oracle-for-everyone, tiny teams, domain-native founders (Martin's echo: a portfolio company that went from fifteen people to two while growing revenue 20%). That converts "small fund" from a weakness into a structural fit; he counts himself an early adopter of the pre-seed "seed-strapping" mentality.

• When everyone zigs, zag to the wedge you can sell today. The Xenode principle: when consensus chases the long-horizon, high-accuracy holy grail, attack the adjacent workflow where lower accuracy already creates customer value and revenue starts immediately — same giant market, dramatically faster feedback. Under uncertainty, prefer the bet that returns its answer soonest.

• The standouts explain, not just build. Across his ~thirty investments, the outliers share an uncanny ability to explain technical problems and their solution in a way that quickly builds rapport and trust with customers and investors. Not technical savants — people who deeply understand the pain point and communicate it clearly. As building gets easier, this (plus your distribution wedge, as Martin added) becomes the real differentiator.

• One-sentence summary (his closing advice): Back the founder — or be the founder — who can communicate a very large vision, just large enough, while having spoken with a hundred customers, keeping a finger on the pulse of today's problem, and owning a unique, capital-efficient go-to-market wedge.

---

Episode 4 — Arkady Kulik

Deep-tech VC — founding partner of rpv — and an entrepreneur since age 18. Invests at pre-seed in the hard stuff: energy, neurotech, medical devices. "You're gonna get a fraction of a fraction at the end of the day — so yes, the ambition really is important."

Most Compelling First Bet Stories

1. The energy-storage bet that passed every test — except the one he didn't run

A company came through his pipeline (prior fund) attacking one of the critical problems in energy storage. He did everything right by the classic playbook: deep scientific and technological diligence, then he and his partner flew to the site and spent a full day with the founder and team. Real device, real problem, real customers. The underwriting checklist came back all green — size of the market: check; groundbreaking technology: check ("it is still there"); founders' capabilities, resilience, greed: check. He wrote the check. The one variable he never tested was the size of the team's venture-scale ambition — because coming off 20+ years as an entrepreneur himself, he asked "is this a good entrepreneur?" (yes) and "is this a solid team?" (100%) but never "is this a good entrepreneur to run a venture-scale business?" That fine distinction — a strong operator who'll happily build an excellent lifestyle-scale company versus a founder wired to chase a unicorn — turned out to be the whole ballgame. The company is doing fine: good revenue, probably tens of millions soon, a great outcome for founders who own most of it. "Are they gonna do well? Sure. Is it gonna be a good investment for me? Maybe not so sure" — it won't return his fund. And fund math is merciless: the fund gets a fraction of the founders' outcome, and the GP a fraction of the fund's fraction. The founder's appetite sets the investor's ceiling.

Why it's compelling: It's the cleanest demonstration in the series so far that you can be right on every classic underwriting axis — team, technology, market, product — and still make a bad fund investment, because the decisive variable lived inside the founder's head. The miss permanently added a line item to his diligence: ambition is not a given; it must be underwritten like everything else.

2. First VC check into Humanity Neurotech — after melting the wall of ice

The deal most investors would flee — a medical device ("people hear medical device: immediately FDA, regulation, no no no, not gonna touch it") — is the recent bet he's proudest of. Humanity Neurotech, led by Blake Gurfein, builds a neuromodulation halo worn around the head that uses a very specific type of magnetic wave to heal inflammation in human tissue, starting with the brain precisely because it's the hardest place for traditional pharma to treat. Arkady had been talking with Blake since before the company launched and wrote the very first VC check into the firm. The scientific diligence went deep — physics and neuroscience, led by his scientific partner Richard Silberstein — but the unusual investment was in the founder himself: 10+ net hours in person and on Zoom versus his usual three or four, because the interpersonal channel was broken. His self-description: "a Russian dude trying to decipher the social code of an Italian American." The fix wasn't a framework; it was an awkward lunch in Palo Alto — "Man, I sometimes feel like there's a wall of ice between us. What's going on?" Just a conversation, openly held, and the ice melted. The famous risks never spooked him: in neuroscience the FDA is a given, so he graded how Blake thought about it — and Blake "had his stuff together from day one," clear on FDA and, more impressively, on reimbursement codes, the bigger problem almost everyone misses. Ambition was never in question: Blake plainly intends a multi-billion-dollar company that heals people along the way. By every standard this was the riskier deal — more risks, bigger magnitudes, a therapy never done before — and he made it before the energy-storage investment. It keeps playing out: the first clinical trial finished a complete success with 80% of patients reporting positive results, and a bigger open-indication trial with Mount Sinai is now underway.

Why it's compelling: It's the perfect inversion of the first story. Every risk you could name in advance was larger, but the two variables the classic model doesn't price — venture-scale ambition and a communication channel he could trust for ten years — were both strong, and they're exactly what's compounding. It also carries the episode's best micro-lesson: he nearly let an unspoken interpersonal chill block his best recent bet, and the entire remedy was naming it out loud.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Underwrite venture-scale ambition as its own risk. The question is never just "is this a good entrepreneur?" but "is this a good entrepreneur to run a venture-scale business?" A strong operator content with a healthy lifestyle-scale outcome will make themselves wealthy and still be a bad investment for the fund. Fund math enforces it: the fund earns a fraction of the founders' outcome and the manager a fraction of that fraction, so the founder's appetite is the investor's ceiling.

• Run a two-gate IC: blockers first, then a weighted scorecard. Gate one exists only to find dealbreakers — six questions that can go wrong on team, three on market, five on product/technology, two on strategy, two on round parameters — with team the most important blocker layer, technology second, strategy/round/market third. Only if nothing fires does gate two score the deal on fixed weights: ~23% team & people, ~8% market & need, ~28% the mix of science/technology/product, ~15% competition, ~11% traction (plus regulatory where it applies), ~15% strategy & round parameters ("we need to understand how we're gonna make money as investors in that specific deal"). And hold the numbers loosely: every VC's weights encode their own perception of risk — his 28% on tech and someone else's 14% "might both be right — for themselves."

• Price the pivot optionality before anything else. In software the ability to pivot is close to unlimited — a failed video game became Slack and IPO'd. In deep tech it's vanishingly narrow — a failed solar-panel company doesn't pivot into nuclear engines; "you basically bankrupt one company and launch another one." The narrower the escape routes, the more diligence weight belongs on the technology itself. (Martin's contrast from the software side: roughly half his portfolio pivoted after the check — which is exactly why product is only ~10% of his underwriting versus Arkady's 28%.)

• Underwrite the relationship like the asset it is. A pre-seed check means seven to ten years with this founder; broken communication at entry compounds into disaster. Trust is functional, not sentimental: "when founders really need your help, they're ashamed to even mention it on a call" — only a real relationship surfaces problems while you can still help. It also converts to money directly: access to secondaries, preferred liquidity in M&A or IPO. His tooling is disarmingly simple — scale founder time way up when something nags (10+ hours vs. his usual 3–4), and name the problem to the founder's face; same rule he keeps at home with his wife: talk about the negative stuff openly before it bursts. (Martin's companion screen: the founder needn't take his advice, but must be curious about it — comments dismissed out of hand predict a decade of bad conversations.)

• Treat regulatory risk as a variable to grade, not a gate to fear. In his sectors the FDA is simply a given, so the diligence question becomes how does the founder think about it? The underrated half is reimbursement: FDA approval doesn't put you "in money" — locked into out-of-pocket payments, "your revenue is abysmally smaller" than with insurance reimbursement codes. A founder fluent in both from day one is the real de-risking.

• Ask where the founder lies to themselves. Good founders are necessarily a little delusional — "there is no way you can be absolutely sane and rational and be a successful founder at the same time." So the diagnostic isn't whether they're delusional but where. Knowingly lying about anything → very strong signal, never do business. Delusional to the point of hearing no feedback → pass. Delusional about hard scientific reality (blood tests from a single drop, a tabletop fusion engine) → pass; founders must accept the physics of the world. Delusional about how big and great they can become, paired with a beginner's mind open to any advice or question → "this is the perfect combination."

• Bend what's malleable; never bet on physics changing. "Human systems are malleable, whether it's governments or industries or relationships. Real physics of real things is less malleable." Uber looked delusional, but Travis bent regulation and an industry around a real unlock that already existed — the GPS chip in every phone. If he'd been building Uber while hoping GPS would show up, "it's a fail." Back founders who bend human systems around real physics; pass on founders waiting for nature to cooperate.

• Map the scientific landscape — it moves slowly, so the work compounds. Read the corpus, retain an expert, or at minimum read the summaries: what was true of energy storage ten years ago mostly still holds, so the map keeps its value in a way software maps never do. Deep-tech competition is more knowable (often only ~3 labs are anywhere near a problem) but broader than direct rivals — a brain-inflammation halo also competes with pharma, surgery, and healthy eating. The map is also your fraud detector: generations of physicists failed to extract energy from a well-known microscopic effect; the founder with no physics degree has not "figured it out" — and people who understood biology never invested in Theranos.

• Filter all advice by stage and sector — including his. VC is "way less of a uniform homogeneous industry than people think" — friends & family/pre-seed/seed, Series A/B, and growth are different businesses, and Series B fintech rules don't transfer to pre-seed pharma. Don't take advice found online at face value; translate it into your specific industry and stage first.

• One-sentence summary (his closing distinction): "Human systems are malleable — governments, industries, relationships. Real physics of real things is less malleable." Invest in the delusionally ambitious who accept physics and bend everything else.

---

Episode 5 — Aram Attar

Twenty years and 50+ deals across three continents — including a $1.6B LBO — before walking away to found The VC Factory and build "mindset-based investing," now the subject of his doctorate. LP in Incisive Ventures Fund II, a bet he made without ever opening the data room. "Intuition is not a mystical gift."

Most Compelling First Bet Stories

1. Walking away from the deal career to bet that VC runs on a broken operating system

After fifteen-plus years as an investor and advisor — fifty-plus closed deals on three continents, a $1.6B LBO among them — Aram started teaching VC in 2018: first students, then aspiring VCs, then working VCs, through his online VC Career Accelerator. Every cohort asked the same question: how do I make better decisions? Hunting the answer, he listened to the era's vocal power law masters — Brad Feld, Fred Wilson, the twenty-to-thirty people who had repeatedly manufactured billion-dollar outcomes — and noticed the anomaly hiding in plain sight: all these people ever talk about is psychology. The Kahneman-grounded diagnosis followed. Expert intuition — the fire captain who empties the house moments before it collapses, the chess grandmaster's instant read — is real, but it only develops in environments with rules (same action, same result) and short feedback loops over thousands of reps. Poker has both. VC has neither: five founding teams of identical quality can produce four failures and one win on luck and timing alone, and the feedback on any check takes years to arrive. Deal volume, the one ingredient VCs do have, just breeds confident pattern-matching — "intuition is not a mystical gift... it's only recognition of what you've done before." So he bet his second act on replacing the industry's favorite tool: built the mindset-based investing framework (improved returns at every capital-allocation step, LP→GP and GP→founder), trained dozens of GPs and hundreds of founders, and enrolled in a doctorate to nail it academically. The honest engine under it all is his own $2B exit: "I was lucky... mostly right time... I haven't done it twice." A man who knows his biggest win wasn't a system went looking for one.

Why it's compelling: He won the traditional game, then walked away on a contrarian thesis about how the game itself is played — that intuition, the tool most investors treat as a superpower, is demonstrably unreliable at the early stage because venture can't teach it. It's the series' first first-bet on an idea, and he's staking his second career, his own LP checks, his published reputation, and now a doctorate on being right.

2. The LP check into Incisive Fund II — written without opening the data room

Aram's LP thesis starts from a data anomaly he calls the emerging VC conundrum. In the Cambridge Associates vintage tables for roughly the last decade, half of each year's top-ten funds are Fund I or IIs — every single year — with another couple at Fund III/IV and only one or two established brands. Yet LPs refuse the pocket, for one dominant reason ("they don't have a track record, so we don't know if they're good") backed by a second (no social proof). His counter: the track-record compass is broken anyway — persistence research puts the odds that a top-quartile fund repeats at ~30%, worse than a coin flip. Before betting real money on that conclusion, he did the work: with a team of three MBA researchers he spent months reverse-engineering how the best emerging-manager pickers — Beezer Clarkson (then at Sapphire Partners), Michael Kim at Cendana Capital, Samir Kaji — actually decide, combing hundreds of podcasts, articles, and videos (AI-assisted, manually verified) and distilling six rules they apply consistently, every one traceable to mindset. Then came the bet on Martin. Martin sent the data room; Aram never opened it. His diligence was Martin's YouTube interviews about his life and how he makes decisions, direct conversations, and watching Martin field questions live at his South by Southwest event — Aristotle's 2,500-year-old test: you judge someone's character by the way they make decisions, not even by the actions. He scored Martin against the five power-law-master traits he had already published (reputation committed in advance), read Martin's history of making his own way at sixteen as the root of low loss aversion and contrarian comfort, and counted Martin's pass-review algorithm — re-examining anti-portfolio deals to separate decision quality from decision outcome — as the clinching signal of a true student of decision-making. Conviction was never the bar: "I don't need super hard conviction... Conviction is I believe in my framework and I'm willing to put money behind it." The early returns he cites are informational: Martin's LP newsletters (the "SaaSpocalypse" analysis) carrying insight he's seen nowhere else.

Why it's compelling: It's the show's premise turned back on the host — a real capital-allocation decision into Martin's own fund, dissected step by step, made with zero spreadsheet diligence. Character-by-decision-process replaced the data room; a researched framework replaced manufactured conviction; and a contrarian read of the persistence data turned "no track record" from a disqualifier into the exact source of his edge. (His first LP check ran the same play: a GP with fifteen years in energy and climate deploying a climate/AI fund, judged by watching what he does with founders before he invests.)

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• The three-step process of power law masters: pause intuition → collect data scientifically → decide promotion-focused. The best repeat investors are good at having the intuition and then delaying it. Step two collects data expressly to fight confirmation bias — you are trying to disprove the intuition, not dress it up. Step three decides with a promotion focus: "they don't really care about the loss, they're trying to win. They're not trying not to lose." Intuition is the first step of the process, never the verdict — if it survives the disproof attempt, keep it; if not, revise it.

• Check whether your environment can even produce expert intuition: rules + fast feedback. From Kahneman: the firefighter and the grandmaster earned trustworthy gut calls from rule-bound worlds with immediate feedback over thousands of reps — poker qualifies on both counts. VC fails both: no rules (five identical-quality teams; four die on luck and timing) and no fast feedback (a year or two to the next round, longer still to failure). VCs have only volume, so they mistake pattern-matching for "a God-given gift, a mystical thing." If the environment can't teach intuition, don't let intuition decide.

• Track record is a broken compass — and that's the opportunity. Persistence in VC is scientifically unconfirmed: a top-quartile fund has roughly a 30% chance its successor is top-quartile — less than a coin flip. Meanwhile half of each year's Cambridge Associates top ten are Fund I/IIs, out of thousands of emerging funds. LPs screening on track record and social proof are structurally locked out of the best-performing pocket — leaving it to anyone whose evaluation method doesn't need those crutches.

• Evaluate emerging GPs as entrepreneurs first, investors second. A Fund I GP is a founder building a firm, so run founder diligence: are they talking to startups, testing the thesis, getting the raise done — or burning hundreds of thousands on lawyers and touring LP resorts? (He's been burned: "I put money in people who were going to resorts and they never raised the fund.") Fund size is a mindset tell — GPs, like founders, tend to ask for too much money; the GP raising a deliberately modest ~$5M fund "to test themselves and show they're good investors" is displaying exactly the groundedness you want.

