The logic of public markets—margin of safety, moats, intrinsic value—partly breaks down in the early-stage world. Here returns are decided by a tiny minority of winners, and most bets going to zero is the norm. This issue returns to the primary work of Thiel, Paul Graham, and angel-investing research to understand the decision rules of a power-law world—what they mean for a long-term investor, and where they fail.
The Framework
Early-stage returns do not follow a normal distribution—they follow a power law: the single best investment in a fund often returns more than all the rest combined.
Source & Quote
Peter Thiel, Zero to One (2014), Ch. 7 "Follow the Money."
"The biggest secret in venture capital is that the best investment in a successful fund equals or outperforms the entire rest of the fund combined."
— Peter Thiel, Zero to One (2014)
Interpretation
Public markets are roughly normal—most companies cluster around the mean. The early-stage world is extremely asymmetric: a Correlation Ventures study of ~21,000 venture financings (2004–2013) found that about 65% failed to return their capital, only ~4% returned more than 10x, and what drives an entire fund's return is the tiny sliver of 50x-plus outliers. The counterintuitive implication: don't ask "could this lose money," ask "if it works, can it be big enough to move the whole fund." A "safe" bet capped at 2–3x is actually a failure here—it consumes capital that could have gone to a potential 100x.
Case Study
In 2004 Thiel put $500,000 into Facebook for ~10% of the company; around 2012 he cashed out for over $1 billion—a single position that returned more than the rest of Founders Fund's early bets combined, a living specimen of the power law. Another: Sequoia invested ~$60M total in WhatsApp and made ~$3B when Facebook acquired it for $19B in 2014—nearly 50x on one deal.
Limits & Checklist
The power law is the VC's friend but the individual's trap. It only holds if you can make enough bets and can access top deals. If you back only three to five companies, the power law means you will most likely miss the one winner entirely—negative expected value. It also does not apply to a public-equity portfolio: concentrating with a "bet on one 100x" mindset in secondary markets disguises risk as opportunity. Beware survivorship bias: you see the Facebooks, not the thousands buried alongside them.
- Do I have enough bets (20+) for the power law to work in my favor?
- If this works, is the ceiling 3x or 100x? Is a capped "safe deal" a negative asset here?
- Is the sample of returns I see distorted by survivorship bias?
Essence & Reflection
In a power-law world, the biggest risk isn't losing one bet—it's missing the single winner that would have changed everything.
Is your portfolio built to "avoid losses" or to "capture extreme winners"? Can those two logics coexist in the same account?
The Framework
The first-principles question behind a startup (and an early bet) isn't technology or funding—it's whether you're building something people actually want, and haven't run out of money.
Source & Quote
Y Combinator's motto, and Paul Graham, How to Get Startup Ideas (2012).
"The way to get startup ideas is not to try to think of startup ideas. It's to look for problems, preferably problems you have yourself."
— Paul Graham, How to Get Startup Ideas (2012)
Interpretation
PG attributes countless failures to "solving a problem no one has." Two operational tools: ① Do Things That Don't Scale (2013)—win your first hundred die-hard users through manual, unscalable effort rather than talking about growth; ② Default Alive or Default Dead? (2015)—at your current growth and burn rate, can your existing money carry you to profitability? If default dead, raising only delays the death. For an investor, this reduces a hollow "story" to two testable facts: retention and runway.
Case Study
Early Airbnb was dying; PG's advice—"go to your users"—sent the three founders to New York to reshoot hosts' listing photos door to door, and conversion recovered: the textbook "do things that don't scale." YC, founded by Paul Graham in 2005, has incubated Airbnb, Stripe, Dropbox, Coinbase, and more; its portfolio's cumulative valuation exceeds $600B, with an acceptance rate around 5%. Its screen has always been plain: the growth curve, and whether anyone actually wants it.
Limits & Checklist
"People want it" is necessary, not sufficient. Thiel's reminder: failure comes more from "poor sales and distribution" than "a bad product"—many products people want die on unit economics or customer-acquisition cost. The framework also skews toward consumer and software; it has weaker predictive power for deep tech and long-R&D-cycle ventures (biotech, hardware, AI foundation models), where "want" is only validated a decade later.
- Is there a small group that loves it madly (versus many who find it "fine")?
- Does the retention curve flatten (users stay), or keep declining?
- Is it default alive or default dead—can current money reach profitability?
- Is growth driven by real demand, or bought with subsidy burn?
Essence & Reflection
Ask "does anyone really want this" first; talk technology, valuation, and funding second—reverse the order and everything after is illusion.
The last time a grand narrative excited you, did you first go find one user who actually uses it and is willing to pay?
The Framework
Real value creation is going "from 0 to 1"—building something unique and near-monopolistic—not going "from 1 to n" by copying existing models into unprofitable competition.
Source & Quote
Peter Thiel, Zero to One (2014).
"All happy companies are different: each one earns a monopoly by solving a unique problem. All failed companies are the same: they failed to escape competition."
— Peter Thiel, Zero to One (2014)
Interpretation
Thiel's counterintuitive claim is that "competition is for losers." Under perfect competition profits are ground to zero; only firms that escape competition and build a monopoly can earn and keep excess profits, and thus invest for the long run. Durable sources of monopoly: ① proprietary technology (10x better than the runner-up); ② network effects; ③ economies of scale; ④ brand. His signature question—"What important truth do very few people agree with you on?"—is really a search for a "secret" others haven't seen, and secrets are the raw material of 0→1. This is isomorphic to Buffett's moat, but Thiel stresses monopoly at the origin rather than its defense.
