The model library

First-principles thinking strips a belief back to the parts you know to be true — physics, arithmetic, verified facts — and rebuilds from those, rather than reasoning by analogy to 'how it's usually done'. It is the technique of the scientist and the engineer: refuse inherited assumptions and ask what the ground floor actually is.

Most 'impossible' constraints turn out to be conventions or someone else's conclusion carried along unexamined. Decompose, keep only what survives scrutiny, and reassemble.

Example situations:

• Costing a product from raw materials and labour instead of from what competitors charge.
• A process everyone follows 'because that's how it's done' that no one can justify from scratch.

Source: The Great Mental Models (Farnam Street)

A map is useful precisely because it leaves things out — but the reduction that makes it usable also makes it wrong in ways that matter when the terrain changes. Financial models, org charts, job titles, and reputations are all maps: abstractions that can drift dangerously far from the ground they claim to represent.

Treat every model as a tool with a shelf life. Check it against the territory, and distrust it most exactly when it is most convenient.

Example situations:

• Trusting a five-year-old process document over what the team actually does today.
• A risk model that worked until the conditions it was fitted to quietly changed.

Source: The Great Mental Models (Farnam Street)

A thought experiment reasons rigorously through a hypothetical — Einstein chasing a beam of light, Rawls's veil of ignorance — to test an idea's implications before paying to test it in the world. It lets you explore the impossible, the irreversible, and the not-yet-built cheaply.

The discipline is to follow the setup honestly to its conclusion, including the parts you'd rather not reach, instead of stopping at the answer you wanted.

Example situations:

• Designing a policy as if you didn't know which affected group you'd end up in.
• Pre-mortem: imagine the project has failed spectacularly and work backward to why.

Source: The Great Mental Models (Farnam Street)

Popper's line between science and everything else: a real theory forbids something, so it can be refuted by observation; an unfalsifiable claim explains every outcome and therefore predicts none. Seeking confirmation is easy and worthless — one genuine attempt at refutation is worth a hundred agreeable data points.

Before trusting a belief, ask what would falsify it. If nothing could, you hold a faith, not a finding.

Example situations:

• A strategy 'thesis' that would be declared correct no matter what the market did.
• A metric that only ever gets interpreted as good news.

Source: Karl Popper

Good decisions can have bad outcomes and vice versa. Estimate the odds and sizes of each outcome, multiply, compare — and update the odds honestly as evidence arrives (Bayesian habit).

A small chance of ruin outweighs a likely small gain: expected value must respect survival first.

Example situations:

• An 80%-likely deal worth 10k versus a 20%-likely deal worth 100k.
• Judging a hire as 'wrong' because one reference was lukewarm and later proved decisive.

Source: Fermat/Pascal via Poor Charlie's Almanack

Genuine randomness produces streaks, clusters, and apparent trends that mean nothing, but human cognition is a relentless pattern-detector that manufactures explanations for chance. Mistaking noise for signal — and then acting on the invented story — is one of the most common and expensive errors in judgment.

Before explaining a fluctuation, ask what pure randomness would have produced. Often the honest answer is 'exactly this'.

Example situations:

• Rewarding or firing a manager over a swing that's within normal random variation.
• Seeing a 'hot streak' in results that a fair coin would reproduce just as often.

Source: Mathematics / statistics

Almost everything you know comes from a sample of a larger whole, and if the sample is small or unrepresentative, the conclusion is unreliable no matter how confidently it's held. Biased sampling — surveying only the reachable, the loud, or the surviving — quietly corrupts most everyday inference.

Ask of any generalisation: how many cases, and how were they chosen? A vivid handful is not data; a skewed thousand is worse, because it looks like data.

Example situations:

• Redesigning a product around the few users who bothered to complain.
• Judging a market from the customers a biased channel happened to send you.

Source: Mathematics / statistics

Wherever luck contributes, outstandingly good or bad performance tends to drift back toward the average — with no cause required. We invent stories (the pep talk worked, the punishment worked) for what is mostly statistics.

Before crediting an intervention, ask what would have happened anyway.

Example situations:

• The star salesperson's record quarter is followed by a normal one — and gets blamed on complacency.
• A struggling clinic improves right after a consultant visit that changed nothing.

Source: Francis Galton

Studying successful companies, funds, or people samples only what survived. The graveyard is invisible, so the traits of survivors look causal when they may be common to failures too.

Ask: what would the full population, including the dead, show?

Example situations:

• Copying a famous founder's risky habits without counting the identical founders who went broke.
• Armouring WWII bombers where returning planes show holes — instead of where the lost planes were hit.

Source: Abraham Wald / statistics

Compounding is the arithmetic of exponential growth: value that grows on top of previous growth. Its power is invisible over weeks and overwhelming over decades — in money, knowledge, relationships, and damage.

The practical corollaries: start early, don't interrupt compounding unnecessarily, and beware small recurring costs or harms, which compound too.

Example situations:

• Choosing between a slightly higher salary and a role where you learn twice as fast.
• A team that loses 2% of trust every sprint through missed commitments.

Source: Poor Charlie's Almanack

In multiplication a single zero collapses the entire result to zero — no amount of value elsewhere can compensate. Many real situations multiply rather than add: one fatal flaw, one broken link, one act of dishonesty can nullify enormous accumulated worth.

Hunt for the zeros. A plan with one multiplicative weakness isn't 'mostly fine' — it's a plan whose expected value is zero until that factor is fixed.

Example situations:

• A brilliant product with a checkout that doesn't work — revenue is zero.
• A talented team whose one dishonest member voids the trust the rest depends on.

Source: Mathematics

On a landscape of outcomes, a local maximum is a peak higher than everything immediately around it — but not the highest peak (the global maximum). Improvement that only ever climbs uphill gets stuck on the nearest hill, and reaching a better summit requires first going down.

When progress stalls at 'good', ask whether you're on a local peak. Escaping it usually means accepting a temporary loss to cross the valley toward something higher.

Example situations:

• A profitable business model that must get temporarily worse to reach a far better one.
• Endless A/B tweaks that optimise a page you should have replaced entirely.

Source: Mathematics / optimization

Geometry: as a shape grows, volume rises faster than surface area (the square-cube relationship), which is why big animals overheat slowly and small ones fast. The abstraction is exposure — how much of a thing is in contact with the outside — and it drives everything from reaction speed to luck to vulnerability.

Increasing surface area increases both opportunity (chances to get lucky, react, connect) and hazard (attack points, things that can break). Choose deliberately which you want more of.

Example situations:

• Publishing more widely to grow luck's surface area — and inviting more criticism with it.
• A system whose expanding attack surface grows its breach risk faster than its size.