• Judge character by decision process — Aristotle's test. "The way you can judge someone's character is by the way they make decisions. It's not even the actions." So make that the diligence: skip the data room, watch their interviews, talk to them, join the investment committee (he usually does, in funds he backs) and ask why they picked that founder, that market. Screen for first principles versus FOMO and social proof; for data-drivenness; above all for self-audit — Martin's algorithm revisiting passed deals to separate decision quality from decision outcome was, for him, the signature of someone deciding the right way.

• The five traits of future power law masters — published in advance, reputation attached: (1) they see what can go right, not what can go wrong; (2) they're willing to swing big — "you can just lose one X but you can miss hundred X, so why don't you go all in?"; (3) they're okay being wrong — low loss aversion; (4) they're not afraid to be contrarian; (5) they have a high tolerance for risk, uncertainty, and ambiguity ("which are the same thing").

• Don't manufacture conviction — bet the framework. "Everyone's talking about conviction now, but... much of the time you're trying to convince yourself anyway. So it's not real conviction." His replacement: "Conviction is I believe in my framework and I'm willing to put money behind it." In a domain with no fast feedback, framework confidence plus survivable sizing beats case-by-case certainty.

• Interrogate the pattern-match: "Why are you here?" When intuition flags a category-switcher — the B2B SaaS veteran suddenly raising a climate fund — pause the reflex and try to disprove it: What did you understand that nobody else did? Do you have another advantage here? "I just prefer it" fails. A genuinely non-consensus angle passes, and is worth extra precisely because every LP running the reflex unexamined has already screened these people out: "Better for me. That's where I get an unfair advantage that nobody else has."

• Test the spin-out halo: brand or person? After track record, the second thing LPs reach for is pedigree — "they worked at a VC firm before." But you must separate the firm's success from theirs. His hard signal: did founders follow the GP out of the old firm, taking the new fund's money in their next round? Absent that, pedigree is not obviously worth more than a thesis the GP can defend from lived experience.

• Resilience over grit. YC's finding — the one trait correlating with founder success was not quitting — is grit: perseverance plus passion (Angela Duckworth). Aram upgrades it: grit "can lead you to do the same thing always" straight into a wall; resilience is better because it allows the pivot, and its test is whether someone can "take failure and make something out of it." Under real uncertainty all you can underwrite is sound decisions — so he'd re-back a sound decision-maker who simply got unlucky.

• Go in with eyes wide open — size the bet so losing is fine. For LPs entering venture: never demand short liquidity ("every time a GP pitches me on liquidity, I'm like relax, dude — I know it takes time to build a company, so I know it takes time to build a great fund"), and commit only money you're okay to lose, because "it's still rare to make a lot of money" — you need roughly top-decile funds to earn the illiquidity premium. He writes deliberately small checks he can afford to lose; the family-office member with 80% of a billion dollars in private equity was rational for the same reason — the other $200M was plenty to live on. (Martin's mirror: 80% low-leverage real estate, 20% venture.)

• One-sentence summary (his closing advice): Evaluate emerging GPs as entrepreneurs first — watch how they build the firm — then get inside how they make decisions, and back the promotion-focused ones: "trying to win... not trying to preserve something," okay to bet their house on their own decisions.

---

Episode 6 — Ihar Mahaniok

Learned to code at twelve in Belarus, spent twenty years shipping software at Google and Facebook, then — no MBA, no finance pedigree, no GP role — bet his own money into 100+ startups. Five became unicorns: Instacart, PandaDoc, and People.ai among them. Now managing partner of Geek Ventures, a $23M Fund I backing immigrant founders in AI and robotics. "Don't pitch me what you're building. Pitch me yourself."

Most Compelling First Bet Stories

1. The Instacart seed check — re-deriving the analogy in a "cursed" category (2012)

When Ihar wrote his Instacart check in 2012, grocery delivery wasn't a category — it was a cautionary tale. Webvan had incinerated hundreds of millions a decade earlier buying warehouses, trucks, and payroll, and VC sentiment on the space was openly negative. His entire bet came down to refusing the obvious analogy: "The key thing that stood out for me for Instacart was that it wasn't that it was similar to Webvan — but that it was similar to Uber." He was already an Uber user, impressed by how seamless it was, and had absorbed the deeper unlock before "gig economy" was even a settled term: random part-time people as an elastic labor supply. Uber had taken the transportation angle; the same marketplace dynamic could carry many other jobs — and groceries was a market as big or bigger, with a universal customer: "Not everybody might need a taxi, but everybody needs food." What converted thesis into check was proof at miniature scale: Instacart was live in a single San Francisco neighborhood with essentially one store (Trader Joe's) — but inside that postage stamp they had demonstrably sold the supply side (a real retailer) and delivered the demand side (real customers ordering). On team, he knew exactly which judgment wasn't his to make: they were YC, he had concluded YC is "actually very good in interviewing the team," so he rented that vetting — and his entry trigger was the moment Khosla Ventures committed to lead the round (Sequoia came in the next round, after him). The moat argument sealed it: unlike pure software, cloning Instacart meant going into the real world. It became one of the first "Uber for X" checks of that whole wave, the first entrant into food delivery — and one of his five unicorns.

Why it's compelling: It's the podcast's premise in a single decision. The crowd had priced the category off its most famous corpse; Ihar re-priced it by re-deriving the reference class himself (asset-light Uber, not asset-heavy Webvan), verifying the core loop at one-neighborhood scale, and having the self-awareness to borrow YC's team judgment where he had no edge. The crowd wasn't wrong about Webvan — they were wrong that Instacart was Webvan.

2. PandaDoc — a ~150x outbound check sourced from an online pitch competition (2012–13)

Of the hundred-plus companies Ihar has backed, PandaDoc is the only one he ever found through an online competition — a fully remote video demo day in 2012 or 2013, which "was a big difference" in a world where Zoom didn't exist yet. The company wasn't even called PandaDoc: it was SaaS for salespeople — quotes and contracts redlined back and forth between vendor and buyer, "a very evolved Google Docs for salespeople" (the founders' quoting product, Quote Roller, later became PandaDoc; e-signatures were added only after his check). They won the competition; he compared them against every other participant and reached out cold — "it was my outbound." With no YC to rent judgment from, the diligence was all his: mutual friends validated that the founders — Belarusian immigrants Mikita Mikado and Sergey Barysiuk — were real, with no red flags. They were less established and less senior than Instacart's founders, but they were second-time builders: they'd run a web agency, knew how to work with B2B customers, and were raising capital for the first time. They already had revenue — roughly $50–100K ARR in the earliest innings of the SaaS wave, when the comparables were basically Salesforce and DocuSign. Then the clincher, pure price discipline: he ran his own competitive analysis and found the Silicon Valley rival with less ARR demanding a higher valuation. PandaDoc was about 5x cheaper on valuation than Instacart — and the price was decisive: "If PandaDoc was valued the same as Instacart, I wouldn't have invested." The founder quality that carried the read was character, not pedigree: "I was impressed by the tenacity of the founders... they were creative, they were resilient." The outcome: roughly 150x — and PandaDoc eventually beat that better-pedigreed, pricier competitor, a result he only learned in hindsight.

Why it's compelling: It's the perfect mirror of Instacart. No accelerator halo, no name-brand lead, no warm intro — an outbound check into an unknown immigrant team sourced from an internet video contest, closed entirely on diligence he could run himself: network reference checks, a tenacity read, real ARR, and a comps-based entry price. It shows his asymmetry math nakedly — the same quality of asset at a 5x cheaper entry is a different bet — and it proves the Instacart check wasn't borrowed conviction.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Pick the reference class yourself — "it wasn't similar to Webvan; it was similar to Uber." Categories get priced off their most famous failure. Re-derive the analogy from the mechanics: Instacart was the proven Uber marketplace dynamic — part-time gig labor as elastic supply, asset-light where Webvan was asset-heavy — applied to a market with a universal need ("not everybody might need a taxi, but everybody needs food"). Choosing the right precedent contained the entire bet.

• Demand the core loop proven at tiny scale. One neighborhood, one store: a real retailer signed on the B2B side and real customers ordering on the B2C side — both halves of the marketplace spinning for real, however small. PandaDoc's version was $50–100K of ARR. He doesn't need scale before the check; he needs the mechanism demonstrated rather than asserted.

• Know which vetting to rent and which to run. On Instacart he consciously outsourced team judgment — "YC is actually very good in interviewing the team" — and let Khosla's lead commitment be the trigger. On PandaDoc, with no one to rely on, he ran it himself: reference checks through mutual friends, plus his own read on tenacity. The skill isn't doing all the diligence yourself; it's knowing which judgments are yours to make.

• Valuation is always an input — entry price is the asymmetry lever. "Upside and risk-reward always takes into account the entry valuation." His comp work surfaced a rival with less ARR at a higher price; PandaDoc sat ~5x below Instacart's valuation, and at Instacart's price he simply wouldn't have invested. The identical company at the wrong price is a different bet — and the ~150x outcome is what a disciplined entry buys.

• The talk-shop test for technical moat. He's an engineer, but he has never reviewed code before investing. The test: a proper CTO with real credentials who can talk shop, building something a cloner would need six months to a year to replicate (in the pre-AI era) — or, like Instacart, a real-world operation code alone can't copy. The auto-fail: an MBA who won't hire technologists, ships landing pages and a spreadsheet, and claims a unicorn. "He will not. Or she will not."

• Fear the accidental platform, not the deliberate clone. "Could Google, Adobe, or Microsoft just build that?" is mostly a phantom risk — big companies never go into a small thing (Adobe moved on e-signature long after DocuSign had already achieved escape velocity). The mortal risk is incidental coverage: when ChatGPT already gives people the answers your software sells, they stop buying your software. Ask "who does this by accident?", not "who could copy this on purpose?"

• Right people × right market × right time — with a crisp "why now." The Cytronic check (2024, incorporation/pitch-deck stage — decided fast): founders who had built Shyp and Airhouse and knew fulfillment's every wart, including exactly what's broken in pre-robotics warehouses; e-commerce and logistics, two intersecting trillion-dollar markets that only grow; and robotics freshly crossed from research project to commercially deployable — "robotics five years ago, I would be very reluctant... robotics is finally here." Two years later the fully automated warehouses are live, with the company publicly launched.

• Underwrite the space and the person; treat the plan as minutiae. Build new robotic warehouses or retrofit existing ones? "This is already a minutia." Early enough, he backs knowledgeable founders in huge markets and trusts them to figure out the details: "If the first iteration doesn't work, you pivot and do something else in the same space." The check has to survive plan A dying.

• The "out of advice" test. His current framework in one line: find founders who know the market so in-and-out that "I feel like I'm out of advice to you... I'm just happy to be along for the ride." One gate he still keeps for himself: the market must be growing — never shrinking, never capped.

• Only invest where you're not needed. "I only invest in a startup where I believe the startup doesn't need me. The startup will be successful anyway." The investor is a non-essential accelerant — 10% faster, 10% bigger — never a load-bearing wall. A deal that requires your help to work is a pass.

• One clear leader — "pitch me yourself" before the company. However many co-founders there are, he's underwriting one person: the CEO who makes the company happen. No clear single leader, no check. So he opens every founder meeting the same way: don't pitch the company — "tell me about yourself before this company... I want to come away impressed... so much that I want to invest in anything you're building. So now tell me what you are building." (The extreme case is real: big companies sometimes acquire a whole firm to install one person — Snowflake buying Neeva to make Sridhar Ramaswamy its CEO.)

• An authentic thesis is both filter and magnet. The immigrant thesis was formalized when Fund I's LPs needed differentiation — but it worked because it was already true: Instacart, PandaDoc, and People.ai were all immigrant-built, and he's an immigrant himself (Belarus → US, 2013). It targets a population with documented outperformance, matches his most natural value-add (a connector's thousands of relationships — the exact thing immigrant founders lack), and it pulls aligned founders in: an oversubscribed founder raising a $15M seed recently made room for Geek's half-million check because of the mission and the community. "It's important to be authentic" — a borrowed lens does none of this.

• Pace the fund to keep the bar — stack-rank, don't binge. Geek Ventures' discipline: about one deal a month, two only for the absolutely unbelievable. Five good companies in one week get stack-ranked against each other and one gets picked. He's honest that the inverse holds too: the "why not" on almost every pass is simply "some other deal was better." Every yes is priced in opportunity cost.

• Version your framework. "My decision framework now is different from a year ago, and obviously from fourteen years ago" — it has to evolve. The current build: AI and robotics exclusively, deeply knowledgeable founders, growing markets, person-first underwriting. A framework is a living instrument recalibrated by every cycle, not doctrine.

• One-sentence summary (his closing advice): Open with the person, not the product — "Don't tell me about your current company. Don't pitch me what you're building. Pitch me yourself" — and invest only when the person alone would have been enough.

---

Episode 7 — Brian Bell

Led AI at Amazon and launched the SageMaker Marketplace, then did his time at Microsoft — "playing Game of Thrones," as he puts it — before walking away about six years ago to write his own checks. Bootstrapped deal flow off Y Combinator, ran an AngelList syndicate that now reaches 16,000 people, and turned it into Team Ignite Ventures. Nearly 400 investments, roughly 100 a year, about half of them YC. Host of the Ignite podcast (approaching 300 episodes) and author of LP: The Insider's Guide to Investing in Venture Capital Funds*, out September 15. "I don't even think I'm an A player. I'm probably like a B player. But I'm a B player who can spot A players."*

Most Compelling First Bet Stories

1. Fund zero — betting that the structure was the problem, not the picking (c. 2021)

Brian's first bet wasn't on a company; it was on the machinery around his own decisions. He'd started the way most people do — angel checks straight into companies, a few LP positions, a seat on other funds' investment committees — and discovered he was having far more fun talking to founders than fighting Microsoft's internal politics. The problem was deal flow: a new angel doesn't have any. His answer was to fish inside someone else's filter. Y Combinator takes roughly 150–200 companies out of about 20,000 applications a batch, so the hardest part of selection is done before you arrive — and he cold-emailed his way into the ecosystem, including a cold note on AngelList to Martin, who took the call while walking a soccer field. Then the structural failures started stacking up, all three of them expensive and none of them fixable by picking better. Rounds closed weeks before demo day. He'd open a syndicate, raise $100–200K, go to wire, and hear that the $20M cap was now a $30M cap ("we doubled revenue and the round's oversubscribed"). Worst of the three: founders who didn't want the deal blasted to 600 people, let alone 16,000. So a man who says flatly "I never thought I'd be a fund manager" raised a $1M rolling fund on AngelList — the vehicle he still calls fund zero — for the single purpose of being able to say yes the same day, with small checks, in rounds that don't wait.

Why it's compelling: It's a rare version of the first-bet story where the diagnosis is the interesting part. The obvious read on losing hot rounds is "I need better judgment" or "I need more capital." Brian read it as a latency problem — his decision speed was fine, his capital's speed was not — and the fix was a structure, not a skill. He also names the two-stage cheat clearly: constrain the pool first (YC's filter is free alpha), then buy the ability to act inside it fast. Everything he built afterward sits on top of a bet that the bottleneck was plumbing.

2. Underwriting by machine — building the model before he knew it worked, and the five-cap check now at ~100x

The second bet was that first-principles machine learning could beat the industry's gut. Brian had led AI at Amazon and knew his way around a Jupyter notebook, so instead of borrowing a checklist he asked the question an ML engineer asks: "How would I set up feature extraction? What are the features that matter? And what are their weights and biases in the model that would output a score that would predict success?" The training data was there if you were willing to go get it — YC's roughly 5,000 alumni with well-documented outcomes, plus his own accumulating record of about 10,000 companies seen a year for six years. He built a system that ingests founder résumés, pitch decks, and the transcripts of thousands of his own calls, extracts about twenty features (friction, founders, moat, traction, whether the problem is category-defining), lets the model fit the weights, and outputs a 1–5 score — human-legible on purpose. Then he wired in the things scores can't express: eight hard gates distilled from his own shutdowns, an eleven-vector fragility score, a red/yellow-flag pass, and a separate power-law score asking whether this is a $10B company if everything goes right. He also had to fight the tool's temperament — early models "told me to invest in everything," in Martin's words — so he pasted in a Marc Andreessen–style instruction demanding truth over comfort, which now "almost painfully tells me what I don't want to hear." The verdict came in the only currency venture accepts. One portfolio company he entered at a $5M cap went on to raise a $650M Series B — about 100x gross. His response was to replay it: run the deal back through the model as of 2022 with only the information available then, discover it had scored 3.8, and ask what the screen needs to learn to catch more like it. Six years in, he says he "very, very infrequently" disagrees with the model now.

