Project to age 80: would you regret not trying? Most likely yes — especially since you have co-founders, domain expertise, and enough savings for a runway. People rarely regret the attempts; they regret the safe choices that haunted them. But this calculus shifts if a failure would create genuine hardship for your family, not just discomfort.
This is more reversible than it feels. At 32 with tech experience, you can re-enter the job market within 6-12 months if the startup doesn't work. The real irreversible element isn't the job — it's the cash. $200K spent is gone. Ask: if this fails in 18 months, can you recover financially within 2 years? If yes, this is a two-way door.
Staying costs you the startup's potential upside plus the learning and network you'd build as a founder. Leaving costs you ~$270K in salary (18 months) + $200K in savings at risk = $470K total exposure. But the startup's expected value isn't just financial — it includes optionality, skill development, and career positioning that a corporate role can't provide. The asymmetry favors the startup if your downside is truly bounded.
$200K / estimated monthly burn (~$12K personal + $8K startup costs) = 10 months of runway. That's thin. Most fintech startups need 18-24 months to find product-market fit. Your margin of safety is insufficient unless you can (a) raise a small pre-seed round, (b) keep personal burn under $8K/month, or (c) generate early revenue within 6 months. Without at least one of these, you're racing the clock.
It's 2028 and the startup failed. Most likely reasons: (1) Ran out of cash at month 11 before achieving product-market fit. (2) Regulatory complexity in fintech took 2x longer than expected. (3) Co-founder conflict over equity/direction emerged at month 6. (4) Partner's anxiety about financial instability created pressure to quit early. Each of these is addressable now — but only if you plan for them.
The models converge on a conditional yes. The decision is sound if your financial runway is adequate. The gap in your plan isn't courage or conviction — it's cash buffer.
→ Before quitting: extend your runway to 18+ months (raise $150-300K from angels/friends-and-family, or save for 6 more months). Resolve co-founder equity/roles in writing. Have an explicit conversation with your partner about the worst-case scenario and what "pull the plug" looks like.
Why are three other VCs in the round? Genuine conviction, or social proof? Check if they did independent diligence or followed each other in. The CEO's charisma is a feature for fundraising but a risk for diligence — charismatic founders are better at making you feel good about a deal than the deal deserves. Separate the person from the numbers. Ask: would you invest if the CEO were boring?
At $150M pre-money on $12M revenue, you're paying 12.5x revenue for a lending business with 35% gross margins. Comparable companies at Series B trade at 6-8x with 55%+ margins. Either the market is wrong (possible), you're overpaying (likely), or the margins will improve dramatically (needs evidence). Where's the margin of safety? If margins stay at 35%, you need revenue to hit $50M+ for the valuation to make sense at exit. That's 4x from here — in lending, in LatAm, with regulatory risk.
Three reputable VCs being in the round feels reassuring. That's the bias talking. In 2021, the most oversubscribed rounds produced some of the worst losses. Ask: what do the VCs who passed say? The bears are more informative than the bulls on a hot deal. If you can't find anyone who passed, you haven't done enough diligence.
Instead of "why should we invest?" ask "what would make this a zero?" Lending businesses blow up when: (1) credit quality deteriorates in a downturn (LatAm is macro-volatile), (2) regulatory changes cap interest rates or require expensive licenses, (3) the company scales revenue by loosening underwriting standards (the 3x growth might be doing exactly this). Has the team stress-tested their loan book against a 2015-style LatAm recession?
In VC, you need this to return 5-10x to matter for the fund. At $150M pre, you need a $750M-$1.5B exit. How many LatAm fintech lending companies have achieved that? Very few. Is there a credible path to that scale, or are you underwriting a "good but not great" outcome that doesn't move the needle for your fund? A 2-3x return on a $5M check isn't worth the illiquidity.
The models flash multiple warning signs: rich valuation with thin margins, social proof substituting for independent conviction, and a lending model that hasn't been stress-tested through a downturn. The company might be good, but the price isn't.
→ Pass at this valuation, or negotiate down to 8x revenue ($96M pre). If they won't budge (hot round, they don't need your money), that itself is information — you're not adding enough value to get favorable terms. Keep the relationship warm and revisit if growth stalls and the next round is flat/down.