Source: Mathematics / geometry

In mathematics, an equation can be transformed into equivalent forms that are far easier to solve without changing its truth. The general move is to re-express a problem — reframe, re-unit, invert, change coordinates — until a version appears in which the solution is visible.

When stuck, don't only push harder on the current framing; look for an equivalent statement of the same problem where the path is clear.

Example situations:

• Turning a confusing 'per month' comparison into 'per use' and seeing the answer instantly.
• Reframing 'how do we grow?' as 'why do people leave?' and finding the lever.

Source: Mathematics

Outcomes are rarely spread evenly: a few customers drive most revenue, a few bugs most crashes, a few decisions most of the year's results. Find the vital few and treat them differently from the trivial many.

Apply it recursively — the top 20% has its own 20%.

Example situations:

• Support load: five defects generate most tickets; fix those first.
• One distribution channel quietly produces most qualified leads.

Source: Vilfredo Pareto

Borrowed from engineering: design a bridge to hold far more than its expected load. Because your estimates and models will be wrong in ways you can't predict, decisions should still work when the inputs are worse than expected.

The question is not 'what happens if I'm right?' but 'how bad is it if I'm wrong by a lot?'

Example situations:

• Committing to a mortgage that only works if nothing goes wrong with your income.
• Scheduling a project with zero slack before a hard deadline.

Source: Benjamin Graham via Poor Charlie's Almanack

In physics, measurements of time and motion depend on the observer's frame; there is no single privileged vantage point. The human parallel is just as real: each person judges from a position they didn't choose and usually can't perceive, mistaking their local view for the view.

To understand a situation, deliberately occupy other frames — the customer's, the opponent's, the newcomer's — because your own is invisible to you and therefore the one you'll forget.

Example situations:

• A feature that's obvious to the team that built it and baffling to every new user.
• Two departments 'seeing the same numbers' and drawing opposite conclusions from them.

Source: Physics (Einstein)

An object at rest stays at rest and one in motion keeps its course until an external force acts. People, teams, and companies carry the same momentum: established routines persist not because they're right but because changing them requires force proportional to the mass involved.

Both halves matter. Stationary things are hard to start; moving things are hard to stop or steer. Budget for the force, and use existing momentum rather than fighting it.

Example situations:

• A legacy process that survives every reorg because no one supplies the force to end it.
• A declining product that keeps selling on brand momentum long after the work stopped.

Source: Physics (Newton)

Velocity is speed plus direction — a vector, not a number. Pure activity (high speed, no direction) feels productive and goes nowhere, while modest, consistently directed effort compounds into distance covered. Motion is not progress.

When evaluating effort — your own or a team's — ask not 'how fast are we moving?' but 'how fast toward what?' Two teams working equally hard can have opposite velocities.

Example situations:

• A busy quarter of shipped features that left the core metric exactly where it started.
• Choosing between a frantic team with no roadmap and a steady one aimed at a clear goal.

Source: Physics

Friction opposes motion and viscosity resists flow; both convert useful energy into waste heat and both are unavoidable in real systems. In organisations the analogues are handoffs, approvals, and ambiguity — each a small drag that is invisible in isolation and decisive in aggregate.

You improve throughput less by adding power than by removing friction. Find the sticky steps and smooth them before you push harder.

Example situations:

• A signup that loses most users not to one wall but to ten tiny points of resistance.
• A 'simple' change that needs four sign-offs, each adding a day of drag.

Source: Physics

Give me a place to stand and a lever long enough, said Archimedes, and I'll move the world. Leverage is any mechanism — a tool, a relationship, capital, code, a distribution channel — that multiplies the effect of your effort far beyond its size.

The skill is locating the point where a small, well-placed push moves something large, rather than pushing hard on the load directly. But leverage magnifies mistakes as faithfully as successes.

Example situations:

• Automating a task once so a thousand future runs cost nothing.
• One reference customer whose word opens a whole market segment.

Source: Physics (Archimedes)

Newton's third law: push on the world and the world pushes back with equal force. The physical principle has a behavioural twin — pressure invites counter-pressure, aggression invites retaliation, and a hard shove on any system tends to generate an opposing force you'll have to absorb.

Before applying force, account for the reaction it will create. The reaction is not a side effect; it is guaranteed by the same act that produced the action.

Example situations:

• A price war whose opening move guarantees the competitor's matching cut.
• A heavy-handed policy that produces exactly the resistance it was meant to prevent.

Source: Physics (Newton)

The second law says entropy — disorder — tends to increase in a closed system. Anything organised (a codebase, a garden, a team's discipline, a clean house) drifts toward mess unless energy is continuously spent maintaining it. Order is not the natural state; it is a purchase you keep re-making.

The practical reading: budget ongoing energy for upkeep, not just for creation. Neglect isn't neutral — it is a slow slide back to chaos.

Example situations:

• A codebase that rots into unmaintainability the moment refactoring effort stops.
• Standards that quietly erode unless someone keeps spending energy to hold the line.

Source: Physics (Second Law of Thermodynamics)

Below a threshold, added input seems to do nothing; past it, the reaction becomes self-sustaining — chain reactions, epidemics, network effects, bank runs. Linear extrapolation fails exactly where it matters most.

Locate the threshold and which side of it you're on before judging 'it isn't working'.

Example situations:

• A marketplace that limps until enough buyers and sellers make it self-sustaining.
• Team morale that absorbs three departures and collapses at the fourth.

Source: Physics / network theory

Engineers assume components fail and design so no single failure is fatal — backups, margins, independent paths. The same discipline applies to finances, staffing, suppliers, and plans.

Redundancy looks like waste right up until it saves you; efficiency maximised is fragility maximised.

Example situations:

• A team where exactly one person understands the deployment system.
• Running finances so one late payment triggers a cascade.

Source: Engineering via Poor Charlie's Almanack

A reaction can be strongly favourable overall yet refuse to start until enough energy is supplied to clear the activation barrier. Change in people and organisations has the same shape: the net benefit is positive, but nothing happens until someone absorbs the upfront cost of getting over the hump.

The lever is rarely more enthusiasm about the end state — it is lowering the barrier to the first step, or supplying a one-time push big enough to clear it.

Example situations:

• A habit that pays off for years still stalls because the first week is all cost and no reward.
• A migration everyone agrees is worth it never starts because no one will absorb the switching hump.

Source: Chemistry (Arrhenius)

A catalyst lowers the activation barrier so a reaction proceeds far faster, yet emerges unchanged and ready to act again. The analogue: certain people, tools, or introductions dramatically accelerate outcomes while spending almost nothing of themselves.

Look for catalytic leverage — the connector who unblocks a deal, the template that makes a task trivial — and protect it, because unlike fuel it is not consumed.