Why it's compelling: This is the episode's real first bet, and it was placed years before any evidence could arrive — venture's feedback loop, in his own image, is "playing poker but you don't find out if you win the hand for five years." He committed to a method on faith in the method, then made the delay itself productive by re-running old decisions against the information he'd actually had, scoring his own F1 — true and false positives and negatives across 40,000–60,000 companies — and letting every loss harden into a gate. The 100x isn't the point; the retraining loop it triggered is. And his conclusion is deliberately unromantic about what the machine is for: "AI is not gonna replace venture capitalists... but a VC powered by AI is a very powerful thing."

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Constrain the pool before you improve the picking. YC admits 150–200 of ~20,000 applications, so the filter has already done work you couldn't do yourself. Martin's version of the point: even throwing darts at 150 YC companies beats careful selection from 150 random ones. The side effect matters as much as the primary one — living inside a high-quality pool recalibrates your bar for everything outside it. Non-YC decks routinely score 1.5–1.6 in his model where the YC range runs about 2.0 to 4.3.

• Fix the structure, not just the judgment. Losing rounds to closing dates, cap step-ups, and founders' distribution limits are not picking errors, and no amount of conviction solves them. The $1M rolling fund existed to make a fast yes possible; it changed his outcomes more than any refinement of taste would have.

• Underwrite as feature extraction: features, weights, one score. About twenty features — friction (adopted from Martin), founders, moat, traction, problem/category-defining, and the rest — with the weights fitted from data and the individual ratings open to his override ("you rated the traction a four; I think it's a five because they're doubling every month"). Output on a 1–5 scale rather than 0–1 because that's how humans hold quality: "this startup's a four and a half versus this startup's a two and a half."

• Stack-rank the whole batch before you meet anyone. All ~200 companies ingested four times a year, sorted best to worst, so attention flows down the ranking. He still reviews everything by hand; the ranking decides the order and the depth. Roughly: 2.2 is an automatic pass, above 3.5 earns a closer look and a meeting, 4.2–4.3 is close to an automatic yes.

• Gates before weights. Eight gates, each one built from a past failure, all of which must be full or partial passes before any score matters — "these gates are kind of there to sort of protect the downside a bit." A high score on a deal with a disqualifier is not a good deal.

• Score fragility separately: where is this a single point of failure? Eleven vectors, including founder fragility (solo founder; or a pair without complementary business and technical skills), market fragility (the thesis requires the market to turn one specific way and dies if it turns another), product and capability fragility (success rests on unproven research or hard physics with no IP), and GTM fragility (dependence on one motion, one partnership, or one platform).

• Keep the scores unblended. Power-law score, weighted scorecard, fragility score, red/yellow flags — four separate outputs, because "is this big enough?", "is this good?", "is this brittle?", and "is something off here?" are different questions with different remedies.

• Read the second-order metrics against the cohort. Growth alone is a decoy: "yeah, they're growing 100% a month, but did you notice that their logo retention is 67% and their net revenue retention is 80%?" Capital-inefficient growth is the same trap. Every number gets its meaning from the comparison set — the operative question is always what excellence looks like in this dataset.

• Make the AI adversarial on purpose. Left alone, the models validate you; the earlier ones "told me to invest in everything." His custom instruction, adapted from Marc Andreessen, is built around truth and accuracy and "don't tell me what you think I want to hear." What he gets back is often unwelcome — the deal he wanted to do, missing three or four things — and that's the value. The relationship he describes is closer to an investment committee than a tool: sometimes he overrides it, sometimes it overrides him, and the decision is the consensus they reach.

• Let the machine cross-examine the data room. Dump the documents in and it surfaces contradictions between what was said on the call and what the materials show. Most of it is founders stretching. About 1% of the time it's an outright lie, usually about a credential.

• Founders first, second, and third — because the pivot is the base rate. Somewhere between 15–20% and 50% of a YC batch ends up doing something other than what it was funded for, and below $1–2M of ARR "you don't know if you have something." Martin's example: a founder took $4M raised on a generative-AI search product, concluded within three months that competition was too fierce, churn too high, and defensibility absent, and moved the whole company — which is now doing better. So you're not underwriting the plan; you're underwriting the person who will notice and move. (His podcast rule is the concrete version: he no longer records founders below ~$1–2M ARR, after too many asked him to pull episodes about businesses they'd left behind.)

• The Nasdaq-bell-ringer test. "Do you envision this person being this Nasdaq bell ringer in ten years? Do you see them being interviewed on CNBC?" It reads like vanity and isn't: the founder has to convince strong people to leave comfortable big-tech jobs, so star quality is recruiting ability. And the prerequisite is having been near excellence yourself — "when you've been in the room with Andy Jassy at AWS and he's grilling you on a project, you kind of know what excellence looks like."

• Timing gets its own line. Too early, on time, or too late — a variable he says he'd write a book about. Google was the seventeenth search engine to arrive, which is why he keeps asking himself whether he'd have written the check if Larry and Sergey had walked in in 1997. (Martin's answer to his own version of that question: Ron Conway's $250K got in, and the tell was that the product was an order of magnitude better on a single search — one query against Excite and AltaVista was enough to switch.)

• Name the kind of business before scoring it. Is there a differentiated wedge? A path to a control point — a system of action or a system of record? Which trajectory is this: distribution-first, research-first, something else? You can't judge how a company compounds until you've said out loud what kind of company it is — and the target is hundreds of millions in revenue, not single-digit millions.

• Ask "is this a feature in somebody else's platform?" — then ask whether it can escape. Bundling risk of the incumbents is an explicit feature in the model, and he suspects it's exactly why people passed on Google ("nice demo, kids" — Yahoo will just fold that into its catalog of the internet). Martin's counterweight: DocuSign was hard to fund because it looked like a feature Adobe could copy in five seconds, and the return came from years of unglamorous work — signing order, email workflow, integrations — converting a feature into a platform. The mistake isn't investing in things that look like features; it's investing in things that look like features and stay features.

• Ask for the ten-year vision and listen past the wedge. "Five or ten years out, you've built this thing and it's really successful — what does the future look like?" You want a vision that runs well beyond the initial wedge: the aspiration to "make a huge dent in the universe."

• Avoid consensus rounds by construction. "If you just do consensus investing, everything's gonna get bid up" — the round everyone wants prices at 100–200x forward revenue, sometimes infinite forward revenue because there is none. The job is finding "that sweet spot of kind of this sure bet that nobody knows about."

• Post-mortem the winners as hard as the losers. On a 100x, replay the deal as of the original date with only the information you had then, see what the model said (3.8), and tune toward catching more like it. On the shutdowns, convert the lesson into a gate. Both directions feed the same loop — and with 40,000–60,000 companies seen, he can compute an actual investor F1 score from true and false positives and negatives, "and see how good you personally are, and how good your AI is."

• Buy the tuition on purpose. "There's this phrase in real estate that it takes a hundred investments to make a real estate investor. In VC, they say it takes ten million to make a VC. I think that's really true." Volume is not indiscriminacy — he invests in about 1% of what he sees, roughly 50 of the ~800 YC companies a year — it's the only way the calibration loop gets enough data to run.

• One-sentence summary (his closing advice): Start building your own algorithm now — work with AI on every decision, learn from the history it can hold, and keep refining a personal weighted scorecard — because "AI is not gonna replace venture capitalists, but a VC powered by AI is a very powerful thing," and the only edge worth having is the sure bet nobody else knows about.

---

Episode 8 — Alex McNaughten

Spent his twenties teaching salespeople how to sell — a training and advisory practice that supported more than 130 New Zealand businesses, and two to three hundred companies across his career, before he turned thirty. Two profitable companies, a house a minute from the ocean, and about half a million dollars a year. He sold it, moved to San Francisco, and co-founded Grw AI (grw.ai) with Alistair McLeay and Daniel Ash, building Taylor, an AI sales teammate. Martin backed the bet. "It was a very asymmetric bet."

Most Compelling First Bet Stories

1. Selling the good life at thirty — trading a beach-view business for a San Francisco seed round

The scene Alex sets is deliberately unglamorous about what he gave up, because that is the whole point of the story. He had just turned thirty. Two companies, both profitable. A sea view and a one-minute walk to the ocean. Roughly half a million dollars a year, which in New Zealand "goes really, really, really far." He was working considerably less hard than he does now, for considerably more money, and spending a lot of time with friends and family. Then two things arrived at once: the AI wave visibly starting, and a reintroduction to Alistair and Dan — two people he had never worked with, "wicked smart, like a million times more technical than me," one of them just out of Cambridge with a master's in machine learning. The choice was not between a good option and a bad one. It was between staying on a trajectory that was already working and going all in on a new technology, with new people, in a new country. He sold the company, sold one of his two properties, and moved. His friends thought he was mad. His family, he says, has simply backed him in everything since he dropped out of college. His partner told him it was his decision and that she would move with him. The nagging counterargument was never fear of failure — it was opportunity cost: turn off the taps and how long until you earn like that again, and what would the invested capital have compounded into meanwhile? What settled it was a regret test rather than a spreadsheet. "The upside's so big here and I will regret forever not going after this." A year into San Francisco, he says the impact on the sales profession is already "many multiples bigger" than everything the services businesses did.

Why it's compelling: Most first-bet stories start from constraint — no money, no options, nothing to lose. Alex's starts from comfort, which is the harder version, and he is unsentimental about the mechanics that made it decidable. He mapped a genuinely capped downside (net worth pauses; go back to what you were doing; keep an interesting story either way), sized an upside measured in millions of salespeople, and then noticed something most people never get to say out loud: "it actually was a really easy decision. I didn't wrestle with it very long." Martin's observation lands on exactly that — the big decisions are almost always easier from the inside than they look from the outside, because the person making them has already done the downside arithmetic that the audience never sees.

2. The trial marriage — buying ninety days of information before betting a decade

Before any of that, there was a smaller and much cleverer bet. Alex wanted to work with Alistair and Dan, and he wanted to work in AI, so he framed the problem as risk reduction rather than conviction-building: "how do I minimize the risk of that going wrong and figure out quickly can we work together for 10 years and be in the trenches for 10 years?" The introduction came from someone he trusted deeply, which helped but was not sufficient. So he took a small piece of money from his existing company and funded what he calls a trial marriage — ninety days of the three of them actually working together, building things and solving real problems for his then-customers. It answered the question that no amount of reference-checking answers, and it paid twice: the work done in those ninety days became the new company's initial seed capital. The same instinct now runs the company's operating decisions. Grw's customers had all come inbound or by referral, so when they started spending on ads they put small experimental budgets across several channels, read the results weekly, and moved money toward whatever the data favoured. In hiring, a psychometric test — cheap, fast, and unglamorous — flagged a red flag on a candidate who was otherwise perfect: trusted introduction, excellent interviews, and, the profile suggested, someone who would not stay through the hard stretch. A follow-up conversation aimed at exactly that area confirmed it, and the hire didn't happen.

Why it's compelling: It is the single most portable idea in the episode, and it is the same move at three different scales — co-founders, marketing channels, hires. Buy a small, cheap option on information before you commit real capital, then decide. Martin's poker translation makes the economics explicit: as the fifth limper with terrible cards, you are paying a trivial amount for a one-in-five shot and, more importantly, for three cards' worth of information. Alex's version is the reason the life-changing bet in story one felt easy by the time he made it — he had already bought down the biggest unknown in it, which was not the market or the technology but the people.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Run the regret test, not the spreadsheet. The opportunity-cost math was real and he could do it — foregone earnings, a paused net worth, capital not compounding — but it never decided anything. The question that did was "will I regret not going big here?" measured against a safe and comfortable life he could describe precisely. When the two answers conflict, the regret answer wins.

• Map the downside literally — most people never do. "I think a lot of people don't do that, and maybe it seems a lot scarier because they haven't actually mapped the downside." Written out, his worst case was: net worth doesn't grow for a year or two, go back to doing what I was doing, so what? Fear survives on vagueness; the map is what kills it.

• Weight the downside by where you are in life. In your twenties or early thirties, a year or two of foregone compounding is "almost immaterial, really, when you look at it objectively." The same downside at a different age is a different number. Age is an input to the calculation, not a mood.

• Buy the cheapest information available before committing. The general framework of the episode, and the one he and Martin converge on. A ninety-day paid trial before co-founding. Small experimental budgets across several ad channels, read weekly, before scaling one. A psychometric test before a hire. In every case: spend a little to learn a lot, then make the real decision with better cards.

• Fund a trial marriage before the wedding. The co-founder-specific version. Reference checks and trusted introductions are inputs, not answers; ninety days of building something real together answers the ten-year question. The bonus is that the trial period can produce the asset — in his case the initial seed capital — rather than only consuming time.

• Use psychometrics as pre-hire information, not as a verdict. The test doesn't make the decision; it tells you where to dig. It surfaced a flightiness signal on a candidate everyone liked, and the value came from the follow-up conversation aimed straight at it — which established that she was the wrong kind of person for an early-stage company, "where, let's be honest, you're shoveling shit for a fair amount of the time."

• Hire salespeople who have sold without an engine behind them. A seller from Google is carrying Google's marketing machine, brand, and support; a seed-stage seller has a half-built product and no marketing budget. "It's a very different kind of person who could be successful doing the latter." Martin's harder-won version: prefer people whose first startup is already behind them — his one big-company hire walked in on day one asking where his assistant was.

• Play the cards actually in front of you. Raising with low revenue is the card you were dealt, so work out how to play that card rather than wishing for another. A crowded market means the table is full and everyone's in the pot — that's information, not a verdict. "You're always gonna have limited information, but you just try and play it the best way you can... and more information's gonna come on the next turn."

• Failure is inevitable; a lost hand is not a lost session. "You can't win absolutely everything, but every failure you learn from" — and every failure in his life has been followed by something good. The poker discipline is the operative half: pay attention to why you're losing, to the other players and the table, stay locked in for the session, and you'll do all right.

• Treat confidence as a function of reps. The identical decision felt enormous at company one — mortgage, roughly ten thousand dollars left — and comparatively easy at company three. "When you just have a pattern of things going right, the confidence just compounds." Useful in both directions: it explains why a decision feels hard, and it warns you not to mistake accumulated nerve for evidence about this particular bet.

• Look for the hole in the crowded category. The AI sales wave had gone almost entirely into top-of-funnel volume — more emails, more outbound — and his read from a decade inside the profession was that "the sales world just didn't need another email spam cannon." The untouched space was everything after: first conversation through close, and the customer relationship after that. Crowded is not the same as addressed.

• Discount the opinions of people with less information than you — then bring them along anyway. "Everyone around you is gonna have an opinion about what you should do, and at the end of the day, they have far less information than you." Make your own call. But the people whose lives your decision moves — a partner's job, friends, family — need the why, the picture of how it will look, and the opportunity in front of you. The decision isn't landed until they're on the journey with you.

• One-sentence summary (his closing advice): Accurately map the downside — it's almost always less frightening than it feels, especially young — make your own call despite everyone's opinions, and then take the people your decision affects along the journey with you, because "it's gonna make it a lot easier."