Case Study
Google's share of search has long hovered near 90%; the excess profits of that monopoly let it fund loss-making "moonshots," while it publicly downplays its dominance to dodge antitrust scrutiny. PayPal, which Thiel co-founded, first dominated the small market of eBay payments and then expanded outward—the "monopolize a small market first" move that recurs throughout the Zero to One playbook.
Limits & Checklist
Monopoly is not eternal. Thiel praises the "last mover," yet Nokia, Kodak, and Yahoo were once unassailable monopolists, eventually pierced by technology and platform shifts. A more dangerous misuse for secondary-market investors: betting on "the next monopoly" at venture valuations in public markets—the dot-com bust of 2000 was the collective reckoning of exactly that narrative. Monopoly also faces the real ceiling of antitrust and regulation.
- Is this business creating something new (0→1), or copying in a red ocean (1→n)?
- Is its edge 10x the runner-up's, or just marginally better?
- Is the monopoly's source a real barrier (proprietary tech / network effects), or a fleeting head start?
- Am I using a "next monopoly" story to excuse an overly high price?
Essence & Reflection
Profit comes from escaping competition, not winning it—the businesses worth owning are the ones rivals cannot copy.
For your highest-conviction investment, do its excess profits come from a real barrier, or from a story competition hasn't yet falsified?
The Framework
For an individual, early-stage investing is a "portfolio game" with strict entry requirements and demands of discipline—not a shortcut riding on one or two lucky bets. Underestimating its thresholds is the most common way to lose money.
Source & Quote
Robert Wiltbank, Returns to Angel Investors in Groups (Kauffman Foundation, 2007).
"The distribution of returns is highly skewed... 52% of all exited investments returned less than the capital invested."
— Wiltbank, Returns to Angel Investors in Groups (2007)
Interpretation
In the same study, angel investing averaged ~2.6x over ~3.5 years (IRR ≈ 27%)—enticing—but the average is dragged up by a handful of big winners, and the median is far lower. This is the individual-scale consequence of the power law, and it produces three real thresholds: ① eligibility—most jurisdictions require an "accredited investor" (e.g., U.S. net worth over $1M or income over $200K); ② illiquidity—capital locked up for 7–10 years with no exit in between; ③ diversification—because of the power law, you need 20–30+ bets for a reasonable chance of catching a winner. More insidious is adverse selection: the best deals are oversubscribed and never reach individuals without access; what lands in front of you is often what better investors already passed on.
Case Study
The "spray and pray" model promoted by AngelList and Naval Ravikant is a direct response to the power law and adverse selection—using syndicates (led by a lead investor) and small, numerous bets to turn the individual angel from "betting on one or two" into something like an indexed early-stage portfolio. By contrast, many individuals stake their savings on a single startup run by friends or family—landing squarely in that 52% that returns nothing.
Limits & Checklist
This is the mirror image of Buffett's "punch card" concentration philosophy—concentrating in predictable public businesses is an edge; concentrating in unpredictable early-stage names is suicide. The limit: even a tech-background "super-individual" has finite time, access, and due-diligence capacity, so the realistic place for direct angel investing is a "small, numerous, can-lose-it-all" satellite allocation—not the main channel.
- Do I meet accredited-investor standards, and can this money go to zero without affecting my life?
- Can I bear 7–10 years of illiquidity, plus possible follow-on capital calls?
- Can I assemble 20+ bets so the power law works for me, rather than betting on three to five?
- Are the deals reaching me top-tier, or what others left behind (adverse selection)?
Essence & Reflection
Early-stage investing isn't "hit one and you're free"—it's a long-term portfolio undertaking demanding eligibility, discipline, diversification, and access.
If you could only make three early-stage bets, and two would likely go to zero, would you still use this money for angel investing? Or does it belong where you genuinely have judgment?
Going Deeper
Can power-law logic transfer to public markets?
Partly. Over the past decade, S&P 500 returns have been highly concentrated in a few giants (FAANG / the Mag 7), which is isomorphic to the power law. But secondary markets differ in two fundamental ways: ① you can buy the whole index cheaply, automatically holding those winners; ② public companies' downside isn't zero but volatility, with liquidity. So for an individual, the safest way to capture the public-market power law is precisely "not to pick"—indexing. Trying to replicate VC's power law by "concentrating on one 100x" in secondary markets usually just bears the volatility without VC's diversification or pricing power.
In the AI era, is going "from 0 to 1" easier or harder?
Both sides are amplified. AI crushes the cost and time of "building a usable product"—the 0→1 barrier drops. But precisely because anyone can build fast, competition intensifies and the "1→n" red ocean arrives sooner, making durable monopolies scarcer. Thiel's framework implies: in the AI era, what separates winners is no longer "can you build it" but proprietary data, distribution channels, and network effects—things AI cannot easily copy. The model itself looks more and more like a commodity; the moat moves up to data and ecosystem.
What does early-stage discipline teach a long-term value investor in reverse?
The biggest lesson is "know which game you're playing." VC captures extreme winners by diversifying because targets are unpredictable; value investing concentrates in predictable quality companies because an edge is established before the bet. Mixing the two is dangerous: using VC's "bet on winners" mindset to concentrate in public stocks is undiversified speculation; using value investing's "concentrate" mindset on unpredictable angels is undisciplined gambling. First establish a target's predictability, then choose the matching bet structure.
What does "do things that don't scale" mean for an AI super-individual?
It's a reminder: in an era of heavily automated tools, what's truly scarce and hard for AI to replace is exactly the "unscalable" human moves—personally understanding the real pain of your first few users, hand-crafting the first version of the experience, building trust. AI can amplify already-validated demand, but it can't discover demand for you. For a super-individual, the leverage comes from AI, but the origin "secret" and the real demand still require unscalable, hands-on effort to unearth.