Example situations:

• One well-connected advisor turns months of cold outreach into a week of warm intros.
• A shared checklist that makes every future onboarding faster without extra effort each time.

Source: Chemistry

A chemical system at equilibrium, when disturbed by a change in pressure, heat, or concentration, shifts in the direction that partly offsets the disturbance. Human systems do the same: press on one variable and the system adjusts to absorb your intervention, blunting the effect you wanted.

Before forcing a change, ask how the system will move to cancel it — and whether you are fighting a restoring force you cannot out-push.

Example situations:

• Cut headcount and the survivors' overtime quietly restores the old cost — and the old backlog.
• Crack down on one channel of a banned activity and it re-routes to another to restore the flow.

Source: Chemistry (Le Chatelier)

A solvent dissolves added solute readily until it reaches saturation; beyond that point extra solute simply falls out, dissolved in nothing. Attention, markets, and teams saturate the same way — they absorb input smoothly until a limit, after which more input is wasted or actively harmful.

Find the saturation point before adding more: past it, extra effort, spend, or messaging precipitates out as noise.

Example situations:

• A tenth notification does nothing the third didn't — the user's attention saturated long ago.
• Pouring more budget into a channel that stopped converting several units of spend ago.

Source: Chemistry

Mixing metals produces an alloy — bronze, steel — that is not an average of its ingredients but something qualitatively stronger or more useful than any of them: small additions transform the whole. Teams, skills, and ideas combine the same way, where the right blend yields capabilities none of the parts possessed.

The lesson is that composition is nonlinear. A tiny proportion of the right added element can change the character of the entire mixture, for better or worse.

Example situations:

• A pinch of a rare skill on a team that changes what the whole group can attempt.
• One toxic hire whose small presence embrittles an otherwise strong culture.

Source: Chemistry

Variation plus selection plus replication relentlessly shapes populations toward whatever the environment rewards, with no designer needed. Markets, memes, and org cultures evolve the same way.

Ask what your environment actually selects for — it is what you will get more of, regardless of intentions.

Example situations:

• Bacteria under half-finished antibiotic courses; competitors under weak regulation.
• A company that promotes firefighters breeds arsonists.

Source: Charles Darwin

In Carroll's story the Red Queen tells Alice it takes all the running you can do to stay in the same place. Because competitors and parasites adapt to your every advantage, standing still means falling behind — perpetual improvement is the price of merely surviving, not of getting ahead.

This is why moats erode, why security is a treadmill, and why 'we already won' is the most dangerous thing a market leader can believe.

Example situations:

• A security posture that must improve every year just to face the same relative threat.
• A dominant product overtaken because it stopped running while rivals kept adapting.

Source: Biology (Leigh Van Valen) / Lewis Carroll

Species and businesses thrive by occupying a niche — specialised fit beats general strength. But niches are defined by the surrounding ecosystem: change the climate, the predator, or the platform, and yesterday's perfect fit is stranded.

Know your niche, who else wants it, and what your dependence chain is.

Example situations:

• A profitable agency built entirely on one platform's algorithm.
• A niche product crushed not by a rival but by its distribution channel drying up.

Source: Ecology via Poor Charlie's Almanack

Living things are wired to preserve themselves, and the drive runs deep enough to override almost everything else under threat. Institutions inherit the same instinct: a department, agency, or company will protect its own existence even when doing so undermines the mission it was created for.

When behaviour looks baffling, ask whether something's survival feels threatened. Self-preservation explains a great deal of otherwise 'irrational' resistance.

Example situations:

• A team that quietly buries an efficiency that would eliminate its own roles.
• An initiative kept alive long past its usefulness because ending it ends jobs.

Source: Biology

Life is built on replication — patterns that reproduce themselves spread, and small copying advantages compound into overwhelming prevalence. The same logic governs ideas, content, code, and business models: the format that copies fastest and most faithfully wins the population, regardless of its merit.

Ask of anything you want to spread: how easily and accurately does it replicate? Design for copyability, and beware harmful patterns that replicate better than good ones.

Example situations:

• A template or playbook that spreads across an org because it's trivial to copy.
• A meme or format that out-competes better ideas purely by being easier to share.

Source: Biology

Evolution is popularly read as red in tooth and claw, but symbiosis and cooperation — cells into organisms, individuals into societies — account for its largest leaps. Groups that solve the problem of working together outcompete collections of individuals who can't, which is why trust and reciprocity are so heavily selected for.

Most durable advantage comes from enabling cooperation at a scale rivals can't match, not from winning zero-sum fights.

Example situations:

• An open standard that beats a better proprietary one by letting everyone build on it.
• A team that outperforms more talented individuals because it actually coordinates.

Source: Biology

From cells to organs to organisms, and from individuals to teams to institutions, complexity is managed by nesting simpler units into higher-order ones. Each level has behaviours and rules that don't reduce cleanly to the level below, which is why you can't fully explain an organisation from its individuals alone.

When diagnosing a problem, first ask which level it lives at. Fixes aimed at the wrong layer — treating a system problem as an individual failing — reliably miss.

Example situations:

• Blaming individuals for outcomes their team structure made almost inevitable.
• A reorg that moves boxes but leaves the incentive layer driving the behaviour untouched.

Source: Biology

Evolution rewards economy: organisms that waste energy lose to those that don't, so a deep bias toward the least-effort path is built into behaviour, mental and physical alike. What looks like laziness is often this ancient efficiency drive choosing the cheapest route to a good-enough outcome.

Design around it rather than moralising about it: make the desired behaviour the low-energy path and it will happen; make it effortful and it won't, however well-intentioned people are.

Example situations:

• Users overwhelmingly take the default option because choosing costs energy.
• A safety step that's routinely skipped because the compliant path is the harder one.

Source: Biology

Munger called incentive-caused bias the most underestimated force in human affairs: 'Never, ever, think about something else when you should be thinking about the power of incentives.' People respond to what is rewarded and punished — including honest people who slowly rationalise what pays.

When behaviour looks irrational, first map who gets paid, promoted, praised or blamed for what.

Example situations:

• A consultant recommends the solution their firm happens to sell.
• A sales team hits quota with deals that churn in three months — check what the commission plan rewards.

Source: Poor Charlie's Almanack

Instead of asking how to succeed, invert: ask what would guarantee failure or make things worse, and systematically avoid those things. Munger: 'All I want to know is where I'm going to die, so I'll never go there.'

Inversion works because avoiding stupidity is easier than seeking brilliance, and because failure modes are often clearer and more enumerable than success paths.

Example situations:

• Planning a product launch: list everything that would make it flop, then check each is prevented.
• Improving a marriage or partnership: list behaviours that would destroy it, and stop doing them.