---

Episode 9 — Mike Ma

Ran marketing at Bank of America (heading digital strategy and innovation), ran brand at Vanguard, and was CMO at Betterment in its early robo-advisor days. Came to venture late — operating partner at Sway Ventures, managing director of Nex Cubed's HBCU Founders Fund, founder of a fintech accelerator — then wrote a couple dozen $5–10K angel checks to test a strange idea: work with founders for a month before investing. That idea became Sidecut Ventures, a "coach first, capital second" pre-seed fund with market-rate social-impact returns ("help the world and return the stupid fund"). Its rule: every founder works with the team for thirty days before a check, and about forty percent don't get one.

Most Compelling First Bet Stories

1. Elastic Energy — the founder who cut his pipeline in half and got the check for it

Ben Hilborn is a technical founder building distributed energy hardware — small boxes sold into solar installers and battery OEMs, "a blood and guts ground game" with a tiny team. He told Mike the thing most founders hide: "I need to learn how to sell." So the thirty days became a sales apprenticeship, from building a probability-weighted pipeline down to the absurdly specific — Ben had to have a beer with another CEO and pitch him, had never done it, so Mike had him grab a beer from the fridge and they rehearsed it on Zoom, two guys at a virtual bar. Then Ben ran one real sales loop during the engagement. Mike Slacked him to ask how it ended and got a stream of bad news: it was terrible, he'd just cut his pipeline by thirty percent — no, actually fifty — this deal wasn't qualified, that one had an undiscovered dependency, and then, dawning on him mid-message: "I can't believe I just told a VC I'm trying to seek money from that I just cut my pipeline in half." Mike's reply: "Ben, that's actually why I'm going to invest in you." Ben had walked in with the usual "bajillion dollars in the pipeline" deck, then showed with his own hands and feet how hard the sale really was, and explained every cut line by line. "Even though numerically you went down, you actually know." Sidecut wrote the check.

Why it's compelling: It is the cleanest possible demonstration of why Mike's process exists. No reference call, deck, or LinkedIn trawl would have produced this data point, because it only exists as a founder's live reaction to a bad outcome — and that reaction was the underwriting. The pipeline number got worse and the investment case got better, because what Mike is actually buying is a founder whose reports he can trust for the next eight years: "Now when you and I have a conversation, I know it's less likely you're full of crap." Martin's read: the founder's radical honesty about the true state of the business raised, rather than lowered, the investor's confidence to wire.

2. Melagen Labs — a small check on an eighteen-year-old, a bigger one two years later

Muhammad Hunain found Mike on a pitch show in New York and asked for coffee afterwards. He was eighteen, building an alternative to aluminum for radiation shielding for satellites and people in space — "better, lighter, faster, cheaper." Mike's honest opening was that he had no business being there: "I'm a go-to-market guy. I'm not a deep space investor." But he could read commercial paper, so he asked to see the LOIs. They weren't worth the paper they were printed on — no acceptance criteria, no theoretical payment, no timeline, "no counterparty risk." He said so and expected that to be the end of it. A month or so later Muhammad came back having gone to satellite makers' VPs of procurement and renegotiated every LOI to carry real counterparty risk. Mike's reaction was roughly "shit, yeah, they're better," followed by the arithmetic that mattered: at eighteen, Mike had been working at American Eagle Outfitters and teaching tennis, and here was a teenager who could sit in a procurement office and say "we need to renegotiate our LOI." He wrote a small angel check on that alone. Two years on, the founder is twenty, the company has well into seven figures of revenue, partnerships with big tech names, and a payload already delivered to launch to the International Space Station in December — and Sidecut has just wired a much larger, not-yet-announced round.

Why it's compelling: Mike had no edge on the science and did not pretend to; he underwrote the one variable he could observe — does this person do what the coaching implies, fast, with his own feet? — sized the check to that limited information, and then let two years of tracked performance justify the real one. Martin's summary is the mechanism: write a small check early, watch the founder respond to coaching and overperform over time, then write the bigger check with the additional information. It is also the moonshot-plus-pragmatist combination Mike says he wants in a single founder: a literal space bet, delivered by someone who fixes his contracts.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Coach first, capital second — flip diligence into thirty days of shared work. The first two meetings look like any VC's. After that, instead of references and data rooms, Sidecut asks "what do you want to work on together?" and spends thirty days in the mud on customer discovery, funnel, or product-led growth, on its own time. "At the end of 30 days, you like us, we like you, we'll write you a check." Most firms do value-add after the wire; he does all of it before, because that's when the information is worth the most.

• Get asymmetric information — know more than the rest of the cap table. Selfishly stated: everyone wants to know more than the market; he wants to know more than the other investors in the same deal. Thirty days of work shows how a founder handles ambiguity, conflict, and indecision — "things you can't find on a pitch deck or a spreadsheet." The founder gets the mirror image: a month of seeing what it's like to work with him.

• Move the toothbrush in — underwrite the marriage. The average US marriage lasts 8.6 years; the average stay on a cap table is about the same. So both sides should live together for a month first. The founders who want that are the founders he wants.

• Information is not wine — it doesn't get better with age. If you'd want to know something before writing the check, get it before, not after. References, decks, LinkedIn are past signal with a half-life; a chef's old Michelin star gets you in the door but doesn't guarantee the new restaurant. "I don't want to cook with frozen vegetables." His sharpest version: founder muscle memory from three years ago, pre-AI, may actually be negative signal.

• Action-oriented self-awareness — "don't tell me, show me." Founders are trained to be persuasive, so words are cheap; he wants to watch the founder do the thing with hands and feet, because after the wire he'll see maybe one percent of what they do. Self-awareness half: a tiny pre-seed company selling into a Fortune 500 with a hundred thousand employees needs to understand what it's actually walking into. Action half: build the pipeline, run the sales loop, have the beer — during the thirty days, where he can see it.

• The deck is precisely what will not happen. "It's the only thing I know." At pre-seed you can't invest off a spreadsheet, so the thing to study is what the founder does when things go sideways — which the thirty-day window reliably surfaces.

• Honesty that lowers the number raises the conviction. A founder who cuts his own pipeline in half and explains every cut is more investable than one with a "bajillion dollars" of unqualified deals — because it proves the future reporting can be trusted. If you really had a bajillion in pipeline, why are you asking for a quarter million? Go sell.

• Killers for good — and killers means go-to-market. The one non-negotiable: "I want killers for good. Go-to-market killers. And good is the impact." If thirty days don't produce conviction that the founders are killers, he's out, regardless of technology.

• You may be right — but I can't underwrite faith. The climate deep-tech pass: four brilliant PhDs who had never sold, a four-to-five-figure pilot, a multi-million contract hoped for, and a corporate VC sponsor whose commercialization path amounted to "just trust us." Having run innovation at Bank of America, he knew that isn't how large companies work. Their scientific conviction might prove correct; it still wasn't something he could price. The round oversubscribed anyway, and that's fine.

• Pick one motion — you're not funded for both. Small local customers for fast learning, or Fortune 500 enterprise — the same team said "both," and at their raise that's a no. Whichever they choose, hold them to what it actually requires: for enterprise, live at the customer's cafeteria until people think you work there.

• Test the counterparty risk in every LOI. A letter of intent that carries no acceptance criteria, no theoretical payment, and no timeline is a letter that risks nothing. Send the founder back to get one that does — and watch how fast they return.

• Write the small check on what you can observe, then size up on tracked performance. He couldn't underwrite space materials science, so he underwrote an eighteen-year-old's response to coaching and wrote an angel check to match. Two years of overdelivery earned the real one.

• Say no after thirty days about forty percent of the time — and write the pass as coaching. The bar stays high even after a month of sunk effort; there's no single metric, "it just comes in." The pass letter is long and instructive, not an admonishment — and founders have come back to report meetings one, two, and three went better because of it.

• Want both the moonshot and the pragmatist — and run a portfolio that mixes them. Asked whether he prefers the audacious founder or the under-promise-over-deliver one: "Yes. I want both." Some bets are shorter-horizon and de-risked at his entry price; some are to-the-moon; the portfolio is deliberately built to give LPs both exposures.

• Test the model with small checks before you build the fund. His own first bet on the coach-first idea was a couple dozen out-of-thesis $5–10K angel checks while still deploying for others — mostly to learn whether founders would even give him the access. They did, they loved it, SPVs followed, and only then did he raise a fund around it.

• One-sentence summary (his closing advice): Get the information before the check, not after — work alongside the founder long enough to see what they do rather than what they say, and only back the ones who show you, with their own hands and feet, that they're killers.

---

Episode 10 — Howard Lindzon

Invented the cashtag — the dollar-sign ticker convention Twitter adopted and every financial platform now uses — and founded StockTwits in the 2008 financial crisis. Seed investor in eToro (2010) and Robinhood (2013, a $100K check from a $6M fund), later led the recap of Alpaca; founder and general partner of Social Leverage, now on its fifth fund. Investing thesis: he talks to a million retail traders a day on StockTwits, so he knows what the "degenerates" want before Silicon Valley does.

Most Compelling First Bet Stories

1. Robinhood — $100K into the idea Silicon Valley hated

The setup took Howard fifteen years to assemble. He came up in the 1999 retail-trading era, when everyone worked off twenty-minute-delayed quotes and thought Yahoo Finance was a miracle; the professionals saw everything twenty minutes first, and retail got its head handed to it in the 2000–01 crash. When Twitter arrived, nobody was thinking about finance — except Howard, who saw that information latency had just gone from twenty minutes to zero. That was an exponential change, and it was the reason he started StockTwits: if the plane lands on the Hudson or bin Laden is caught, a properly used Twitter feed now beat the institutions. But the plumbing never followed the information. Web 2.0 was "build it, worry about the law later," and that works for Uber; it does not work when the SEC can shut you down, so nobody in the Valley wanted to build a brokerage. Howard didn't either — "I was too wimpy, not a good enough entrepreneur" to take on FINRA — so in 2008 he went to Jack Dorsey and Ev Williams, via Fred Wilson, and told them to stop selling ads and become the transaction machine: click a cashtag, trade. Fred called it genius; Jack and Ev told him to build on their API. Seven years later, still no one had built a mobile brokerage.

Then Vlad Tenev and Baiju Bhatt called, out of money. Their company, Chronos Research, had raised about a million dollars trying to knock off StockTwits, pivoted a bunch, and now had one idea left: a mobile-first brokerage. Howard flew to Silicon Valley and their designer, an ex-Facebook hire named Joe, showed him the app — not connected to the SEC, not connected to FINRA, just the design of a trading app. Howard's reaction: "If you build this, this is huge — but you gotta go through FINRA, you gotta go through the SEC." They weren't scared. They were Stanford math guys, and they said they'd build it. He called his partner Tom from the plane and said do the maximum we can do. The fund was $6M; they wired $100,000 at roughly an $8–9M valuation.

What made it a first bet rather than a consensus one was that Silicon Valley actively hated it. Every VC had committed to building a better Vanguard — Wealthfront, Betterment, assets under management — because they'd been burned on trading and believed the world didn't need another E*Trade. Howard knew this firsthand from failing to help eToro raise money in the Valley in 2010. His thesis was one number: Schwab was spending about $150 in marketing to acquire a customer, and in that era a good app grew for nothing because people showed it to each other. "If we get to a million users and it costs us zero, it's a $150 million company — Schwab will just buy it to cut the marketing costs." He wrote that to his LPs. He was dead right on the mechanism and wrong only on magnitude: Robinhood was a billion-dollar company by around 2015, when a billion was unheard of outside Uber and Airbnb, and is, in his telling, a hundred-billion-dollar company today. His one regret is selling a chunk between one and ten billion.

The second act is the term sheet gambit. Six months in, Vlad and Baiju wanted to raise at what was then a crazy $60M valuation, and Social Leverage — between funds, with about a million dollars available — would have needed roughly $8M to lead. Howard called Fred Wilson: what do I do? "Write a term sheet, you'll figure it out." Over the July 4th weekend he and Tom sent a term sheet for $8M at $60M with a sixty-day close, planning to beg for an SPV. The founders shopped it, as they should have, and Index Ventures' Jan Hammer wrote a term sheet off theirs with a nine-day close. "All of a sudden we were screwed" — except that by having written the paper, Social Leverage had earned a carve-out: Vlad and Baiju protected them for about a million dollars in the round. They ended up with roughly $1.2–1.3M in across rounds, plus SPVs along the way.

Why it's compelling: It is the series' purest example of a low-information bet that was also, in Howard's phrase, the most-information bet: "I'm supposed to have made that investment based on everything I had done before." He had watched the retail problem since 1999, built the social layer himself, pitched the trade button to Twitter in 2008, and backed eToro in 2010 — so when two broke founders showed him a disconnected design mockup, the missing information was already in his head. The Valley's "no" wasn't an opinion he weighed; it was the reason the price was $8M. And the term sheet story shows what a small fund does when it can't afford its own conviction: commit first on paper and let the paper create the option.

2. Alpaca — pass at YC's price, lead the recap eighteen months later

Howard is a consumer investor, not a tech investor, but Robinhood taught him something about plumbing. Robinhood was a lightweight app built on Apex, the old clearing firm, and the plumbing was terrible; when a US financial company gets big enough it builds its own clearing, which is exactly what Robinhood did. So he started hunting for the tech-plumbing brokerage — "what would be the pipes for the next 500 Robinhoods?" — and got an intro from Ed Sim to Alpaca. Yoshi Yokokawa's team, around 2015–16, was building AI trading bots: really smart, really early, and not for him. But he told them that if they built the plumbing Robinhood needed, there was an opportunity, because "there's gonna be a thousand Robinhoods and everybody's gonna need this." They came back having understood it, went to YC, and YC gave them $3M at a $17M valuation. Howard nearly had a heart attack. He was very price-sensitive, he's "an anti-YC guy," and he said good luck.

About eighteen months later Yoshi called back: they were out of money and hadn't built the right product. Meanwhile something had changed on the ground. Young developers were playing with Robinhood and with pipes-based businesses, and one of Howard's LPs' kids called him to say he was building trading tools on Alpaca and the product was great. Howard checked in with Yoshi, confirmed the situation, and said: "Listen, we're the right investors." Social Leverage led a recap — something they never do, "almost like private equity" — with True Ventures, a $6M round at about a $12M valuation. Yoshi, to his credit, fell on his sword, went back to every existing investor, and got the company priced right. Howard then structured the table: Santo Politi of Spark Capital, who had done eToro with them and knows financial markets cold, and later Portage, the Canadian fintech fund that is also a Social Leverage LP. Seven or eight years on, in Howard's telling, Alpaca is doing roughly $250–300M annualized, in a hundred countries, powering three to four hundred brokerages including Binance and Kraken for US stocks and options, and last raised at around a $3B valuation — "on the path to be a hundred-billion-dollar company. Because there's gonna be a thousand Robinhoods."

Why it's compelling: The thesis never changed — only the price and the proof did. Howard was right about the pipes in 2016, walked away from a valuation he couldn't justify, and was handed a second look at a fraction of the price with a real signal in hand (a kid he trusted building on the product). His honest caveat is the useful part: passing on price "was a mistake in general," and he isn't good at recaps, which is why the decision leaned on a partner with patience and a repeat co-investor with the chops to clean it up. Martin's read: the confidence came from having built the entire information stack — StockTwits, eToro, Robinhood's Apex problem — so that a re-priced entry plus the right people at the table made the call "hard not to do."

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• "You're supposed to make this investment" — the obligation test. Low information and most information are not opposites. When a deal lands in the exact spot your whole career has prepared you for, "if I don't do that investment, it's almost worse than having done the investment." Part of the job is making sure that when the bets you're supposed to make come along, you make them — even with thin data — because that's what the previous fifteen years were for.

• Read the era before you take credit. ZIRP, the cloud, the iPhone, open APIs, zero customer-acquisition cost, everyone friendly — "it was the easiest time to ever have made money." He insists on separating the setup from the skill: the era did most of the work, and a framework that doesn't account for that will misprice the next one.