Your team is 45 people and "already stretched thin." That's the constraint. Adding EU expansion doesn't just add work — it splits focus across two markets, two regulatory regimes, two time zones. The bottleneck isn't opportunity; it's execution capacity. Before expanding, ask: what's the constraint on growing faster in the US? If you haven't maxed out your home market, why split resources?
Every engineer building GDPR infrastructure is an engineer not building features that accelerate US growth. Every dollar spent on EU go-to-market is a dollar not spent on US sales. At 40% YoY growth, your US business is compounding fast — the opportunity cost of slowing that compounding is enormous. What's the expected EU revenue in year 1? Probably $500K-$1M. Is that worth risking your 40% growth rate?
First order: expand to EU, get new revenue. Second order: team splits focus, US product velocity slows, existing customers notice. Third order: US churn increases slightly, growth rate drops from 40% to 30%, and your Series A valuation takes a hit. Meanwhile, the EU business takes 18 months to reach meaningful scale. You've traded a strong position for two mediocre ones.
What works at $8M ARR in the US might not work at $0 in Europe. EU is not a homogeneous market — Germany, France, and Spain have different buying cultures, languages, and compliance requirements. Your playbook doesn't transfer 1:1. You'd essentially be a startup again in a new market, but with the overhead of a 45-person company. Small companies expanding too early into international markets is one of the most common scaling mistakes.
Hiring an EU team and building GDPR infrastructure is semi-irreversible: the infrastructure cost is sunk, and letting go of local employees in Europe is expensive and slow (EU labor law). This isn't a "try it and see" decision — it's a 12-month, $1M+ commitment before you know if it works. If you're going to do it, you need to commit fully, not test with half-measures.
Every model points the same direction: don't do it yet. Your constraint is execution capacity, not market opportunity. Expanding now means doing two things poorly instead of one thing well. The German customer is one data point — don't let a single voice drive a strategic decision.
→ Say "not yet, but soon" to the German customer. Set a clear trigger: "We'll expand to the EU when we hit $15M ARR and have 70+ people." Meanwhile, serve the German customer from the US (many SaaS companies do this successfully). Use the next 12 months to build the GDPR foundation incrementally without a full launch.
Instead of reasoning by analogy ("others do it this way, so we should too"), strip away assumptions until you reach bedrock truths that cannot be reduced further. Then rebuild your reasoning from those truths.
Elon Musk was quoted $65M for a rocket. Instead of accepting the market price, he broke it down to raw materials (aluminum, titanium, carbon fiber) — costing ~2% of the price. SpaceX built their own rockets for a fraction of the cost.
Overusing it on problems where analogy is perfectly fine. Not every decision needs to be derived from physics. First principles thinking is expensive — reserve it for high-stakes, novel problems.
Flip the problem. Rather than "how do I build a great company?", ask "what would guarantee this company fails?" Then systematically avoid those things. Often reveals blind spots that forward-thinking misses.
"All I want to know is where I'm going to die, so I'll never go there." Charlie Munger's approach: identify what destroys value (excessive debt, dishonest management, unscalable unit economics) and simply avoid it.
Using inversion alone without also thinking forward. It's a complement to positive reasoning, not a replacement. You can avoid all failures and still not build anything meaningful.
First-order thinking: "This action will produce X." Second-order: "And then X will cause Y, which leads to Z." Most people stop at the first order. Competitive advantage comes from thinking further.
First-order: rent control makes housing cheaper. Second-order: landlords reduce maintenance and stop building new units. Third-order: housing supply shrinks, making the problem worse long-term.
Analysis paralysis — going to the 5th or 6th order when the uncertainty makes it meaningless. Two to three orders is usually sufficient.
When multiple explanations are equally consistent with the evidence, prefer the one with the fewest assumptions. Not because simplicity is always right, but because complex explanations have more ways to be wrong.
"When you hear hoofbeats, think horses, not zebras." A patient with a headache more likely has tension or dehydration than a brain tumor. Start with the simple explanation and escalate only when evidence demands it.
Dismissing complex-but-true explanations because they're uncomfortable. Occam's Razor is a tiebreaker between equally supported theories, not a license to ignore evidence.
Every model, plan, or representation is a reduction of reality. The menu is not the meal. The financial model is not the business. Useful models acknowledge their own limitations.