Source: Poor Charlie's Almanack

Affection distorts: we overrate what we like, forgive its flaws, and adopt its views wholesale. Salespeople, recruiters, and demagogues all work by getting themselves liked first.

When you notice you like the messenger, audit the message twice.

Example situations:

• Approving a weak proposal because it comes from a charming colleague.
• Buying into a company because you admire its founder.

Source: The Psychology of Human Misjudgment

The mirror image of loving: dislike makes us dismiss good ideas from bad sources and escalate small slights into feuds. Wars and workplace politics both run on it.

A useful discipline: state your opponent's position so well they'd accept the summary — then judge.

Example situations:

• Rejecting a competitor's genuinely better practice because it's theirs.
• Dismissing feedback because you dislike the person delivering it.

Source: The Psychology of Human Misjudgment

Unresolved doubt is uncomfortable, so we leap to conclusions — fastest when stressed or puzzled, which is exactly when we should be slowest. Cults and high-pressure sales exploit this deliberately.

If a decision can wait, doubt is information: let it work.

Example situations:

• Signing the lease at the end of an exhausting day of viewings.
• Diagnosing the first plausible cause of an outage and stopping the search.

Source: The Psychology of Human Misjudgment

Reaching a conclusion is satisfying, so the brain latches onto the first plausible answer and quits searching, then defends it against alternatives. It's efficient for trivia and dangerous for anything important, because the first idea to arrive is rarely the best one available.

Munger's discipline: treat the first conclusion as a hypothesis to be tested, not a verdict. Force a genuine second and third option into consideration before deciding.

Example situations:

• Diagnosing the first plausible cause of a bug and shipping a fix for the wrong thing.
• Hiring the first decent candidate because evaluating more feels like effort.

Source: The Great Mental Models (Farnam Street) / Charlie Munger

Once we've decided, said, or done something publicly, we bend new evidence to stay consistent with it — the human mind, as Munger says, works like a human egg: once one idea gets in, the rest are locked out. Habits, good and bad, ride the same machinery.

Beware especially of conclusions you announced loudly.

Example situations:

• Doubling down on a failing strategy you presented to the board.
• Keeping a hiring bar exception because admitting the mis-hire is worse than living with it.

Source: The Psychology of Human Misjudgment

Once a conclusion feels like ours, the mind works as a defence lawyer, not a judge: it hunts support and explains away conflict. Darwin's discipline was the antidote — he recorded disconfirming observations immediately, before his mind could dismiss them.

Operationally: state what evidence would change your mind before gathering evidence, and assign someone to argue the other side seriously.

Example situations:

• Reading ten reviews and only remembering the ones that support the purchase you already want to make.
• A founder who only interviews happy customers about a struggling product.

Source: Psychology of Human Misjudgment

Curiosity drove science before incentives did, and it counteracts doubt-avoidance and first-conclusion bias by keeping the question open a little longer. It is the one tendency to feed rather than guard against.

Cultivate it deliberately: follow the anomaly, read outside your field, ask the naive question.

Example situations:

• The engineer who chases the 'impossible' log line and finds the real bug.
• Reading a rival industry's playbook and importing its best trick.

Source: The Psychology of Human Misjudgment

Humans follow — and demand — reciprocal fair conduct, even at personal cost; people will burn value to punish cheaters. Systems perceived as unfair generate sabotage far beyond the direct grievance.

In any change (pricing, layoffs, rules), the fairness story matters as much as the economics.

Example situations:

• A pay policy that's economically rational but reads as favouritism, wrecking morale.
• Customers boycotting over a fee they could easily afford but find insulting.

Source: The Psychology of Human Misjudgment

High-trust relationships and societies move faster and cheaper because they don't pay the constant tax of verification, contracts, and defensive checks. Trust is slow to build and quick to destroy, and once lost it's replaced by friction — every interaction now needs proof, escrow, and oversight.

Treat trust as core infrastructure, not a soft nicety. The returns to being trustworthy compound, and the costs of a single betrayal ripple far beyond the incident.

Example situations:

• A team that ships fast because members don't re-litigate each other's work.
• A market that seizes up when counterparties stop trusting each other's word.

Source: The Great Mental Models (Farnam Street)

Munger (after Buffett): 'It's not greed that drives the world, but envy.' People accept a bad deal that beats their neighbour's over a good deal that trails it; compensation systems and bubbles run on this.

When a decision suddenly feels urgent because someone else just won, stop.

Example situations:

• Chasing an acquisition because a rival made one.
• A bonus pool that leaves everyone better off but half the team furious.

Source: The Psychology of Human Misjudgment

A small gift, concession, or compliment creates a disproportionate urge to give back; negotiators open with extreme asks so their 'concession' triggers yours. The same reflex escalates hostility tit-for-tat.

Accepting anything from a counterparty is never free; decide what it may cost first.

Example situations:

• The vendor's conference invitation before the contract renewal.
• Conceding too much because the other side 'moved first'.

Source: The Psychology of Human Misjudgment

Past success makes us repeat the strategy in changed conditions; prestige brands, attractive messengers, and lucky streaks all borrow credibility they didn't earn. Shooting the messenger is the dark side — bad news gets associated with its bearer.

Separate the thing from its packaging before judging it.

Example situations:

• Trusting a bad product because a prestigious firm endorses it.
• Repeating the playbook from your last startup in a completely different market.

Source: The Psychology of Human Misjudgment

Through conditioning, a stimulus that reliably precedes reward or pain comes to trigger the anticipatory feeling directly — Pavlov's dogs salivating at the bell. Humans do the same across brands, people, and places, forming automatic hope or fear responses that bypass judgment entirely.

Because the reaction is felt before it's reasoned, the defence is to notice the automatic pull and ask whether the thing actually merits it, or whether you're just responding to a bell.

Example situations:

• Warm feelings toward a brand built entirely by pleasant associations, not product quality.
• Dreading a task because it's paired in memory with an old bad experience of it.

Source: Psychology (Ivan Pavlov)

Facts too painful to accept get distorted or denied — failing ventures, addictions, dying relationships. Denial is strongest exactly where the stakes are highest, which is why the numbers should be looked at by someone without the pain.

If you can't bear a conclusion, that's a reason to suspect it's true.

Example situations:

• A founder ignoring cohort data that says the product isn't retaining.
• A family not naming a member's addiction for years.

Source: The Psychology of Human Misjudgment

Most drivers rate themselves above average; owners value their goods above market (endowment effect); and we overvalue our own conclusions because they are ours. Hiring interviews reward self-confident twins of the interviewer.

Munger's fix: force yourself to consider the disconfirming case, and judge people on track record, not impression.

Example situations:

• Refusing a fair acquisition offer because 'our' company must be worth more.
• Backing your own estimate over the base rate for projects like yours.