• Reduce the thesis to one number you know and the crowd doesn't. Schwab paid $150 per customer; a good app in that era acquired customers for zero; one million users at zero CAC was a $150M company Schwab would buy for the marketing savings alone. He wrote it to his LPs before the check. Being right on the mechanism and wrong on magnitude is the shape of a great early bet.

• Hunt where the consensus has committed elsewhere. Every VC had bet on "a better Vanguard," and even a better Vanguard isn't ten times better on margins that thin. The trading idea wasn't just unfunded — it was hated — which is why a pre-product design mockup was available at $8M. Silicon Valley's collective allocation is a map of where not to compete.

• The plumbing comes after the app — so bet on the pipes for the next 500. Watching Robinhood outgrow Apex taught him that once the app wins, it builds its own clearing; the next wave of apps still needs someone else's. "What would be the pipes for the next 500 Robinhoods?" turned the pain of one portfolio company into the thesis for the next.

• "If you build it, they will come" — but only where you actually know the customer. He talks to a million people a day on StockTwits, so in Robinhood and Alpaca he knew who the users would be. He's explicit that this doesn't generalize: "I don't have expertise across a hundred different subjects," and he did not see prediction markets coming.

• Write the term sheet, you'll figure it out. Fred Wilson's advice when Social Leverage wanted to lead a round it couldn't fund: commit on paper, then go find the money. The term sheet got shopped and beaten, but its existence earned the carve-out. Writing the paper creates options that waiting for capital never does.

• Price discipline, honestly audited. He passed on Alpaca at $17M post and got a second chance at $12M — but he calls the price sensitivity "a mistake in general." Don't let a good outcome from a pass convince you passing on price is a strategy.

• Take the recap when the thesis is intact and the founder falls on his sword. Recaps are "almost stupid," private-equity work, and not his skill — so the conditions were strict: the original thesis unchanged, a live signal from a user he trusted, a CEO willing to go back to every investor and reprice, and a partner (Tom) and co-investor (Spark) with the patience and chops to do the cleanup.

• Stay in your lane, then stack the table with repeat co-investors. Like Fred Wilson, Social Leverage stayed under $100M so a $1–2M check still affects the outcome, and it can work with the same people over and over. Founders know that if they execute, the next rounds will come from partners who've done it together before — Spark from eToro, Portage as an LP.

• Discover your risk score — it's DNA, not a setting. The only way to learn your risk profile is "to lose money or make a lot of money." Kids track sleep scores and readiness scores but not this one; at some age it gets etched in, and you should own it rather than fight it — mingle with other scores, but set up your financial life to yours.

• Open the account and play — with $10 if that's what you have. Fractional shares mean a real portfolio costs almost nothing, so the earliest lesson is "red, green": how you actually feel when it moves. The question isn't whether you lose 30%; it's what you do when you're down 30% — chase it back or slow down.

• It's what you do after the first big mistake. Everyone blows up a first $100 or $10,000 account. "It's making four big mistakes in a row that really gets you in trouble." The next phase for every new trader with access to the pipes is risk management and position sizing — and a mentorship group around them to say, this is a normal reaction, how did it make you feel?

• Draft behind the great white shark — and be careful whom you follow. He is not winning the Tour de France and he is not the great white shark; there's a lot of money and time saved right behind the leader. So manage your ego, get as close to the great white sharks as you can without getting eaten, be humble and nice, and after ten or fifteen years the alpha they leave on the table drips to you. Two warnings: don't get in front of them, and the biggest mistake is "your friend did it, so you're gonna do it."

• Domain experience over generalists — back people who'll work the problem for life. "I'm not looking for generalist founders." StockTwits took him a decade; he wants founders who will be trying to solve this one problem for the rest of their lives. Free mentorship is everywhere now — it took his generation twenty years to find a mentor and takes a kid two months — but that raises the starting point, not the required depth.

• One-sentence summary (his closing advice): Build some domain experience, be very careful whom you're following, and find your risk score early by playing small and losing real money — because access to the pipes is now universal, and the only edge left is knowing yourself and your customer better than the crowd does.

---

Episode 11 — Sohun Sanka

Twenty-five years old and six years deep in TikTok Shop, seen from all three sides of the table — brand, agency, and platform. Built one of the first scaled TikTok Shop services inside a nine-figure omnichannel marketing agency, then joined Euka, the first TikTok Shop affiliate software, as head of growth at twenty-three, and ran go-to-market as employee number one at two competing TikTok Shop SaaS companies. He then walked out of a rising comp package, lost his co-founder before the company existed, and founded Clankers — a services business that teaches TikTok Shop brands and agencies to build their own internal tooling with coding harnesses instead of buying more software. No outside capital; living at his parents' house to hold burn near zero.

Most Compelling First Bet Stories

1. Leaving the agency at 23 for the seat that could see the whole category

His TikTok Shop career started by accident. He was at an omnichannel marketing agency doing over $100M a year managing brands on Amazon and Walmart, and TikTok Shop "fell in my lap because they needed someone younger to build out more of an influencer focused marketing engine." The platform had just launched in the US and he had no idea it would matter; the CEO did — "this is going to be the next big thing" — so Sohun worked directly with leadership and built the practice out, making "a ton of mistakes at the beginning" he still winces at, and ended up one of the first agency strategists anywhere running a scaled TikTok Shop service for nine-figure enterprise brands.

That work made him visible at exactly the moment the category was hiring. Because TikTok Shop was so new, "everyone was looking for a head of TikTok Shop or a TikTok Shop leader," he had offers, and he was on track to become a manager where he was. Instead, at twenty-three, he took a head-of-growth job at Euka — the first TikTok Shop affiliate software, essentially a CRM for finding and working with TikTok Shop creators, and at the time a small company in a small industry.

The reason wasn't the title or the money. It was the vantage point. Every brand entering TikTok Shop had to pass through that software to start its program, and he could see that: "if I can go and be at the nexus of where brands incubate and then scale, I can get so much perspective... I can see the back end of every single business and start to really understand the system." He was buying the seat that would let him make his next decision — about where TikTok Shop was heading — with information no single brand or agency could assemble.

Why it's compelling: At the moment of maximum market demand for the skill he already had, he traded the obvious move for an information position, and took the pay and title risk to get it. The bet paid in a currency that doesn't show up in comp: a back-end view of the whole category, distribution to everyone in the space (in an agency "you're just competitive with other agencies"), and what he calls a kitchen cabinet of founders and advisors. Every later bet he makes is underwritten by what that seat showed him — including the one where he leaves the industry's payroll entirely.

2. Quitting for a co-founder who never came — and starting anyway

Compensation kept climbing at each software company, "which made it even harder and harder to leave"; across the moves he estimates he has given up on the order of a million dollars in total comp. Two reads pushed him out anyway. The first was technological: with Claude Code, Codex and the other harnesses arriving, the moat around pure SaaS in marketing had gone very thin. E-commerce software isn't agent-to-agent enterprise complexity — as a friend of his put it, these are "form to table" products, you put stuff in a form and it outputs a table — so anyone can now mimic the logic and rebuild them. The second was the flip side: if software is cheap, the future is hyper-personalized internal tooling pulling from public APIs and MCPs, which was precisely his own pain point as a non-technical marketer stitching together other people's products.

What nearly stopped him was the co-founder question. He believed every founder needed a good one, and he had no high-school-or-college partner to call. What he had was a family friend introduced by an uncle who had been a good advisor — enough to mitigate the risk that the guy would "maliciously stab me in the back," and nothing more. They planned a SaaS: an orchestration layer that surfaced the gaps in a brand's marketing program, MVP a glorified dashboard APIing into Google, Facebook and TikTok Shop. Sohun, not technical, had no way to evaluate the man's technical acumen — and he knew exactly how much that mattered, having worked at two directly competing software companies where the difference between the two CTOs changed the product, the go-to-market and the customer experience, a difference it took him ten months on the inside to even see.

He refused to wait it out: "I don't believe in delaying a year to figure it out." He told his co-founder he was quitting in two weeks. The co-founder said a month, then burned out. The SaaS died before it existed and Sohun was already unemployed — and building software the conventional way was closed to him, because "when you think about going into software, if you want to build anything complex you have to go find some sort of institutional capital and raise, because I don't have 500k to a million in my bank account." To raise he would have needed a team, a thesis, and a co-founder to hedge the risk on. He had none of it.

So he re-asked the question in cash terms: "what is the easiest thing I can do today that I have the best advantage on, that is highly profitable and will put cash into my pocket?" The answer came from the seat he had bought two years earlier. Nobody else had been head of go-to-market inside TikTok Shop SaaS — an industry only three years old — and run agency services. That combination meant he could take a product apart end to end and explain it to a developer, a designer or a buyer; he simply couldn't build it. So he wrapped the capability in a service: treat the existing softwares as data layers, and teach brands and agencies to build their own internal tools with Claude Code. Total cost to start: a Claude Code subscription, Figma, Slack, and about $500 to form an LLC through Stripe Atlas.

The downside math is what let him move that fast. Living at his parents' house held burn near zero, the startup cost was a couple thousand dollars, and the fallback was concrete: "could I reasonably get a job in two years and get paid even more than what I was making now? And I was just like, yeah. So virtually it was a zero risk decision on my part." Demand wasn't hypothetical either — as a TikTok Shop thought leader on LinkedIn he already had people asking him to consult, latent demand he had been "vesting up" through his own brand equity for years.

Why it's compelling: This is the rare first-bet story where the premise dies after the chips are already in. He quit on a plan that required a co-founder, lost the co-founder, and found out in real time whether the bet had been on the plan or on himself. What made that survivable is that he had engineered the downside before he needed it: no rent, no dependents, $500 of formation cost, and a labor market that would pay him more in two years than it did the day he left. Martin's read is that this is Alec Torelli's framework from Episode 1 transposed into a different game — the door you walked out of stays open at the same rate, so the real cost of the attempt is time, not money. Sohun's addition is the timing argument: take the shot while nobody depends on you, because a spouse, a mortgage and obligations to other people don't just change your appetite for risk, they change your likelihood of succeeding.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Take the seat with the best view of the system. When a category is too new to have a right answer, optimize the next decision rather than this one: go where the whole market passes through and the back end of every business is visible. The nexus seat beats the bigger title because it is the only place the system becomes legible.

• Move when the moat moves, not when the job gets worse. He left a compounding comp package because coding harnesses had made "form to table" software trivially cloneable — "is this industry really the industry that I want to stay in for the long run?" The trigger for a change of position is a change in what is defensible, which arrives long before the paycheck notices.

• Abstract the job before you take it, then price only the residual. Running a podcast interviewing founders, plus being employee number one — drafting NDAs with lawyers for partner CEOs, sitting in on hires for departments outside his own, sourcing leads with no paid budget — meant he had already done most of the founder job. So he inverted the question: what are the things I'd have to do as a founder that I'm not already doing? The list came back short (invoicing, legal, accounting, forming an LLC) and every item was purchasable — roughly $500/month for a back-office firm, $500 for Stripe Atlas. A decision gets easy when the unknown part is small and has a price tag.

• Interrogate every "you need X" — ask which constraint it was solving. A mentor pushed him on the co-founder orthodoxy: why do you need it? His own answer was that software is hard enough that focus splits across product vision, design, managing devs, marketing and selling, so you must be hyper-specialized to compete — which means the requirement only holds while building is the hard part. If building gets easy and buyers' decision-making becomes the bottleneck, that's a go-to-market constraint, and the rule dissolves. Received requirements are compressed answers to old constraints; decompress them before obeying them.

• Cap the downside structurally, then check the second bite. Not "how brave am I" but "what is actually at risk": burn near zero by living at home, a couple thousand dollars of setup, and a re-entry job in two years at higher pay than he left. With those three in place the decision stopped being a risk calculation at all.

• Take the shot inside the low-obligation window. "It's time to go all in while I have time. If I have a spouse in the future, if I have a mortgage, if I have other obligations to other people, those things are going to affect my likelihood to succeed." Obligations don't only change your tolerance for risk — they change the odds, so the window itself is an asset that depreciates.

• Bank latent demand before you need it — and notice whose brand the leads come from. He brought leads into seed-stage employers with his own LinkedIn following rather than the company's: "that's why these companies are giving me equity." Demand you generated personally is portable, and the fact that it's portable is the tell that you're already carrying founder risk. The corollary he acted on: if you keep posting, the pipeline exists before the company does.

• Don't delay a year to resolve what only time can reveal. He could not assess his prospective co-founder's technical ability at face value — it had taken ten months inside a company to see what separated one CTO from another. Rather than stall for information that only arrives by doing, he started and let reality answer it; it answered in weeks.

• When the plan dies, ask the cash question. Not what is most ambitious, most defensible or most fundable, but "the easiest thing I can do today that I have the best advantage on, that is highly profitable and will put cash into my pocket." It converts a collapse into a next move quickly, and cash flow buys the time to be ambitious later.

• Spend 80% of your time on the single constraint — and do it by hand. "If the constraint of your business is sales, are you spending at least fifty-five hours a week?" And if you don't know how to sell, it can just be DMing every person Claude says fits your customer persona — manually, on purpose. His warning is about the failure mode that looks like work: "because things can be automated, people spend too much time looking for ways to automate things, or if it's not automatable, they kick the can down the road." Nobody actually sends warm outreach to a thousand people a week, which is why it remains the cheapest and most predictable way to book a call.

• The eighteen-hour filter — protect momentum from cognitive split. Anything that doesn't move "more customers" or "happier customers" within one order of magnitude in the next eighteen hours gets ignored. He turns down creator deals, referral revenue and SaaS advisory roles on that basis: "even just one to two hours away from this can completely derail momentum in one direction." Focus, in his framing, is what lets a smart person scale almost any business — roughly seventy concentrated hours a week with an agent alongside, and "there is literally nothing you can't get done."

• Anchor to a stack of proof, not a feeling. When the "can you actually do this" doubt arrives, he looks tangibly at the pipeline and the content he has shipped, then asks what has materially changed since the last time it worked: has the algorithm shifted? have five better competitors entered and targeted exactly my customers? "Usually that isn't the case." His certainty is compounded reinforcement rather than self-belief — which is also why he favors work that returns feedback in weeks: "I might not be as perseverant as other entrepreneurs, but I've been fortunate enough to do work and get feedback within a couple of weeks or a month that what I'm doing is right."

• One-sentence summary (his closing advice): Get your downside to near zero and take the shot while nobody depends on you — then say no to everything that doesn't move your one binding constraint in the next eighteen hours, because at this point focus, not capability, is the scarce input.

---

Episode 12 — Kirby Winfield

A four-time founder with two IPOs and two acquisitions who then decided that wasn't enough and raised funds to make the bets for other people. Founding general partner of Ascend.vc in Seattle, now deploying out of fund three; before that, the founding investor in residence at the Allen Institute for AI, where two years among AI research PhDs taught him how to tell a researcher from a company builder. An English major and self-described "recovering marketer" who writes inception-stage checks almost exclusively into technical founders — and who, in 2022, backed a pre-incorporation defense hardware company after telling his LPs, in public, that he was not a hardware investor. That company, Overland AI, has since raised $100M and won a $20M Marine Corps production contract.

Most Compelling First Bet Stories

1. Overland AI — breaking his own two stated rules on a company that didn't exist yet

In 2022 Ascend had just closed fund two on a thesis Kirby was pounding the table about: enterprise AI, "SaaS 3.0," AI-native vertical software. The consensus at the time was that vertical AI was a thin wrapper on somebody else's frontier model; his bet was that far more value would accrue to AI-native software than to the three or four labs that would win the model layer. Alongside that thesis he had made two proclamations "to our LPs and anyone else who would listen": he was not a hardware investor, and defense tech was not a sector he had considered at all.