Risk models rated mortgage-backed securities as safe because historical data showed low default rates. The models couldn't capture a systemic shift — the territory had changed, but the map hadn't.
Using "all models are wrong" as an excuse to avoid modeling at all. The point is not to abandon models, but to remember their limits and update them.
Everyone has areas where they have deep, earned knowledge and areas where they're tourists. The danger isn't ignorance — it's not knowing where your competence ends. Operate inside your circle; be cautious outside it.
Warren Buffett avoided technology stocks for decades — not because tech was bad, but because he knew he didn't understand it deeply enough. He stayed inside his circle (consumer brands, insurance, banking) and outperformed.
Using it as an excuse to never expand your knowledge. The circle should grow over time — the point is to be honest about where it is today.
When someone does something harmful, the most likely explanation is usually not evil intent but rather ignorance, laziness, or simple mistakes. Assuming good faith first leads to better relationships and more accurate assessments.
Your colleague didn't CC you on an important email. Malice interpretation: they're undermining you. Hanlon's interpretation: they were rushed and forgot. The second is almost always more accurate.
Being naive about genuinely bad actors. Hanlon's Razor is a default assumption, not an absolute rule. When evidence of malice accumulates, update your model.
Jeff Bezos classifies decisions as Type 1 (irreversible, high stakes — walk through carefully) or Type 2 (reversible, low stakes — decide fast and iterate). Most decisions are Type 2, but organizations treat them all as Type 1, slowing everything down.
Amazon launches products quickly because most launches are reversible — if a product fails, kill it. But acquisitions (like Whole Foods for $13.7B) get deep analysis because they can't be undone.
Treating everything as reversible. Some decisions compound — a bad hire might be "reversible" technically, but the damage to team culture lingers.
Every choice has a hidden cost: the best alternative you didn't choose. Spending 3 hours in a meeting isn't just 3 hours — it's the product work, thinking time, or rest you sacrificed. Make the invisible visible.
Investing $5M in Company A doesn't just cost $5M — it costs the potential return from Company B, C, or D that you can't fund now. The best investors obsess over opportunity cost, not just absolute returns.
Paralysis from constantly calculating alternatives. At some point you have to commit. Opportunity cost is for evaluating options, not for perpetual indecision.
Engineers build bridges to handle 3x the expected load. Value investors buy at a discount to intrinsic value. The principle: always leave room for error, surprise, and bad luck. The world is less predictable than your model.
Benjamin Graham's core principle: if you calculate a stock is worth $100, don't buy at $95. Buy at $65. The 35% discount is your margin of safety against estimation errors, market shocks, and unknowns.
Being so conservative you never act. An infinite margin of safety means zero decisions. Calibrate the margin to the stakes and uncertainty involved.
Before starting, assume the project has already failed spectacularly. Ask everyone: "It's 12 months from now and this was a disaster. What went wrong?" This overcomes optimism bias and surfaces risks people are too polite to mention upfront.
A startup pre-mortems their product launch: "We launched and nobody cared." Possible causes: wrong audience, bad timing, feature nobody needed. This exercise revealed they were building for developers when their actual users were ops teams.
Using it to kill bold ideas. Pre-mortems identify risks to mitigate, not reasons to avoid action. If every risk kills the project, you're using it as a veto tool.
Plot tasks on two axes: urgency and importance. Important + Urgent: do now. Important + Not Urgent: schedule (this is where strategic work lives). Urgent + Not Important: delegate. Neither: eliminate. Most people spend all day in quadrants 1 and 3, neglecting quadrant 2.
Responding to Slack messages feels urgent but is rarely important. Building a hiring pipeline feels non-urgent but is critical. The matrix forces you to protect time for what actually compounds.
Mechanically categorizing everything instead of developing the judgment to recognize importance intuitively. The matrix is training wheels.
For big life decisions, imagine yourself at 80 looking back. Which option would you regret not trying? This framework cuts through short-term fear and social pressure by anchoring to your long-term values.
Jeff Bezos used this framework to decide whether to leave his well-paying Wall Street job to start Amazon. At 80, would he regret not trying? The answer was obvious. He quit the next day.