Source: The Psychology of Human Misjudgment

Beyond denial, humans display excess optimism even about neutral matters (Demosthenes' point). Plans assume best cases; lotteries and unrealistic roadmaps are bought with the same coin.

The antidote is arithmetic: base rates, reference-class forecasting, and margins of safety.

Example situations:

• A project plan where every workstream must go right to hit the date.
• Revenue forecasts that assume zero churn and instant sales ramp.

Source: The Psychology of Human Misjudgment

We overreact to losing (or almost getting) something: gamblers chase near-misses, managers escalate sunk commitments, and people take big risks to avoid small certain losses. Take away a privilege and the reaction dwarfs the original gratitude.

Ask: would I buy this position today at this price? If not, holding it is the same decision.

Example situations:

• Holding a crashing stock to 'get back to even'.
• A bidding war where losing the auction feels worse than overpaying.

Source: The Psychology of Human Misjudgment

When unsure, we look to others' behaviour as evidence of what's correct, which is often efficient and occasionally disastrous (bubbles, panics, bystander effects). The pull is strongest under uncertainty and stress, and when the others resemble us.

Ask: would I still do this if nobody else were doing it? Is the crowd here actually informed, or just mutually reassured?

Example situations:

• Every competitor adds an AI feature, so the roadmap gets one too — without a user problem attached.
• An overheated housing market where 'everyone is buying'.

Source: Psychology of Human Misjudgment

People are readily led: a leading question, a confident assertion, or an implied expectation can plant a belief or even a false memory that feels self-generated. Suggestibility rises with authority, uncertainty, and group pressure, and it's the mechanism behind loaded surveys, manipulative interviews, and manufactured consensus.

Guard the inputs to a decision. How a question is asked shapes the answer, so neutral framing and independent, un-primed opinions are worth deliberately engineering.

Example situations:

• A survey whose leading wording produces the 'finding' its author wanted.
• A meeting where the boss's stated preference silently becomes everyone's opinion.

Source: The Great Mental Models (Farnam Street)

A 30k car makes 1k floor mats feel cheap; a terrible option makes a mediocre one look good; small gradual changes slip beneath notice (the boiling frog). Salespeople and negotiators set the contrast deliberately.

Always compare against the absolute standard or the full market — not the anchor you were handed.

Example situations:

• Accepting a bad salary because the first offer was insulting.
• A codebase degrading one 'small exception' at a time.

Source: The Psychology of Human Misjudgment

Light stress can improve performance, but heavy stress produces fast, extreme, often faulty decisions — it supercharges social proof, doubt-avoidance, and denial. Pressure is deliberately manufactured by manipulators for exactly this reason.

Rule of thumb: the more stressed you are, the fewer irreversible decisions you should make.

Example situations:

• Panic-selling in a market crash.
• Accepting bad terms because the deadline was engineered to be tomorrow.

Source: The Psychology of Human Misjudgment

An idea or fact isn't more important because it's more available — but the brain acts as if it were. Dramatic anecdotes beat base rates; the last incident dominates the risk register.

Munger's checklist habit exists precisely for this: procedure forces the unavailable factors into view.

Example situations:

• Overinsuring against last year's rare disaster while ignoring the common one.
• Judging a candidate by one memorable interview answer.

Source: The Psychology of Human Misjudgment

Asked how likely something is, the mind substitutes an easier question — how much does it resemble my mental prototype? — and in doing so neglects base rates, mistakes vivid detail for probability, and sees false patterns. A description that 'sounds like' a category feels likely even when the category is rare.

The corrective is arithmetic: start from the base rate, then adjust. Resemblance is a clue to identity, not a measure of probability, and treating it as one is how stereotypes and false conjunctions mislead.

Example situations:

• Judging a quiet, bookish person more likely a librarian than a salesperson — ignoring how few librarians there are.
• Rating a detailed, plausible scenario as more probable than the simpler event that contains it.

Source: Psychology (Kahneman & Tversky)

A striking anecdote or two feels like sufficient evidence, and the mind readily builds a general rule from a sample far too small to support it. Vividness substitutes for quantity, and one memorable case outweighs statistics it has no right to.

Before generalising, count. Ask whether the cases are numerous and representative enough to bear the weight of the conclusion, or whether a compelling story is masquerading as a pattern.

Example situations:

• Declaring a strategy 'proven' after it worked once or twice.
• Writing off an entire approach because the first attempt happened to fail.

Source: The Great Mental Models (Farnam Street)

Unused abilities fade, and faded abilities disappear from your usable repertoire just when needed. Munger's answer is routine practice of the fundamentals, like a pilot in a simulator, and keeping models in active use through the checklist.

Competence you haven't exercised recently should be trusted less.

Example situations:

• A 'fluent' engineer who hasn't coded in five years estimating a rewrite.
• Emergency procedures nobody has drilled since onboarding.

Source: The Psychology of Human Misjudgment

Alcohol and drugs degrade judgment and — worse — degrade the ability to notice the degradation, often teaming up with denial. Munger lists it among the standard causes of ruin worth avoiding entirely.

Never make or trust significant decisions made under influence, including your own.

Example situations:

• Deal terms agreed over a long boozy dinner.
• A high performer's slow-motion decline everyone politely ignores.

Source: The Psychology of Human Misjudgment

Some skills decay with age while others hold; what preserves function longest is continuous practice and joyful learning. Plan for succession honestly rather than assuming permanence.

Judge current capability by current evidence, not past reputation — in yourself too.

Example situations:

• A founder-CEO with no succession plan at 78.
• Deferring to a legendary expert whose knowledge froze a decade ago.

Source: The Psychology of Human Misjudgment

Milgram's subjects delivered 'lethal' shocks on instruction; copilots have flown planes into the ground rather than contradict the captain. Titles, uniforms, and confidence trigger compliance independent of correctness.

Build dissent into the system: someone must be paid to say the emperor is naked.

Example situations:

• A junior analyst suppressing a fatal flaw the partner missed.
• Medical or legal advice followed without a second opinion because of the letterhead.

Source: The Psychology of Human Misjudgment

The mind craves narrative — cause, effect, hero, arc — and will manufacture a coherent story from random or merely correlated events. Stories are how we remember and persuade, but the same instinct fabricates false causation and makes us more certain than the facts warrant.

A good story feels like understanding even when it isn't. Ask whether the evidence actually supports the causal chain, or whether your narrative instinct has simply supplied the connective tissue.

Example situations:

• A tidy post-hoc explanation for a market move that was mostly randomness.
• A founder's origin story that makes luck look like inevitable strategy.