Then he was introduced to Byron Boots, who was still running his robotics lab at the University of Washington. Boots and his team had built software that let vehicles navigate off-road, running on off-the-shelf sensors and fabricating a little of their own hardware to bolt onto a chassis. Every part of the setup argued against a check. Vehicular autonomy was widely considered solved — the self-driving companies had been at it a decade — so the idea that you needed a new class of models to navigate unmapped, dynamic battlefield terrain was itself a new idea. It was the only investment Ascend had ever considered that had hardware in it. It was defense, in a city where, as Martin put it, anything that smelled like defense or public safety "wasn't gonna win you any pats on the back from your local LPs." And the introduction came with a scar: it had reached Kirby through a friend who had been advising Boots, who had been introduced to him by a prominent Seattle venture investor who passed.

Three things got him to yes. The first was borrowed technical diligence he could actually trust: the referring friend was Rob Coneybeer, co-founder of Shasta Ventures and a serious engineer in his own right, who had already spent time with Boots and handed Kirby most of the technical work up front — "as a non-technician, you know, I'm an English major and a recovering marketer. I'm always relying on folks in our network to help me get smart about those things." The second was the AI2 years: he had a framework for interrogating a PhD to find out whether the person could build a company rather than only a technology, and he used it on the life question rather than the science one — what did Byron actually want? Was this innovation for innovation's sake? Something to hand off? Or a generational company? The surrounding signal was strong: Boots had been sending his PhD students downstream to Cruise and Waymo for years, so he would be a talent magnet, and the team was winning the DARPA RACER challenge, so the technology worked.

The third was first principles, and it's the part Kirby keeps coming back to. "How many vehicles does the Army, Navy, Marine Corps or whoever deploy on the ground? Hundreds of thousands. Some of those are very expensive, and many of them are being deployed by warfighters in combat situations where there's real human lives at risk." Big enough, urgent, and greenfield — nobody else was doing it. That cleared his first qualifier: if it works, does it matter? And looking downstream he could see, faintly, that capital was starting to want capital-intensive defense businesses — Anduril at the time was worth perhaps $2B, not yet the proof it would become. "I don't think it was obvious in 2022, but you could kind of squint and see maybe a path to doing something special."

Why it's compelling: He broke two publicly stated rules, in a sector his own LP base was cold on, behind another investor's pass, on a company that had not been incorporated when the money went in — and the thing that could have ended it was a race: "they could have lost the DARPA RACER challenge and been out of business." What makes this a clean first bet rather than a lucky one is that he can name precisely which risk he was taking. He wasn't underwriting whether the technology would work; Coneybeer's diligence and the RACER results answered that. He was underwriting whether an academic wanted to become a founder, and whether a market everyone in his city found distasteful would turn out to be real. Both of those are questions about people and timing, which is exactly where he believes his edge lives.

2. The converged-security bet — writing a check that made him feel sick, on purpose

The second story is live, which is what makes it useful: nobody knows yet whether it works. Ascend recently backed a company doing converged security — physical security for critical infrastructure, meaning human guard teams working alongside drones and an AI-enabled software stack.

On paper it fails almost every screen an analytically minded investor would run. It is a solved problem, and has been for twenty years. There are massive incumbents already doing it, and they are not worth very much, which means the top-down market size doesn't compute. "A quant VC coming in and looking at that opportunity, we just go, TAM's not there." Kirby's counter-read was that the space had never been disrupted, the macro tailwinds behind it were enormous, the threat surface is growing, and the founder — a former CTO at NASA and Los Alamos who had raised venture for a cybersecurity company before — was right about the problem. And even holding all of that, the risk he could not eliminate was not technical or commercial but financial: "they could do everything right and then still not raise money."

So he made the discomfort itself the deciding input. The framework he says he took out of Overland and now applies more frequently is the sickness test: do I feel sick to my stomach making this bet? The answer he wants is yes. "I want to feel sick. I want to feel nervous. If it's obvious — if I'm doing a neo-lab investment in 2026 — it just feels too easy." Martin's restatement is the plain-English version: you want to be a little uncomfortable when you write the check, because there are plenty of people telling you you're stupid, you're crazy, it's already solved.

Why it's compelling: It is the rare case of an investor naming his own emotional state as a variable in the model and then inverting the sign on it. Most people treat nausea as a signal to slow down and gather more information; Kirby treats its absence as evidence that a deal has already been priced by everyone else. Paired with the first story, it shows the same underwriting logic applied forward instead of backward: Overland looked wrong in 2022 for reasons that turned out to be reflexes rather than analysis, and converged security is his deliberate attempt to find the next thing that only looks wrong. The honest part is that he says out loud what the failure mode would be — a company that executes and still can't raise — rather than talking himself into believing the risk isn't there.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• "If it works, does it matter?" The first qualifier for any idea, run before market sizing or diligence. Assuming you love the founder's background and they fit your archetype, the next question is simply what the ceiling is — is there a ceiling at all? On Overland the answer was hundreds of thousands of ground vehicles, human lives at risk, urgent and greenfield, and that settled it. The corollary is a job description: "our job as VCs is not to be too worried about if it doesn't work. If it does work, can it be the next Anduril?"

• Do I feel sick to my stomach? Treat discomfort as the entry signal. If a bet feels obvious it's too easy and has already been priced by everyone else. The specific worry he wants to feel is not "will they execute" but "am I worried that two or three years from now, the thing they're working on just won't be backable?" Absence of that feeling in 2026 is, for him, a reason to look again rather than a reason to proceed.

• Fund enough powder to turn the cards over — then ask whether aces are enough. The poker version of round construction: "what cards do we want to see turn over?" The round has to be big enough to actually flip the specific cards that answer the thesis — no hand-to-mouth $100K increments — but the second half of the test is the one people skip: if the cards come over aces, is that enough to get the next round? A milestone the next investor doesn't care about isn't worth funding.

• Clear milestones are easier to underwrite than blurry ones. Software-enabled hardware for defense behaves more like life science than like software: tranched, gated, with milestones far clearer than horizontal AI's. Investors who cut their teeth on SaaS are trained to be comfortable with blurry milestones, so the explicit, sequenced kind should read as comforting rather than intimidating — the work becomes underwriting whether this team hits those marks.

• Underwrite research risk, not execution risk — which is why you back technical founders. "I'm willing to underwrite technical research risk. I don't want to underwrite technical execution risk." With the right technical team, the product almost always gets delivered and almost always works. The point of that is to control the cause of death: "we want to be in the business of backing founders where the technology isn't what kills the company." What should be allowed to kill it is being too early, missing problem/solution fit, or a problem that turns out smaller or less urgent than believed. Non-technical founders doing technical things with a technical co-founder, or technical founders without the pedigree, is "a tough road to hoe."

• The academic-founder screen: ask what they want in life, not what they've built. For a lab spin-out, building technology and working in academia is very different from raising venture and scaling a multi-billion-dollar company, so the diligence is a conversation about ambition: is this innovation for innovation's sake? Something you want to hand off? Or do you want to build a company, potentially a generational one? Martin's version is the same gate from the other side — for most spin-outs the answer is that the founders just want to keep being researchers, and that's a no. The rest of the screen is four properties: a big enough vision, some secret knowledge, being a talent magnet, and projecting to be able to raise from tier-one venture — because a $5–10B company will need that money.

• You're investing in the company they'll build, not the company you'd build. His earliest mistakes came from reading founder-level disagreements as red flags: I'm not sure I agree with how they're positioning it, I wouldn't go to market that way. "You're not investing in the company that you would build, you're investing in the company that they're gonna build. And if that's what you're investing in, then really you're just investing in the founder." The practical bar is now close to a floor rather than a filter — as long as the founder isn't doing something that violates the LPA and has the four properties above, the disagreement is a category error, not a signal.

• Multi-stage participation at pre-seed is an optionality check, not signal. "They'll tell you that the partner is gonna be on the deal, and then three months in it's the associate." Taking comfort from a big name on a seed cap table is a mistake he has made and now refuses; sometimes it's a negative signal. The corollary is expectation management on the way up: don't get too excited when a company lands a tier-one Series A, because their venture math works exactly like yours — a firm writing twenty Series A checks out of a fund projects two of them to return it.

• Decide naked — don't wait to see who else is in. He tries to get his check in on the note without knowing who else is investing. "I do care. We all care. But I don't want to know. Because if I feel like I need to know who else is investing, then I shouldn't make the investment — then I'm not doing my job, I'm not doing what I'm special and good at. It's my taste, it's my gut, it's my algorithm. And the minute I start outsourcing that to someone else, I'm doing my LPs and myself a disservice." He's candid that it's hard: the morning after the wire you wonder whether anyone else will back this company, and whether you're insane. Sometimes they get into YC two days later and raise at 3x; sometimes it's a couple of angels and a lot of work ahead.

• Outside the Valley, in AI, do your own competitive diligence. If you're backing a team that isn't in the Bay Area and the space is AI, find out exactly who else has been funded doing anything that smells like it. If you aren't one of the first three or four funded teams in a category, you're in a red ocean outside the echo chamber, and it becomes a capital-raising problem regardless of quality — "you could be better than the other teams. It doesn't matter, half the time." And you have to do the work yourself, because the founder will always tell you nobody else is doing it. Martin's addition: he is routinely astonished by how many founders have no idea who else is building in their space — he'll hear a pitch he has heard ten times, told it's the only one of its kind.

• Be smart about the founder, not about the field. His single sharpest correction, and the source of his anti-portfolio: "It's not my job to be smart about what you're doing. It's my job to be smart about you and your capacity to build something meaningful." His biggest misses — data center energy management, fleet management in automotive — were generational companies he talked himself out of because he thought he had to be a subject-matter expert to invest. The discipline is to get out of your own way and back the most compelling founder you see in a given period doing something not antithetical to what the fund is for.

• Want the ball on a cold streak. The temperament test for inception-stage investing, and his answer to where the confidence to allocate capital came from. Running his own companies gave him what he calls irrational confidence that things work out; backing other people means they usually don't. What carries you is the gambler's disposition: "you have to be like a jump shooter on a cold streak who wants the ball at the end of the game with the game tied. Even if I missed eighteen of the last nineteen shots, give me the ball." He is built to get over losses quickly and is driven by big wins — a temperament he doubts would scale to growth-stage investing, but considers close to required at inception.

• One-sentence summary (his closing advice): Founder first, a hundred thousand percent — it is not your job to be smart about what they're doing, it's your job to be smart about them and their capacity to build something meaningful, and if the bet doesn't make you feel a little sick, it's probably too obvious to be worth making.

---

Episode 13 — Gregory Raiz

An engineer by training whose first job was Microsoft, and who walked out of it in 2003 with a laptop and conviction. What he ground away at became Raizlabs, a Boston product design and mobile engineering firm he built to roughly eighty people over fourteen years and sold in 2017 to Rightpoint, a private-equity-backed digital agency, staying on through the earn-out. He then ran Techstars Boston as managing director, and now runs Founders Edge, a pre-seed fund investing at the intersection of AI and user experience. He calls himself a "recovering entrepreneur." His distinguishing asset is The Founder Index — three-plus years of proprietary research, built on surveys of more than 3,000 entrepreneurs (about a tenth of them with seven-, eight-, or nine-figure exits) plus hundreds of academic papers — built to answer which founder attributes actually predict success, and used to de-bias his own screening.

Most Compelling First Bet Stories

1. The first angel check — closed by making them prove it on his own mortgage

He was still an operator when he wrote it. Raizlabs had sold to a private equity buyer, he was working through the earn-out, and he had a little cash and a lot of empty square footage — so the company moved into a large space and ran what he calls a pseudo-accelerator, a non-equity program that brought startups in to work alongside them. That arrangement turned out to be the real asset. "The amount of information is pretty limited" in angel investing, but "when you see how teams are actually operating and what value they're actually driving, how they're interacting on a day-to-day basis, you get true insight on how teams execute."

One of the teams in that space was in prop tech. They were older than the archetype — "a lot of folks would have said, hey, maybe you're not the quote unquote Harvard dropout, the quintessential Zuckerberg or Peter Thiel early-stage founder" — but they had extraordinary founder-market fit: they had sold close to a billion dollars of property on the brokerage side, and their thesis came from having lived the inefficiency. To get a mortgage you go through a middleman who charges a couple of thousand dollars and doesn't optimize your rate. Software could do it better, and nobody was doing it.

Greg was not a prop-tech investor and had no deep understanding of the space, so he called BS on them. "This sounds like a good idea, but how do I really know?" Their answer was to make him the experiment: we'll save you tens of thousands on your mortgage. The product barely existed — a Google form that collected some information and got plugged into a back end — but he had a broker of his own, so he ran a live bake-off. His broker produced a rate. They produced a rate. They saved him about $30,000 on his refinance. "And I literally said, okay, that's your angel check."

In his telling, that company recently exited for close to $300 million.

Why it's compelling: The bet was made with almost no conventional information — a thin MVP, no traction, a sector he didn't know, founders who failed the fashionable pattern — and he closed the gap with the cheapest possible experiment: he became the customer. The point isn't that a $30,000 saving proves a business; it's that a claim which can be tested on the spot for free should be, and that the founders' willingness to be tested was itself data. The two things he says he carried out of it are now core to his fund's thesis, and both cut against the archetype that made him hesitate: more experienced founders tend to outperform, and founder-market fit — really knowing the domain and the go-to-market inside a vertical — is a large execution advantage.

2. Betting three years on the premise that founder quality is measurable

The angel investing that followed did not go well. "I lost a bunch of money angel investing. It was not really fruitful." His diagnosis was about the room rather than the deals: he had joined seven angel groups on the theory that he should surround himself with people smarter and more experienced than him, and found that many angels "tend to not take it particularly seriously. They're not digging into the data, they're not doing the research." What he got was vocabulary — safes, convertible notes, the mechanics — not signal. "There's a lot of like, I like this idea, I like this founder."

So he went to Techstars in search of institutional knowledge, reasoning that an organization that has seen hundreds of thousands of companies must know what actually drives founder success. What he found instead was that every managing director had their own differentiated approach — all of them smart, none of them quantified. He is an LP in two or three former Techstars MDs' funds and still can't reconcile their methods into one answer.

At that point he stopped looking for the data and decided to generate it. The hypothesis was specific and falsifiable: there should exist questions that, if you ask a founder, are more predictive of higher success. His reason for believing it could be true was the one empirical fact he trusted — serial founders outperform — which implied there must be something they know, some network, or some executional capacity that de-risks the investment, even if nobody had named it. He wrote a battery of candidate questions, some of them deliberately odd: have you ever played a competitive sport? do you play a musical instrument? how long have you known your co-founder? how much domain experience do you have? where did this idea come from? did you patent it? is this your first company or your seventh? Then he surveyed more than 3,000 entrepreneurs, roughly 10% of them with seven-, eight-, or nine-figure exits, and looked for what separated them. He has been re-running it inside Founders Edge for about three years — another batch over the summer — now using AI tools to read hundreds of papers across tens of thousands of entrepreneurs.

Part of the motivation was auditing himself. On exiting Raizlabs he looked inward at his own résumé — immigrant founder, Microsoft alum, co-founder met in college, entrepreneur father — and noticed they were all entrepreneurial tropes. "How much of that is actually significant data and how much of that is coincidence?" Some funds have an explicit thesis around immigrant founders; he wanted to know whether that was quantifiable or folklore. "I think there's a lot of myths in investing, and so I'm trying to figure out what is actual signal and what is noise."

The payoff story is the cleanest demonstration of why the instrument exists. A founder applied cold through the website with a pitch deck they could not make heads or tails of — confusing, badly built — but the Founder Index score came back very high, the model insisting there was something interesting here. Buried on page nine of the deck was the reason: he was a third-time founder building his third company in the same domain, having bootstrapped the first two, and had simply never made a pitch deck before. "That's an easy thing to fix. We can show you how to build a pitch deck. It's really hard to show you how to build and scale a company."