Using it for small decisions where it's overkill, or rationalizing reckless choices as "I'd regret not doing it." It works best for genuine fork-in-the-road moments.
"Show me the incentive and I'll show you the outcome." People respond to incentives — financial, social, psychological. If you want to predict behavior, look at the incentive structure, not the stated intentions.
Employees were incentivized to open new accounts (sales quotas). They created millions of fake accounts. The employees weren't evil — the incentive structure made fraud the rational response.
Assuming all behavior is rational and incentive-driven. Humans also act on emotion, habit, identity, and social pressure. Incentives are the strongest predictor but not the only one.
Once you form a belief, you unconsciously seek information that confirms it and dismiss information that contradicts it. This applies to investment theses, political views, hiring decisions, and self-image.
An analyst who's bullish on a company will unconsciously weight positive data points more heavily. The fix: actively assign someone to argue the bear case, or force yourself to write the counter-thesis first.
Accusing others of confirmation bias while being blind to your own. Everyone has it. The goal isn't to eliminate it (you can't) but to build processes that counteract it.
You've invested $2M in a failing project. The rational question is "should we invest more given future prospects?" But the emotional pull is "we can't waste the $2M we already spent." The sunk cost is gone either way — it shouldn't affect the forward-looking decision.
The British and French governments kept funding the Concorde long after it was clear it would never be commercially viable — because they'd already invested too much to "waste." The term "Concorde fallacy" comes from this case.
Using "sunk cost" as an excuse to abandon things prematurely. Sometimes persistence is rational — the key is to evaluate based on future expected value, not past expenditure.
In ambiguous situations, humans look to others for cues on correct behavior. This is why restaurants with long lines attract more customers, and why market bubbles form — everyone assumes others know something they don't.
When a top-tier VC leads a round, other investors pile in — not because they independently evaluated the deal, but because "Sequoia invested, so it must be good." Social proof drives capital allocation more than most VCs admit.
Dismissing all social proof as irrational. Sometimes the crowd is right, and copying others is an efficient information shortcut. The danger is when you follow the crowd without checking their reasoning.
Our brains can't handle randomness. So we weave narratives: "The stock went up because of the Fed meeting." Maybe. Or maybe it was random. Post-hoc narratives feel true and satisfying, but they often assign causation to correlation.
"They succeeded because they had great product-market fit and a visionary founder." Survivorship bias + narrative fallacy. Many companies with identical characteristics failed. The story is constructed after the outcome is known.
Becoming so skeptical of narratives that you can't make decisions. Stories are how humans communicate and coordinate. Use them, but hold them lightly.
Losing $100 feels about as bad as gaining $200 feels good. This asymmetry shapes decisions everywhere: people hold losing stocks too long (hoping to avoid realizing the loss), avoid beneficial risks, and overvalue what they already have (endowment effect).
Investors sell winners too early (locking in gains) and hold losers too long (avoiding the pain of realizing losses). The rational approach is the opposite: let winners run, cut losers. Loss aversion makes this psychologically brutal.
Using it to justify never selling anything. Sometimes a loss is a loss and should be taken. The model explains the bias — it doesn't say the bias is always wrong.
Beginners lack the knowledge to recognize their own incompetence, so they overestimate their ability. Experts know enough to understand how much they don't know, so they're less confident. Competence and confidence are inversely correlated — until deep expertise.
The most vocal crypto "experts" on social media are often people who bought their first Bitcoin 3 months ago. Meanwhile, actual protocol developers and researchers qualify every statement with uncertainty.
Using it to dismiss anyone who's confident. Sometimes confident people are right. The model describes a statistical tendency, not a universal law.
Reinforcing loops amplify (growth begets more growth). Balancing loops stabilize (thermostat adjusts temperature). Understanding which loops drive a system tells you whether it will explode, collapse, or reach equilibrium.
More users join WhatsApp → more of your contacts are on it → more reason to join → more users. This reinforcing loop drove exponential growth until market saturation (a balancing loop) kicked in.
Assuming all reinforcing loops are infinite. Every reinforcing loop eventually hits a balancing force — resource limits, competition, regulation, or saturation.
Individual ants follow simple rules (follow pheromones, carry food). No ant understands the colony's architecture. Yet together they build complex structures, optimize supply chains, and solve problems. The system has properties that no individual component has.