Source: The Great Mental Models (Farnam Street)

Humans are wired for language, and the categories and metaphors a language provides frame how problems are perceived — you reason with the concepts you have words for. Framing a choice as a 'loss' versus a 'cost', or a person as a 'resource' versus a 'colleague', changes the decision without changing the facts.

Watch the words. Whoever sets the vocabulary of a debate has already shaped its conclusions; reframing in different language often reveals options the first framing hid.

Example situations:

• A layoff framed as 'right-sizing' that anaesthetises the decision it describes.
• A negotiation reshaped by whether the number is called a 'discount' or a 'fee'.

Source: The Great Mental Models (Farnam Street)

Humans produce chatter that sounds like content — and organisations can mistake fluent twaddle for expertise. Munger's bee that dances a nonsense dance still gets followers.

Protect decision forums: weight evidence and track records, not eloquence and airtime.

Example situations:

• The meeting dominated by the most articulate, least informed person.
• Strategy documents full of impressive words and no falsifiable claims.

Source: The Psychology of Human Misjudgment

Adding 'because…' to a request dramatically raises compliance, sometimes even when the reason is empty. Used well: always explain the why, and knowledge sticks and travels. Used badly: hollow reasons manufacture unearned compliance.

Give real reasons with every instruction, and demand real ones behind every claim.

Example situations:

• Orders rolled out with no rationale, quietly sabotaged by a confused team.
• Accepting 'because that's our policy' as if it were an argument.

Source: The Psychology of Human Misjudgment

Franklin's maxim, echoed by Munger: 'If you would persuade, appeal to interest and not to reason.' People are moved far more by what serves their incentives, identity, and feelings than by the logical merit of a case, however airtight. Winning the argument and changing the behaviour are different achievements.

Before trying to convince, map the other side's interest and speak to it. And in reverse: when something is being sold to you with pure reason, check whose interest the conclusion actually serves.

Example situations:

• A technically correct proposal that dies because it ignored what the approver stood to gain or lose.
• Getting adoption of a change by showing each team how it makes their own work easier.

Source: Benjamin Franklin via Poor Charlie's Almanack

Facing uncertainty or a bad situation, people feel a strong urge to do something — to intervene, to demonstrate effort, to relieve the discomfort of passivity — even when the evidence says the best move is patience. Action bias produces needless surgery, over-trading, and meddling that makes things worse.

The corrective is to make inaction a legitimate, explicit option. Ask whether a proposed action beats doing nothing on the merits, or whether it just relieves the itch to act.

Example situations:

• Over-trading a portfolio into worse returns than simply leaving it alone.
• A manager 'fixing' a team that was already recovering on its own, and stalling it.

Source: The Great Mental Models (Farnam Street)

Most harm you experience is error, incompetence, or inattention — not conspiracy. Assuming malice poisons relationships and blinds you to the fixable process problem underneath.

Keep the exception in view: incentives can make carelessness look a lot like strategy.

Example situations:

• A teammate 'ignores' your email that actually landed mid-crisis.
• A vendor ships late because of their supplier, not to squeeze you.

Source: Robert J. Hanlon

Munger's own coinage: the biggest disasters and manias come not from one bias but from several reinforcing each other — social proof plus envy plus overoptimism plus authority makes a bubble; add deprival-superreaction for the crash. Cults and open-outcry auctions are lollapaloozas by design.

When many pressures all push the same direction, your confidence should go down, not up.

Example situations:

• An auction: rivalry, loss-fear, social proof and sunk costs converging on one bid.
• A company-wide conviction no one can trace to evidence.

Source: The Psychology of Human Misjudgment

Because time, money, attention, and materials are limited, every use precludes others and everything worth having must be rationed somehow — by price, queue, or effort. Scarcity is the root of economics and the reason trade-offs are unavoidable rather than a failure of planning.

Manufactured scarcity also distorts behaviour: perceived shortage raises desire independent of real value, which is why 'limited' and 'ending soon' work on us even when we know better.

Example situations:

• A roadmap forced to choose because engineering hours are the true scarce input.
• A 'limited edition' whose scarcity, not its quality, drives the demand.

Source: Economics

Where supply is constrained and demand grows, price (or queues, or quality decay) must rise; where supply floods in, margins compress. Many puzzling situations are just supply-and-demand wearing a costume.

Ask what the scarce resource really is — often it's attention, trust, or talent rather than the obvious good.

Example situations:

• Hiring 'unicorn' engineers in a market where every company wants the same profile.
• A city that restricts building permits and is then surprised by rents.

Source: Economics 101

Every yes is a no to something else. Decisions should be compared against the best available alternative — not against doing nothing, and not in isolation.

Most bad allocations of time and money come from never asking what else the same resource could do.

Example situations:

• Spending six engineer-months on a feature is really spending whatever else those six months could have built.
• Keeping money in a failing project because it 'might turn around' while better uses go unfunded.

Source: Economics 101

There is no free lunch: gaining on one dimension almost always costs you on another — speed versus quality, flexibility versus efficiency, reach versus depth. Pretending a decision has no downside just hides the cost until it arrives uninvited.

Good judgment names the trade-off explicitly and chooses it on purpose. The dangerous decisions are the ones sold as pure upside.

Example situations:

• A 'no-compromise' plan that's really an unexamined compromise on time or cost.
• Choosing speed to market knowing, and accepting, the technical debt it incurs.

Source: Economics

Optimising for a single objective pushes a system toward the extreme of that objective — and whatever wasn't in the objective function gets spent to get there. Maximise short-term profit and you can consume trust, quality, and resilience without noticing until they're gone.

Ask what a metric is quietly costing. The gains from hard optimisation are real; so are the invisible things being converted into those gains.

Example situations:

• A team optimising ticket-closure speed that quietly destroys support quality.
• An algorithm maximising watch time that maximises outrage as a side effect.

Source: Economics

An efficient system wastes nothing — but the 'waste' it removes is often the very slack, redundancy, and buffer that let it absorb shocks. Pushed to the limit, efficiency and resilience trade directly against each other, and a perfectly efficient system has no capacity left to survive surprise.

Deliberately keep some inefficiency where failure is costly. The buffer that looks wasteful in calm conditions is what keeps you standing in a storm.

Example situations:

• A just-in-time supply chain with zero inventory that stops dead at the first disruption.
• A schedule optimised to 100% utilisation that can't absorb a single sick day.

Source: Economics

Adam Smith's pin factory: dividing a job into specialised tasks makes each worker far more productive than any generalist doing the whole thing. Specialisation drives nearly all economic productivity, but it buys efficiency with fragility, because specialists depend on a whole chain of other specialists to function.

The trade is throughput now for resilience later. Highly specialised systems break badly when a link fails or the environment shifts and the narrow skill no longer fits.