Why it's compelling: Every other guest in this series closes the information gap deal by deal. Greg's bet was that the gap is closeable in general — that founder quality has measurable structure, and that three years of unglamorous survey work could recover it. He risked being wrong in an expensive and public way (a fund whose whole differentiation is a thesis the data might not support) and he kept finding results that cut against the market he has to raise from: school is predictive of funding but not of success, prestige is an ecosystem bias, and the archetype investors reward is not the archetype that wins. The bad-deck founder is the proof that the instrument pays for itself — it caught a signal the standard interface was actively hiding.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Become the customer — make them prove the value on you. When you can't evaluate the domain, convert the claim into an experiment you can run yourself for nothing. He had his own mortgage broker, so he put her head-to-head with a Google-form MVP; their answer saved him about $30,000, and that was the diligence. The generalization is that a testable claim should be tested immediately, and a founder's willingness to be tested is itself signal.

• Watch teams operate before you write the check. The pseudo-accelerator in his own office existed for this, and Techstars was the same mechanism at scale: day-to-day proximity tells you how a team executes, communicates and creates value in a way no meeting does. "The amount of information is pretty limited" otherwise.

• Quantify first, pattern-match second. "Humans are really bad at pattern matching, and pattern matching leads to bias." So on cold inbound the order is inverted: score the founder on data before forming an impression, and only bring judgment once the quantified read is in. The explicit purpose is to de-bias his own investing so the check goes to the highest-aptitude founder rather than the most familiar-looking one.

• Founder-market fit and reps, measured together. Experience predicts, but only in the right direction: a third-time founder building in the same vertical as their prior companies is strongly predictive of another win, while a third-time founder in a completely new domain performs "same as average." And attempts are not wins — "being a third-time entrepreneur with no wins under your belt is less important than being a third-time entrepreneur with several wins under your belt," because the more wins you have, the more likely the next one is bigger.

• Prestige predicts funding, not returns. "School is predictive of funding, not of success." A degree from a prestigious university makes a founder more likely to be funded because that is a bias in the investing ecosystem — it does not predict higher returns. The same applies to the Microsoft-or-Google line on a résumé.

• The metrics question as an execution proxy. A deliberately simple question with a graded answer: how do you track your metrics? I don't / I have a spreadsheet / I use a third-party tool / I built my own tool. The more sophisticated the answer, the more likely the team is out in the arena building rather than talking about building.

• "Show me something you built" — a bad deck is fixable, an inability to build isn't. Pitch decks are "kind of BS... a little bit of startup theater." An exceptional founder answers the show-me question instantly — here's the GitHub repo, here's the site, here's the company I sold — and a founder who struggles to answer it has told you something about execution. He would rather back someone who builds great product and needs help with a deck than the reverse, and has: a founder whose deck was incomprehensible scored at the top of the index and buried the reason on page nine.

• The ante is build execution. His most expensive repeated mistake in angel investing was the charismatic, strong CEO not paired with the technical talent to build the product — "I can probably name seven or eight companies that just went belly up because they couldn't recruit the technical talent to build." Technical build capability is a disqualifier, not a weighting.

• Ask whether the co-founder relationship has been tested. Team cohesion and founder–co-founder trust are among the most predictive variables he's found, and the failure mode is enormous: by the study he cites, twenty to thirty percent of startups die of founder conflict. So an early question in the survey was, counterintuitively, has your co-founder let you down? — and a yes reads positive. "When you've actually tested that relationship and you've been resilient past that, that really is a sign of a stronger founder–co-founder relationship." The relationship to worry about is the untested one: "we met at co-founder matching and we built a prototype over a weekend and please fund us."

• Count the co-founders — convincing smart people to work unpaid is the proof. The data points nearly linearly toward more co-founders being better, provided the team is strong; teams of three, four or five outperform. The mechanism isn't headcount, it's evidence: early co-founders aren't paid much, so persuading four or five very smart people to spend their limited life resources on your business is itself a strong signal about the business and about you.

• Back a solo founder only on proven reps. Founders Edge has done exactly one solo-founder deal, with someone who had eight prior companies, two of them quite successful. Martin's version of the same gate: the question with a solo founder is always why are you solo? — if you've never convinced anyone smart to come along, scaling will be hard — unless they've already proven they are a talent magnet, or you've caught them early enough that they simply haven't pressed on it yet. Both are skeptical of the current fashion of trying to be the first solo founder to build a billion-dollar company; as Martin puts it, you might back a proven leader who happens to be solo, but the idea that any of them could run the company alone doesn't survive contact.

• Question whether you believe it or were told it. His first piece of closing advice, and the engine behind the whole research program: "There's a bunch of things I thought were true, and then I was like — do I think this is true because other investors have told me this is true? Or do I think this is true because I've actually done a little bit of digging to verify?" Look for signal and data; treat received investing wisdom as a hypothesis with an owner you can't see.

• Curiosity is the flywheel, not the hobby. He plays with AI tools, codes his own models and tinkers with circuit boards because he finds it fun — "I'm a kid again playing with Legos, only instead of Lego bricks it's LLMs and AI models and circuit boards." The investing return on that is indirect but compounding: genuine curiosity creates deeper connections with founders, founders refer other founders, and deal flow follows.

• "Confidence" may be the wrong word — "what's next" is the engine. Asked where the confidence came from to leave Microsoft, to write angel checks, to raise a fund in a hard market, he pushes back on the premise. The successful CEOs he talks to don't perceive themselves as having cleared the bar; "they're just executing." What moves him is the next rung — of thought, process, data, execution or team — rather than a settled belief that he'll win. Martin's version is adjacent: for him leaving Microsoft wasn't confidence either, it was curiosity about a category he couldn't chase from inside.

• One-sentence summary (his closing advice): Look for signal and data rather than consensus, interrogate every belief for whether you verified it or merely inherited it, and stay genuinely curious — because curiosity is what keeps the instrument improving and what makes founders want to work with you.

---

Episode 14 — Eric Jorgenson

The synthesizer. Eric Jorgenson assembled Naval Ravikant's scattered tweets and podcast answers into The Almanack of Naval Ravikant, a book downloaded more than a million times, and then did the same for Balaji Srinivasan (The Anthology of Balaji) and Elon Musk (The Book of Elon). Before that he spent about a decade inside one startup that, in his words, "sold for a t-shirt." Today he is CEO of Scribe Media, which helps people write and publish their own books, and a GP of an early-stage rolling fund. His whole career is a bet on curation over creation — finding other people's ideas and making them usable.

Most Compelling First Bet Stories

1. The Almanack of Naval — the biggest check of his life, paid for by a flooded apartment

He was about thirty. The startup he had given ten years to had just been sold for next to nothing, and he wasn't sure what came next. The Almanack was not a leap out of a job; it was a side project that had been compounding for years — "tweeting turned into a blog and a blog turned into a book," a series of nights and weekends "that I just took progressively more seriously."

The leap was the money. The check to cover publishing costs "was the biggest check I'd ever written in my life. It was more expensive than my car." He was still renting, and the only reason he could afford it was that his apartment had flooded and the landlord gave him a couple of months of free rent while they rebuilt it. Every objection was obvious and he heard all of them: nobody ever makes money writing books; you're over-investing in a passion project and will never get the money back; why spend three years on this when you have other ideas — couldn't you have done it in three months?

His answer was to redefine what success meant. The bar was to break even and make a product he was proud of. And he wrote for "an audience of like one or two" — himself, and Naval. He pictured Naval looking over his shoulder, knowing that one day he would have to hand him the finished book and hope to hear you did a good job with this. That imagined reader "really changed my threshold of what I considered quality, and made me work and invest way harder in it."

The outcome was "unbelievably asymmetric" — a million-plus downloads, and a single rare accomplishment that turbocharged everything after it: "there's just a world of difference in who I met and how quickly I met them" once the book was out. It also made the next two books possible. Going from zero happy customers to one changes your internal world, and "navigating that maze all the way through one time kind of gives you the sense that you're gonna make it through and it makes you more confident entering the maze again."

Why it's compelling: It is the purest example in the series of a bet sized by the person and not by the expected value. He had no forecast of success; he had a quality bar and a reader in his head. That is the power law from the inside: he couldn't see the tail outcome, so he optimized the only thing he controlled — being excellent — and let the distribution do the rest. It is also a case study in how cheap an asymmetric bet can be made: years of nights and weekends kept the job intact, and the one real cash outlay was funded by a windfall.

2. Longshot — backing the next leg of the launch-cost curve

His investing thesis is built on curves. Founders worth backing, in his view, have often watched a threshold for a decade and can tell you it crosses next year — the way Steve Jobs watched hard-drive storage and battery curves to know when an iPod could replace the Walkman. The same dynamic is everywhere: CRISPR's tooling finally reaching the throughput and clarity to be useful; research breakthroughs a professor won't commercialize; space supply chains where a component that cost $100,000 to build in-house now costs $1,000 off the shelf, so a problem that needed $100 million now fits in a seed round.

Longshot is where he took that one step further. Once you see that SpaceX will drop the cost to orbit by two orders of magnitude, you see where the next opportunities are — but Longshot's bet is on the curve itself. SpaceX, in his telling, will bottom out somewhere around $500 per kilogram. Longshot has an architecture it believes can theoretically get to $50 or even $10 per kilogram: a ground-based launch system that throws only the cargo, instead of putting the cargo in a rocket that has to lift all of its own fuel. "What is the next leg of that curve going to look like to drop that cost even more?"

Why it's compelling: Most "why now" investors ride someone else's curve. This is a bet on a company trying to be the next leg of one — a much higher-uncertainty wager on physics and engineering, but also one with, in his framework, no obvious ceiling. It pairs neatly with the Almanack: both are bets where the downside was a capped, affordable loss and the upside, if it came, would be off the chart.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Do the thing you'd do for its own sake. People repeat the line that you do best at what you're drawn to and would do for fun, but "it's really hard to actually live that way." The Almanack is his proof: he did it expecting no reward, because "the process was going to be its own reward," and the reward turned out to be enormous. Define success as something you control — break even, and be proud of it.

• Write for an audience of one. When you can't predict how the market will respond, pick the one judge whose opinion you'd care about and imagine them looking over your shoulder. It sets a far higher quality bar than any market forecast would have.

• Live on the power law — every other rule is noise. Book writing and early-stage investing are both bad businesses on average and incredible ones at the top 10%, 1% and 0.1%. Paraphrasing Peter Thiel: getting into that tiny fraction matters so much that "there are no other rules... at the end of the day, are you in that fat edge of the power law or not? And everything else is noise."

• Put the wood behind one arrow. The extra value of going from the top 1% to the top 0.1% is much higher than the value of starting another project and "firing another lead bullet." MrBeast's version: it's much easier to get 10 million views on one video than a thousand views on each of ten thousand videos. So make one really, really good thing.

• Name the ceiling. His first question in diligence is what could cap the outcome, and whether this founder or market has it. Founder ceiling: will they take a $50–100 million offer instead of pushing on to a $1–10 billion company? Market ceiling: does the market size and competition limit this to a $100–500 million business, where dilution and a switch from growth multiples to earnings multiples cap the value? He looks for people, markets and technologies with unlimited upside — or at least orders of magnitude more than his alternatives — because without a ceiling, there is always a next buyer and the growth keeps going.

• The sensible-trajectory test. Is this company the natural next chapter of the founder's life — related to their PhD, their first job, their first company that didn't work — or are they chasing markets, drop shipping to crypto to AI? The founder who has loved spaceships since childhood, grew up in their dad's metal shop and has spent thirty years on asteroid mining is likely to succeed early and, more importantly, to keep going through the messy middle.

• Back the threshold-watcher. The most compelling pitch is "I've been watching this threshold for ten years, we haven't been there, but we're going to cross it next year — which is why I'm starting this company now. And here's what that unlocks." The unlock can be AI, a research breakthrough nobody is commercializing, or simply a supply chain that has made a $100,000 part cost $1,000.

• Find the company moving the curve — then ask about the next leg. Seeing the cost curve drop shows you where the next opportunities are. The bigger bet is the company that bends it again (Longshot, aiming to take cost to orbit from roughly $500 to $10–50 per kilogram).

• Farm asymmetry with cheap, repeated bets. The biggest lever on his risk appetite is hammering "into my bones that you can make these relatively cheap asymmetric bets over and over and over again." In domains with asymmetric payoffs and high ceilings, if you're making the right ones, "it's all gonna work out," and no single failure is catastrophic. So have fun with each one, be optimistic, move fast, and move on to the next.

• Don't mix decision modes. "Downside capped, upside not" asymmetry farming and Warren Buffett–style don't-lose-money investing are "very different headspaces." Mixing them leads to weird decisions and rationalizations. He batches them, writes through each decision, and reminds himself which heuristics belong to which kind of bet. (Martin's version is his barbell: 10–20% in early-stage venture, 80% in low-leverage real estate where he makes no decisions, and the cash flow from the second is what lets him take risks in the first.)

• Price the full human cost of an investment. How much maintenance will it need — yearly buy/sell decisions, managing real estate, a board seat? Is the payoff meaningful compared with everything else you have going on? He wrote about this in his post "Opportunity Cost and Attention Thresholds," which he says Nick Maggiulli developed into The Wealth Ladder.

• Grow your psychology along with your portfolio. You should become desensitized to bets your past self would have found life-or-death scary. If you don't coach yourself through that, you'll abandon a strategy that was working just because the numbers got bigger. Scale up as results arrive — from 1% to 5% to 10% to 15% of a portfolio. (Martin's path fits: $10–25K angel checks, $250K checks in Fund I, $500K in Fund II, each step on the track record of the last.)

• Confidence comes from closed loops — and from a secure base. Going from zero happy customers to one is "a world of difference"; getting through the maze once makes you confident entering it again, and closing that loop many times builds the conviction that you can figure things out. Underneath it is something psychological: close, loving relationships mean "my long-term happiness is not dependent on whether or not any individual of these bets are successful."

• One rare early win is a turbocharger. Many careers pivot on one thing you become known for early — a PhD, an early exit, an athletic achievement, time at a rocket-ship company. For him it was the book, and it changed who would talk to him and how fast a real conversation started. It's worth making bets that could become that thing.

• One-sentence summary (his closing advice): Make cheap, asymmetric bets in domains without a ceiling, do each one so well that you'd be proud to hand it to the person who matters most, and let the power law — not your forecast — decide which one changes your life.

---

Episode 15 — Mahesh Narayanan

The first check into the hard stuff. Mahesh Narayanan is Managing Partner and founder of Neuvation Ventures, a pre-seed fund that underwrites science risk — medtech, diagnostics, digital health and deep tech, with its roots in brain health and neurodiversity — after roughly fifteen years running life-sciences startups. Before Neuvation he invested with the Autism Impact Fund and helped build an accelerator for neurodiversity technologies with Dan Feshbach, which is where he met the founder of Impruvon Health. The episode is personal for the host: in May 2022 Mahesh emailed Martin the Impruvon deal, Martin wrote the check on the strength of Mahesh's diligence, and the two of them then watched the company struggle for eighteen months before it became one of the fastest-growing businesses either of them owns.

Most Compelling First Bet Stories

1. Impruvon Health — a year of watching, then a check the day the first pilot paid

Mahesh met Justin, Impruvon's founder, in April 2021 — almost a year before he invested. There was no company yet in any real sense: no box, no customers, "everything was pretty much in his head." What there was, was a reason. Justin had lost his brother to a medication overdose and wanted to make sure it didn't happen to anyone else. For Mahesh that is the first signal he looks for in any founder — "how close is this founder to the actual problem itself? And for him it was very much personal" — and it matched his own reason for being in the category, a deep family connection to neurodiversity.