No single trader sets the market price. Yet millions of individual buy/sell decisions create a price discovery mechanism that's astonishingly efficient (most of the time). Market prices emerge from decentralized interactions.
Using "emergence" as a hand-wave to avoid analyzing mechanisms. Emergence doesn't mean "magic" — it means the mechanism operates at a higher level than the components.
In any process, there's one constraint that limits overall throughput. Improving anything else is wasted effort until the bottleneck is addressed. Find the bottleneck → exploit it → elevate it → repeat.
A fintech startup has great product and marketing but terrible onboarding (60% drop-off). Spending more on marketing just pushes more users into a broken funnel. The bottleneck is onboarding — fix that first.
Assuming there's always exactly one bottleneck. In complex systems, there can be multiple constraints interacting. The model simplifies usefully but don't expect surgical precision.
What works at small scale often breaks at large scale, and vice versa. A 5-person startup communicates informally; a 500-person company needs process. An ant can fall from any height and survive; a human can't. Scale changes the rules.
The "move fast and break things" culture that built Facebook's early product became a liability at 3 billion users, when "breaking things" meant destabilizing elections. What worked at startup scale became dangerous at platform scale.
Assuming that what works for large companies works for small ones (or vice versa). Amazon's operational playbook won't save your 3-person startup.
The first hour of study is highly productive. The 12th hour, much less so. The first engineer on a project adds huge value. The 50th adds less than the 49th. Knowing where the curve flattens tells you when to stop investing and reallocate.
The first 3 features of a product capture 80% of user value. Features 4-20 each add marginal value but multiply complexity and maintenance burden. Knowing when you've hit diminishing returns prevents feature bloat.
Assuming diminishing returns apply everywhere. Some domains have increasing returns (network effects, knowledge accumulation, trust-building). Know which game you're playing.
Archimedes: "Give me a lever long enough and I can move the world." In systems, leverage points are where a small change shifts the whole system. In business: code, media, and capital are leverage because they scale without proportional effort.
One developer writes code once; it serves millions of users. Naval Ravikant calls this "permissionless leverage" — you don't need anyone's permission to create software, write content, or invest capital. These are the modern levers.
Confusing leverage with shortcuts. Leverage amplifies both good and bad outcomes. Financial leverage (debt) amplifies returns in good times and losses in bad times. Use it deliberately.
Instead of "this will work" or "this won't work," ask "what's the probability this works?" Then evaluate whether the expected value (probability × payoff) justifies the cost. Update probabilities as new information arrives (Bayesian thinking).
Annie Duke (professional poker player and decision scientist): a good decision can have a bad outcome and vice versa. You pocket aces and lose — that doesn't mean the bet was wrong. Evaluate decision quality by process, not outcome.
False precision — saying "there's a 73.2% chance" when you barely understand the domain. Probabilities should reflect genuine uncertainty. Use ranges, not false point estimates.
Black Swans are events that are (1) rare, (2) have extreme impact, and (3) are retrospectively predictable but not prospectively. The 2008 crisis, COVID, the internet. You can't predict them, but you can build systems that are robust (or even antifragile) to them.
A global pandemic was on every risk register but nobody acted on it. When it hit, companies with cash reserves and flexible operations survived. Those running lean with zero buffer collapsed. The Black Swan rewarded preparation, not prediction.
Calling every surprising event a "Black Swan." If it was reasonably foreseeable and you just didn't prepare, that's not a Black Swan — that's negligence.
Fragile breaks under stress. Robust survives. Antifragile gets better. Your muscles are antifragile (stress makes them stronger). A startup that learns from failures is antifragile. Design systems that benefit from volatility rather than just surviving it.
Put 90% in extremely safe assets and 10% in extremely risky bets. The safe portion protects you; the risky portion has unlimited upside. You're protected from blowups while benefiting from Black Swans. Taleb's barbell makes the portfolio antifragile.
Seeking stress and volatility for their own sake. Not all stress is productive. Antifragility requires the right kind of stress, at the right dose, with recovery time.
After an exceptionally good year, expect a more normal one. After a terrible quarter, expect improvement — not because anything changed, but because extreme results often reflect luck that won't persist. Don't over-reward peaks or over-punish troughs.