Example situations:

• A team so specialised that one person's absence halts the whole pipeline.
• An economy hyper-efficient in normal times and brittle when one supply link snaps.

Source: Economics (Adam Smith)

The flip side of specialisation is a web of dependence — each participant relies on many others for inputs they no longer produce themselves. This interdependence creates enormous collective capability and also shared vulnerability, since a shock to one node propagates through everything connected to it.

Map your dependency chain before trouble finds it. What looks like your risk is often someone else's risk you inherited by depending on them.

Example situations:

• A product outage caused not by you but by a third-party service you rely on.
• A supply chain where one distant factory's closure stops your line.

Source: Economics

What matters is relative, not absolute, efficiency: each party gains by specialising where its opportunity cost is lowest and trading for the rest. This is why delegation pays even when you'd do the task better yourself.

The question is never 'am I better at this?' but 'is this the best use of my time?'

Example situations:

• A founder who still does the bookkeeping because they're faster than the bookkeeper.
• Two teams splitting frontend and infrastructure instead of both doing both, badly.

Source: David Ricardo

Adding more of an input — effort, people, money, features — raises output at a shrinking rate once the easy gains are captured, and eventually the marginal return approaches zero or turns negative. The first hour, the first hire, the first feature does far more than the tenth.

Know where you are on the curve. Early on, push hard; near the flat, the same input is better redeployed somewhere still steep. Ignoring this is how projects gold-plate.

Example situations:

• The tenth reviewer on a document catching nothing the first three didn't.
• Piling engineers onto a late project and watching it get later (Brooks's law).

Source: Economics

Scale spreads fixed costs, improves purchasing power, and feeds experience curves; past a point it also buys distribution, brand, and data advantages smaller rivals cannot match. Munger: scale advantages are among the most durable moats.

But scale adds bureaucracy — the diseconomies arrive as surely as the economies.

Example situations:

• A rival can price below your cost because their volume is 10x yours.
• Consolidating suppliers to hit discount tiers.

Source: Economics 101 / Poor Charlie's Almanack

Debt lets you act now with resources you haven't yet earned, accelerating growth when things go well. But leverage is symmetric: it multiplies losses exactly as it multiplies gains, and it removes your flexibility precisely when you most need it, because obligations are fixed while income is not.

The danger isn't debt itself but the loss of optionality: a leveraged position can be forced to act at the worst possible moment. Financial, technical, and time debt all behave this way.

Example situations:

• Technical debt that speeds this quarter and taxes every quarter after.
• A leveraged bet that would have paid off if it hadn't been margin-called first.

Source: Economics

Schumpeter's engine of capitalism: growth comes not from doing the old thing better but from new methods and products that wipe out the old ones entirely. The process is generative and brutal at once — the same wave that creates enormous value destroys the businesses, skills, and jobs built on the previous order.

Assume your current advantage is somebody's demolition target. The relevant question isn't whether disruption comes but whether you're on the creating or the destroyed side of it.

Example situations:

• A profitable product line quietly obsoleted by a cheaper, worse-but-improving newcomer.
• A skill that was a career moat until a new tool made it a commodity.

Source: Economics (Joseph Schumpeter)

In textbook competition, rivals bid margins down until no one earns more than the cost of capital — a great deal for customers and a grind for producers. Real profit accrues to those who escape competition: through a monopoly position, differentiation, or a niche where the usual rivalry doesn't reach.

So when a business is unusually profitable, ask what shields it from competition — and when yours isn't, ask whether you're trapped in an undifferentiated fight to the bottom.

Example situations:

• A commodity market where everyone works hard and no one makes money.
• A niche leader earning fat margins precisely because no one else bothers to compete.

Source: Economics

Capitalism ensures good returns are competed away unless something durable protects them: brand, network effects, switching costs, scale, patents, regulation. A business plan without a moat is a plan to earn average returns at best.

Ask not 'is this good now?' but 'what stops others from copying it?'

Example situations:

• A hot product whose only edge is being first — with six clones shipping.
• Choosing between two firms: one with locked-in customers, one winning purely on price.

Source: Warren Buffett / Poor Charlie's Almanack

Originally about money — debased coins circulate while good coins get hoarded — Gresham's law generalises: when a low-quality version is allowed to pass as equivalent to a high-quality one, the bad crowds out the good. Anywhere quality is hard to distinguish and cheating goes unpunished, standards decay toward the worst tolerated.

The defence is to make quality legible and to stop treating unequal things as equal. Otherwise the honest and the excellent quietly exit the arena.

Example situations:

• A marketplace where fake reviews are tolerated until honest sellers can't compete.
• A hiring process where confident bluffing reads the same as real competence.

Source: Economics (Thomas Gresham)

Churn is the rate at which you lose what you've gained — customers, employees, users, trust — and because it compounds against you, a leaky bucket can neutralise heavy acquisition effort while the top-line still looks busy. Growth is net of churn, and net is what compounds.

Before spending to add more, ask why what you have is leaving. Reducing churn is usually cheaper and more durable than out-running it with new inflow.

Example situations:

• Pouring budget into acquisition while a high cancellation rate drains the tank behind it.
• A team that's always hiring because it's always losing people to the same fixable cause.

Source: Economics / systems

A bubble forms when rising prices attract buyers whose buying pushes prices higher still, in a reinforcing loop powered by social proof, envy, and the story that 'this time is different'. Detached from underlying value, it inflates until the marginal believer runs out — then reverses just as violently.

Bubbles are lollapaloozas of psychology and money. The tell is not high prices but a narrative that explains away every reason for caution, and a crowd whose confidence rises as its footing weakens.

Example situations:

• An asset everyone buys because it keeps going up, for no reason but that it keeps going up.
• A hiring or spending frenzy in a hot sector that unwinds the moment sentiment turns.

Source: Economics

You don't have to be an expert in everything; you have to know precisely where the edge of your understanding lies and be honest when a decision falls outside it. The size of the circle matters far less than knowing its boundary.

Outside the circle, either don't play, or explicitly treat yourself as a beginner: get help, reduce stakes, and expect to be wrong.

Example situations:

• Being offered an investment in a hot sector you can't explain in plain words.
• A doctor asked for advice about a specialty they haven't practised in twenty years.

Source: Poor Charlie's Almanack

When several explanations account for the same evidence, the one with the fewest assumptions is most likely right and easiest to test. Complexity must earn its place.

It is a tie-breaker, not a law: if the simple explanation keeps failing to predict, add complexity reluctantly.

Example situations:

• Server is down: check the cable and the deploy before suspecting a kernel bug.
• Sales dropped the week a competitor launched — and also the week your checkout broke.

Source: William of Ockham

First-order thinking stops at the immediate effect; second-order thinking asks what happens next, who reacts, and how the system adapts. Most competition and most policy failures live at the second order.