The original idea was a locked medication dispenser plus software for intellectually and developmentally disabled (IDD) individuals, a population that struggles with medication compliance far more than typical patients. Mahesh liked the problem and didn't like the market: there are only a handful of IDD clinics in the US, enough for "a great idea, we can build it for a small population and that's it," not a fundable company. So instead of passing or investing, he pulled Justin into the first cohort of the accelerator he was helping run — "you have an idea, but we need to accelerate this a little bit more and get to a point where you can actually be fundable" — and pushed him outside what he knew, into pilot studies with health systems, long-term care and geriatric facilities.

Two things turned "a good idea" into "an investable idea." The first was a number nobody had seen: in the pilots, medication compliance went from a community baseline of roughly 30–40% to 80–90%, "which is unheard of," in populations that do not want to take medication — and with no overdoses. "This is gonna really change how we manage care within these type of facilities." The second was money. In 2022 the first unpaid pilot converted to a paid one, and that conversion was the trigger: "once he's got the first customer, paying customer, from a unpaid beta to beta, that's when I reached out and said, okay, now I'm ready to make that first investment."

The check was small for structural reasons on both sides. The round was small, the founder had raised most of it already and didn't want to take more, and Mahesh's own fund had only just held its first close — so the final allocation he brought to the table, split with Martin's $50K, was both the biggest check he could write and all the company needed. Martin is candid that his own process was faster — sixteen questions, answered quickly — because he was "riding a little bit on your diligence." The year of proximity was Mahesh's.

Why it's compelling: Mahesh weighted the founder at about 70% of the decision, and that is exactly what paid off, because the business Impruvon became was not on anyone's roadmap — including his. He didn't underwrite the box or even the market he invested in; he underwrote a founder who would make it happen "whether I'm there or not," confirmed with a year of watching him work, and waited for two hard proofs (behavior change and a paying customer) before sizing a check to what the round and his brand-new fund could carry. It is the cleanest case in the series of founder-heavy underwriting buying the right to be wrong about the product.

2. The seed nobody wanted — and the state that turned a device company into a software company

The hard part came after the check. There was "a good 18 months" in which the company struggled to raise more money and find product-market fit. The founders wanted to raise $3 million; Mahesh and Martin talked them down, because there wasn't enough traction to raise that much at a pre-money of roughly $10–11 million "where we're not all getting [diluted] down to nothing." The round was split into two $1.5 million seeds. Even the smaller first one took six or seven months; Martin sent it to Abby Levy at Primetime Ventures, who liked the senior-care market but found it too early.

The bottleneck was regulatory. Each state had to approve the device for use in its facilities, and Justin spent eight to twelve months doing it himself — "knocking on the CMS doors," about six months per state, Massachusetts first and slowest. Once he had raised a little capital he did "the smart thing" and hired someone who knew the process. With the documents in place and a couple of approvals behind them, states started approving each other's lead: the company went from two states in its first twelve months to almost thirty in the next twelve.

The pivot came out of the payment problem. Facilities didn't have the budget for a hardware-software system, so the company went to insurance — and CMS turned out to be more interested in the software side of compliance than the box. A state impressed with the compliance numbers asked whether Impruvon could roll the software out across every system it covered. "Are you willing to pay for it?" "Yes, we will pay for this." The result was a multi-year, prepaid, multimillion-dollar contract to build something adjacent to the original product — a real gamble, Mahesh says, after one state. Before building it, Justin went to other states and asked what they would need, because what worked for Massachusetts might not work for California or Arizona, and built it once for all of them rather than over and over for each. Today Impruvon sells statewide medication-compliance software to state governments, with or without the hardware, and — per Martin on air — has grown from about half a million dollars of revenue to well over ten to twelve million in roughly a year and a half, helped by a tailwind that didn't exist at the first check: states looking for software to reduce Medicaid fraud risk.

Why it's compelling: Every stage of this story is a decision a less coachable founder could have refused — cutting the raise in half, giving up doing approvals himself, saying yes to a contract for a product he hadn't built, and resisting the urge to custom-build for each state. Martin's line on the contract is the point: "there are actually CEOs who would have said no to that deal." The investors' conviction didn't come from foreseeing any of it. It came from having watched, before the check, how this founder took feedback.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Proximity to the problem is the first signal. "How close is this founder to the actual problem itself?" A founder who has lived the problem — here, losing a sibling to an overdose — brings a motivation that market conditions can't remove. Mahesh's own fund exists for the same kind of reason.

• Back the founder who will build it with or without you. The real question isn't whether you can help; it's "how much do I believe that this founder is not going to give up on this idea, even if I do." He digs into what gets the founder up every morning, and it "cannot be money motivated, it cannot be fame motivated" — those "go away overnight for most founders." When the answer is right, the decision simplifies to "do I jump on for the ride or do I sit in the sidelines and watch him?"

• The expansion test: can it plug into another market? It doesn't need to be a platform technology, but he has to see the product working in a market beyond the one it was built for. "If I don't see it being able to just... plug it into another market... then oftentimes that means the market's just too niche." Impruvon passed it twice: IDD clinics to long-term care, then long-term care to state Medicaid programs.

• Coachability is a round trip — and the clock is a week. Founders "don't have to agree with everything I say," but they have to listen, think about it and "come back with something better." If he points out a problem and doesn't hear back within about a week — either "this is not a problem" or a couple of solutions — "that's a deal breaker," because it means the feedback was dismissed rather than discussed. Impruvon's founder set the bar: told a role wasn't his strength, "within a day or two he'd come back with five candidates." (Martin's version swaps obedience for curiosity: he wants "a curiosity for my comments and a respect for being challenged," and an uncurious founder who dismisses hard questions is "a big flag" for a ten-year relationship — even though some of his best founders have ignored his advice entirely and won anyway.)

• You can read all three in the first meeting. Is this a market worth investing in, is the motivation durable, are they willing to listen? "It's really not hard for most investors to figure this out."

• Name the moment a good idea becomes an investable one. For Impruvon it was two proofs together: an outcome nobody had seen (compliance from 30–40% to 80–90% in a population that resists medication) and the first customer converting from an unpaid pilot to a paid one. Before that, it was "just a good idea."

• Re-weight the scorecard by where the risk lives. On a software company the moat is "how big of the market you're gonna capture and keep" — execution — so the founder and team carried about 70% of the decision. On harder life sciences he splits it roughly 30% team, 30% IP, 40% data: still people-first, but now the questions are how the asset is protected and whether the data holds up.

• Unreplicated science isn't evidence yet. "We've done it" isn't enough: "No one else has done it. So I can't believe you yet. Someone else needs to validate this." He looks for publication and outside replication, then the quality of the data — one animal model or several, GMP or non-GMP manufacturing.

• Count the regulatory touchpoints, not just the label. Will it be one 510(k) submission and done, or a path that goes back to the FDA every year? "The more you interact with them, the more expensive your trials are gonna be" — meet three times and the cost can be thirty times. The more straightforward the path, the less of the round disappears into research overhead.

• Rent the compliance you don't need to own. Impruvon handled HIPAA by storing its data with a provider that was already compliant rather than certifying itself — faster and cheaper, and one less regulatory process for a small team.

• Size the round to the traction, not the ambition. Raising $3 million at a ~$10–11 million pre-money on thin traction would have diluted everyone heavily; splitting it into two $1.5 million seeds let the second one price on the progress the first one bought.

• The first approval is the expensive one — then hire the specialist. The founder doing state-by-state approvals himself took eight to twelve months; a hire who knew the process, plus the precedent of states that had already approved, took the company from two states to almost thirty in a year.

• Follow the payer — it may be buying something bigger than your product. Facilities couldn't afford the system, so the company went to insurance; CMS cared about software compliance more than the device, and a state's request to roll it out system-wide became the business. When a customer offers to prepay for the adjacent product, take the conversation seriously.

• Build once, sell to many. Before building for the first state, ask the next states what they'd need, build broad, and customize at the edges — building per customer is "an infinite loop" that kills margins. (Martin's version: if every customer wants a different set of custom agents, "you just created a consulting company. That's not scalable.")

• A good CEO knows when to jump in and when to get out of the way. "When something's going wrong, he knows to jump in and fix it, and something's going right, he knows to step out of the way and just let it run with whoever's running it." Paired with listening to many soundboards and executing fast, it's what he calls the key components of a good CEO.

• One-sentence summary (his three-part screen): Back a founder who is close enough to the problem that they'll never quit it, coachable enough to come back with something better within a week, and building something that can plug into a bigger market than the one it started in — then wait for one hard proof before you write the check.

---

Episode 16 — Martin Jacobson

The first Martin-and-Martin episode. Martin Jacobson is a Swedish poker professional who won the 2014 World Series of Poker Main Event and its $10 million first prize, coming from eighth of nine in chips at the final table. Before poker he trained as a chef, skipping college for culinary school and taking a job at a Michelin-starred restaurant in Barcelona at twenty. He has been an angel investor for over fifteen years, and is now a first-time founder as CEO of Cuppacoin, a performance platform that lets coffee operators run incentive campaigns that tie frontline staff to store targets.

Most Compelling First Bet Stories

1. The November Nine — eighth of nine in chips, three months to prepare, thirty simulated final tables

In 2014 the Main Event paused for months between setting the final nine and playing the final table. For Jacobson that gap was "your biggest dream, but it's not something that you prepare for, even though you prepare for it your whole career": a once-in-a-lifetime spot in a field of seven or eight thousand, a slight chip disadvantage, and a fixed window to get ready. Poker is a game of information, so he spent the window manufacturing it, by hand, in an era before AI or deep video archives.

He started by asking for help. He formed a Skype group with about thirty of his strongest friends, all successful high-stakes players, and asked how they would spend the next three months — "thirty opinions is better than one." Then he built the information. He read every hand report PokerNews had published on each opponent, researched their backgrounds — how long they had been playing, how they were doing financially — and wrote a player profile for each one. He handed the profiles out to the group, assigned each friend an opponent to play, and they ran the final table about thirty times. The scenarios varied on purpose: Jacobson as chip leader, as short stack, as middle stack with one player holding everything and trying to bulldoze the table.

He is clear that none of it predicted the outcome — "you don't know what's gonna happen... anything can happen." What it bought was readiness. "When I was there I felt so confident and... whatever happened, I was sort of prepared for mentally because I knew I had done this training." He went in tweeting the Seneca line he lived by, luck is what happens when preparation meets opportunity, and climbed from eighth to first.

Why it's compelling: It is the cleanest example in the series of turning a fixed waiting period into an information edge. The cards would stay random, so Jacobson worked on everything around them: who the opponents were, how they decided, and how he would react under every stack shape. As Martin points out, the method — opponent profiles plus simulated reps — is exactly what anyone could now do in an afternoon with AI. Jacobson did it manually, with thirty people, years earlier.

2. Leaving a capped game — from world champion to first-time CEO of Cuppacoin

After winning the Main Event, Jacobson had a problem success creates: there was "not much more to accomplish." Poker is a zero-sum, capped game, and reaching the top was eroding his motivation to put in the work. It was also comfortable — the recognition, the community, the routines — which makes it "an easy one to leave, but also a hard one to leave." He says he is not especially money-driven ("I'm driven by solving meaningful problems"), and he weighed joining a VC firm full time before deciding the founder's life suited him better: "I just love the zero to one."

The bet arrived through an acquaintance running a franchised coffee shop in London who couldn't get his staff aligned with store targets and kept losing them. He ran a manual version of the idea in his own store, and it worked: staff stayed longer, felt recognized, and for the first time the incentives were visible and fair. Jacobson recognized the problem from two directions. From poker: "incentives is everything... everyone is driven by incentives at the end of the day. They might not be financial, but there needs to be some incentive. Otherwise, why would you care?" From his hospitality background and his angel portfolio: staff churn is a problem across the whole industry. He also saw that the payoff wasn't only the slow one, churn, but immediate store metrics the owner already tracks — sales and Google reviews.

He didn't go in knowing his role. "In the beginning I wasn't sure what my role would be," and then he "kinda fell into the CEO role," driving the business with his co-founder and putting in "a little bit" of his own capital. After years as an angel with "very little control," he likes having his hands on the day-to-day, and treats it as being back where he was in poker fifteen years ago — learning by trial and error, problem-solving under pressure.

Why it's compelling: Jacobson walked away from a game he had mastered for one he is new at, and the reason was the shape of the game, not the money: a capped, zero-sum contest he had already won versus an open-ended one with meaningful problems to solve. The founder bet itself rests on a short chain of evidence — a manual pilot that worked in one store, a problem he knew firsthand was industry-wide, and a mechanism (incentives) poker had taught him to trust.

Decision Frameworks for Low-Information, High-Uncertainty Decisions

• Preparation is the information edge — build a dossier on every counterparty. When the cards are random, the edge is everything else you can know. He read every reported hand of every opponent and profiled their experience and finances. "Luck is what happens when preparation meets opportunity."

• Simulate the scenarios before the real table. About thirty run-throughs of the final table, with friends playing assigned opponents across every stack shape — chip leader, short stack, one player bulldozing. The point isn't prediction ("anything can happen"); it's that nothing on the day is new, so you execute with confidence instead of reacting.

• Thirty opinions beat one. Before deciding how to prepare, he asked thirty of the strongest players he knew how they would spend the window. Crowdsource the plan from people who are better than average, then own the execution yourself.

• Commit to your process — outside opinion is noise. Few people had ever been in that seat, so few were qualified to say what was right, and the right approach is individual. Know your temperament (he is an admitted over-preparer, "for better or worse") and build the process that suits it.

• Judge the decision by its EV, not its result. "Every decision has an EV." Decisions are fully in your control; results include outcomes that aren't. As a founder, focus on what you know and what you can control, and stick to the process.

• Be honest in both directions. It's easy to forgive a good decision that lost. The harder audit is the one that went your way: sometimes the result works out in your favor and it was still a bad decision — and admitting that is "much, much harder."

• Compress the feedback loop. What he took from angel investing into the CEO job: shorten the cycle between a decision and the information that tells you whether it had any value. Poker answers in minutes and venture in a decade; a CEO can engineer something much closer to poker.

• Incentives are everything. "Why would the frontline staff care about your sales or your reviews if there's no incentive?" Evaluate a business idea by asking whose incentives it fixes. Not all incentives are financial, but there has to be one.

• Start from a manually proven pilot, then ask whether it generalizes. The idea came to him already working by hand in one London store; his job was to judge whether that store's problem was the whole industry's. His hospitality background and portfolio said it was.

• Leave the capped game. Once you have reached the top of a zero-sum game with a ceiling, the motivation to do the work fades. Move to a game where the upside isn't capped and the problem is meaningful to you.

• Pick the seat by the work you love. He considered becoming a full-time VC, but the zero-to-one work — taking an idea into the "nitty-gritty" — suits him better than advising from the outside.

• With risk comes opportunity — not taking risks is the riskiest thing. His life is a chain of high-risk decisions: culinary school while peers studied academics, a Michelin-starred kitchen in Barcelona at twenty, poker, angel investing, founding. "There's no opportunity if you never take any risk"; avoid it and "you leave so much potential at the table." He cites Seth Godin's version: not taking risks is one of the riskiest things you can do.

• Keep a fallback under the early risks. Without college, he could always go back to cooking — "it's easy to get a job in a restaurant" — so leaving for poker "wasn't that risky for me at the time."

• Watch your risk tolerance drift. In his experience, and his peers', risk tolerance falls with age, "biologically" — so be mindful of it and correct for it, rather than let it quietly make your decisions for you.

• Nothing is fifty-fifty just because there are two outcomes. Most people call any two-outcome event a coin flip ("it's a fight, anyone can win, it's fifty-fifty"); the edge is handicapping it properly — "it's eighty-twenty." Understanding probabilities is one of the advantages he carries everywhere.

• One-sentence summary (his closing advice): Get comfortable taking risk — prepare until nothing on the day is new, judge yourself on the quality of the decision rather than the result, and remember that with risk comes opportunity, because if you never take any you leave too much potential on the table.

---