A fund manager who beats the market by 30% in one year is likely to underperform relative to that benchmark next year. It's not that they got worse — the initial outperformance included a luck component that naturally fades.
Assuming everything regresses. Genuine structural advantages (moats, network effects, talent) can sustain above-average performance. Distinguish luck from skill before applying this model.
We study successful companies, surviving species, and winning strategies — and draw conclusions from an incomplete dataset. The failures are invisible. Any conclusion drawn only from survivors is biased toward overestimating success rates and misidentifying causes.
The military examined returning bombers and found bullet holes concentrated on wings and fuselage. They planned to reinforce those areas. Statistician Abraham Wald realized the opposite: planes hit in the engines didn't return. Armor the places with NO holes.
Dismissing all success stories as survivorship bias. Sometimes successful patterns are real. The fix isn't to ignore winners — it's to also study the losers who did the same thing.
When someone bears the downside of their decisions, they make better decisions. When they don't (asymmetric risk), expect careless advice and reckless behavior. Look for who pays the price of failure before trusting their judgment.
Investment bankers earned bonuses on deals that later blew up. They had upside but no downside — skin in the game was asymmetric. When the crisis hit, taxpayers absorbed the losses. The incentive structure guaranteed recklessness.
Requiring skin in the game for every opinion. Academic researchers, journalists, and advisors can provide valuable perspectives without personal financial exposure. The model is most important for decision-makers, not commentators.
A moat is a structural advantage that's hard to replicate: network effects, switching costs, brand, patents, cost advantages, or regulatory capture. Without a moat, profits attract competition until margins disappear.
Visa's moat is a multi-sided network effect: more merchants accept Visa → more consumers carry it → more merchants accept it. Plus switching costs (replacing infrastructure) and scale economics (processing cost per transaction shrinks with volume). This moat has lasted 60+ years.
Calling any advantage a "moat." First-mover advantage, a good team, or current market share are not moats — they're temporary leads. A moat must be structural and durable.
New technologies and business models inevitably displace existing ones. The automobile destroyed the horse industry. Streaming killed Blockbuster. This isn't a bug — it's how capitalism evolves. Incumbents resist; disruptors embrace.
Traditional banks built branches and relationships. Fintechs built apps and APIs. The banks' strengths (physical presence, trust) became weaknesses (cost, slowness). Creative destruction doesn't eliminate the function (banking) — it replaces the form.
Assuming every new technology is destructive. Most innovations are sustaining (improving existing products) not disruptive (creating new markets). True creative destruction is rarer than the hype suggests.
Even if you're better than everyone at everything, you should still specialize in what you're most disproportionately better at. Because time is finite, focus on activities where your relative advantage is greatest and trade for the rest.
A CEO might type faster than their assistant. But every hour spent on admin is an hour not spent on strategy, fundraising, or key decisions — where their comparative advantage is massive. Delegate despite absolute skill.
Using it to permanently avoid developing weaknesses. Comparative advantage is about allocation now; you can still invest in expanding your capabilities over time.
80% of results come from 20% of causes. In venture capital, one investment returns more than all others combined. In sales, a few clients generate most revenue. Identifying and doubling down on the vital few is more important than optimizing the trivial many.
Peter Thiel's rule: a VC fund's best investment should return more than the entire rest of the fund combined. This is why VCs seek outliers, not "good" returns. The power law means average is irrelevant — only the tail matters.
Applying the 80/20 ratio literally everywhere. The ratio varies. And in some domains (safety, compliance), the "trivial many" matter critically. Not everything follows a power law.
In strategic interactions, your optimal choice depends on others' choices. Key concepts: Nash Equilibrium (no one benefits from changing strategy alone), Prisoner's Dilemma (individual rationality leads to collective irrationality), and repeated games (cooperation emerges when the game repeats).
Two airlines on the same route. If both keep prices high, both profit. If one cuts prices, they steal market share temporarily — but the other retaliates, leading to a price war where both lose. This is a Prisoner's Dilemma in action.
Over-applying game theory to situations with incomplete information. Real-world "games" are messier than textbook examples — players are irrational, information is asymmetric, and the rules keep changing.
37 MODELS ACROSS 6 CATEGORIES
This library is continuously updated. Suggest a model →