A cheap test for any plan: write down the first-order effect, then force yourself to write the reaction to it.

Example situations:

• Cutting prices wins customers (first order) until competitors match and the whole market earns less (second order).
• Subsidising a shortage increases demand for the scarce thing.

Source: Howard Marks / general systems thinking

Reinforcing loops amplify (compounding, virality, panics); balancing loops stabilise (thermostats, prices, satiation). Most surprising system behaviour is a loop you haven't mapped, often with a delay that hides the connection.

Ask of any plan: what does this change feed back into, and with what lag?

Example situations:

• Hiring freezes → overwork → attrition → more overwork.
• Price cuts fund growth that lowers unit costs, enabling further cuts — until a competitor's loop runs faster.

Source: Systems thinking (Forrester/Meadows)

Equilibrium is the state a system drifts to when its opposing forces balance — and apparent stillness usually hides forces in active tension, not their absence. Push on one side and the system moves to a new balance point, frequently in a direction that partly cancels your push.

Ask what forces are actually holding a situation in place. Durable change means altering the balance of forces, not just the visible outcome, which snaps back the moment you stop pushing.

Example situations:

• A market price that looks stable but is a live standoff between supply and demand.
• A 'settled' team dynamic held in place by pressures no one has named.

Source: Systems thinking

Every system has a limiting step — the bottleneck — and total throughput is governed by it alone. Effort spent speeding up non-bottleneck stages produces no improvement in the whole and often just piles up inventory in front of the real constraint.

Find the constraint, exploit it fully, and subordinate everything else to it (Goldratt's theory of constraints). When you relieve one bottleneck, the constraint moves — so locate the new one before optimising blindly.

Example situations:

• Adding developers when the true limit is a single overloaded reviewer.
• Upgrading every step of a pipeline except the one slow stage that gates output.

Source: Theory of Constraints (Goldratt)

Quantity has a quality all its own: a thing that works at one size can break or behave entirely differently at ten or a thousand times that size. Communication, coordination, and structure that suit a small system fail at a large one, and vice versa — what scales is rarely what worked when small.

Before assuming a success will grow, ask what breaks at 10x. Many strategies are size-specific and stop working precisely because they got bigger.

Example situations:

• A hands-on management style that works at ten people and collapses at a hundred.
• An architecture that's elegant at low load and falls over when traffic multiplies.

Source: Systems thinking

An algorithm is a defined sequence of steps that reliably produces a result — and for recurring problems, a well-designed procedure outperforms fresh judgment each time, because it captures hard-won lessons and removes the noise of mood, memory, and fatigue. Checklists and rules are algorithms for human decisions.

For anything you do repeatedly, the leverage is in improving the procedure once rather than re-deciding each instance. Automate the routine so attention is free for the genuinely novel.

Example situations:

• Replacing ad-hoc triage with a checklist that catches what tired humans miss.
• A hiring rubric that beats gut-feel interviews by scoring the same things every time.

Source: Computer science

Interactions among simple parts can produce complex behaviours that exist only at the level of the whole — wetness isn't in a water molecule, traffic jams aren't in any car, culture isn't in any employee. Emergent properties can't be found by studying components in isolation.

So some problems are unreachable from the parts: you have to observe and intervene at the system level. Reductionism that ignores emergence explains away the very phenomenon you care about.

Example situations:

• A toxic culture that no single person embodies but the whole clearly has.
• Market behaviour that emerges from traders and lives in none of them individually.

Source: Complexity science

Every system has irreducible elements: a minimum number of steps, a floor on time, an essential complexity that can be moved around but never removed. Trying to compress past the floor doesn't eliminate the requirement; it just relocates it, usually into failure or hidden cost.

Distinguish the genuinely irreducible from the merely traditional. You can strip away the accidental, but demanding that an irreducible minimum be beaten is how projects break — nine women can't make a baby in a month.

Example situations:

• Compressing a schedule below the time the work irreducibly takes — and shipping broken.
• A 'simplification' that just pushes the essential complexity onto the user instead.

Source: Systems thinking

Military commanders who visit the front see what reports and maps leave out; distance and hierarchy filter and distort information, so decisions made only from the rear are made half-blind. The general who walks the line, like the executive who talks to customers directly, corrects for the distortion of secondhand accounts.

Before a consequential decision, go and look. The map, the dashboard, and the summary have all been simplified by someone; the front has not.

Example situations:

• A leader who reads only status reports and misses what every frontline worker knows.
• Fixing a product by watching real users struggle instead of reading the metrics deck.

Source: Military strategy

When one side is far stronger by conventional measures, the weaker side loses any symmetric fight — so it changes the rules, using tactics (guerrilla action, speed, unconventional moves) that neutralise the giant's advantages instead of confronting them. Strength in one game is irrelevant if you're made to play another.

For an underdog, the question is never 'how do we win their fight?' but 'what fight can we make them play where their size is a liability?'

Example situations:

• A startup beating an incumbent by competing where the incumbent's scale is a handicap.
• A small team winning on speed and focus against a large, slow, well-resourced rival.

Source: Military strategy

A power forced to fight in two directions divides its resources and attention, and is often beaten in detail — defeated on each front by an enemy it could have handled alone. History is littered with campaigns lost to the two-front trap.

The general lesson is to avoid opening a second front while the first is unresolved: concentrate force, finish one fight before starting another, and beware being maneuvered into simultaneous battles.

Example situations:

• A company fighting a price war and a costly internal reorg at the same time, losing both.
• Taking on two hard strategic bets at once and starving each of the focus it needed.

Source: Military strategy

Conventional strength struggles against an insurgency with no front line and deep popular roots — brute force often creates more resistance than it removes. Winning requires addressing the conditions that sustain the opposition (legitimacy, grievances, support), not just attacking its visible fighters.

The transferable idea: against a decentralised, self-replenishing problem, force applied to symptoms backfires. You have to change the environment that keeps regenerating it.

Example situations:

• Cracking down on a workaround that just breeds new workarounds until the cause is fixed.
• Fighting a grassroots backlash with pressure that only deepens the grievance feeding it.

Source: Military strategy

If any attack triggers a counterstrike devastating enough to destroy the attacker too, rational parties are deterred from attacking at all — the very certainty of catastrophe produces stability. Paradoxically, strength that guarantees mutual ruin can be more peace-keeping than strength that promises one-sided victory.

The same structure appears in business and negotiation: credible, symmetric retaliation deters aggression, while a first strike that invites proportionate reprisal makes everyone worse off.

Example situations:

• Two rivals who avoid a price war because each knows it would ruin them both.
• A standoff held stable by the shared certainty that escalation destroys everyone.

Source: Military strategy / game theory