Concepts
The same idea shows up in a book, a letter and a lecture. Each concept here is one record, with every source that carries it and the L3VLUP page where it is taught.
Incentives
Who is paid for what explains behaviour better than any stated intention. Read the incentives first.
7 ideas
How an activist builds a case
Relative performance, a worked valuation of what remains, a cost or margin bridge, and a governance failure to pin it on.
5 ideas
Capital allocation
What a company does with the cash it makes: reinvest, acquire, buy back, pay out or hold. The decision that decides what the company becomes.
4 ideas
Decision quality against outcome quality
A good decision can have a bad outcome and a bad decision a good one. Judge the decision on what was knowable when it was made.
4 ideas
Duration and covenants
The terms of a loan, when it comes due and what it allows, matter as much as how much was borrowed.
4 ideas
Long-term orientation
Choosing cash flows years out over results this quarter, and telling investors so in advance.
4 ideas
Price against value
The market quotes a price every day. Value is a separate question, answered by the business rather than the quote.
4 ideas
Quality of earnings
Whether reported profit is backed by cash, and which adjustments were built to flatter.
4 ideas
Buybacks and intrinsic value
A repurchase creates value only below a conservative estimate of intrinsic value, and destroys it above.
3 ideas
Liquidity against solvency
An illiquid firm cannot pay today; an insolvent one cannot pay ever. They need different rescues.
3 ideas
Margin of safety
Paying enough less than a conservative estimate of value that being somewhat wrong still leaves a return.
3 ideas
Network effects
A product that gets more useful as more people use it. They start in one small market and rarely across a whole one.
3 ideas
Proxies against goals
A process, a survey score or a metric starts as a stand-in for the goal and quietly becomes the goal.
3 ideas
Risk control as the common factor
Traders who agree on nothing else agree about limiting losses.
3 ideas
Survivorship
One person’s record is a story about the people who did not make it until proven otherwise.
3 ideas
Task against purpose
A job is a purpose made of tasks. Automating the tasks changes the job without removing the purpose.
3 ideas
Term premium
The extra yield for holding a long bond over rolling short ones. A yield is expected short rates plus this.
3 ideas
The correlation assumption
A structured product’s rating rests on how its loans are assumed to default together. When that assumption fails, the rating does.
3 ideas
Tokens
A model never sees words, only tokens, which is why it miscounts letters, why context has a size and why usage is billed as it is.
3 ideas
Accretion is not value
Whether a deal adds to earnings per share depends on how it was paid for. Whether it was worth doing is a separate question.
2 ideas
Accrual accounting
Revenue is recorded when earned and cost when incurred, so profit and cash diverge by design.
2 ideas
Adjusted metrics
An adjusted metric is an argument about what should not count. Ask who benefits from the exclusion.
2 ideas
Base rates and anchoring
The two biases that cost money: ignoring how often something usually happens, and leaning on the first number seen.
2 ideas
Cheap measurable signals
A cheap, measurable signal beats an expensive intuitive one at scale, in scouting, in quant investing and in hiring.
2 ideas
Chokepoints
A supply chain with a single irreplaceable step is a single point of failure for everything built on it.
2 ideas
Cost of capital
The blended return every provider of capital requires, used to discount cash flows that belong to all of them.
2 ideas
Deposit beta
How much of a change in the policy rate a bank passes on to depositors. Low betas are where net interest margin comes from.
2 ideas
Frictional costs
Fees, spreads and trading reduce what investors as a group can earn from the businesses they own.
2 ideas
Intrinsic value
The value of any asset is the present value of the cash it will produce. Everything else in a valuation is machinery for that sentence.
2 ideas
Maturity mismatch
Borrowing short and lending long. A solvent bank can still fail if enough depositors ask for their money at once.
2 ideas
Models built on calm history
A model fitted to years of moderate outcomes cannot see the one that is not moderate.
2 ideas
Physical commodity trading
Physical traders make money on logistics and access, not on a price view.
2 ideas
Platforms
A business that sets the rules for a transaction it does not itself perform, and decides which side pays.
2 ideas
Proprietary data as the moat
In an industry, the firms with deep data of their own capture more of a technology shift than the makers of the tools.
2 ideas
Return on reinvestment
Earnings growth is return on capital times the share reinvested. Spending is only growth if the return is there.
2 ideas
Secrecy as a warning sign
Secrecy presented as protecting the technology is the signal that there may be no technology.
2 ideas
Short selling and fieldwork
A short seller checks the world against the filings because the filings are what is in doubt.
2 ideas
Short squeezes and settlement
A squeeze is a funding problem before it is a price problem, and the collateral call lands before the trade settles.
2 ideas
Skill and market structure
A skill is only worth what the market structure around it allows.
2 ideas
Specification gaming
A system optimises the objective you wrote, not the one you meant.
2 ideas
Stating what would prove you wrong
A thesis is testable only when it names the evidence that would refute it, in advance.
2 ideas
The anatomy of a bubble
Displacement, boom, euphoria, distress, revulsion. The same sequence across four centuries, needing marketable assets, money and speculation together.
2 ideas
The conditions for compounding
Compounding needs an uninterrupted run more than it needs a great year. The conditions matter more than the picks.
2 ideas
The context window
Everything the model knows at answer time is in the window, and attention across it degrades before the window fills.
2 ideas
The cost of being public
Quarterly reporting, guidance and the market’s scorecard shape decisions. Founders and activists both argue about whether that is a price worth paying.
2 ideas
Working capital
Receivables, inventory and payables. Growth ties up cash in them even when the income statement looks healthy.
2 ideas
A prompt is a brief
Task, role, background and output format. Most disappointing answers are underspecified questions.
1 idea
Ask about the past, not the future
Ask what people already did, never what they would do.
1 idea
Bias and variance
Too simple a model misses the signal; too flexible a model memorises the noise. Diagnose which from train and validation error.
1 idea
Bias reproduced at scale
A model trained on a biased outcome reproduces the bias at scale and with a straight face.
1 idea
Breadth before specialisation
In unpredictable fields, breadth and late specialisation win more often than an early narrow path.
1 idea
Capital, not asset purchases
Buying a bank’s bad assets helps only if you overpay; giving it capital helps directly.
1 idea
Competition against profit
Competition and profit rarely occupy the same market.
1 idea
Complexity as cost of change
Complexity is what you must hold in your head to make a safe change. Deep modules with shallow interfaces reduce it.
1 idea
Consensus
Getting several machines to agree on one value when any may crash and the network may lie about which.
1 idea
Database indexes
A sorted structure that turns a full scan into a handful of page reads, paid for in write speed and disk.
1 idea
Depth over breadth in applications
Effort and results are not linear. Depth on a few firms beats applications to many.
1 idea
Diagnosing a metric
North star against guardrail metrics, and the structured hunt for what made a number fall.
1 idea
Evaluation
Without a test set and a metric you have a demo, not a system.
1 idea
Financial discipline has a limit
Cost programmes and buybacks run without limit against an engineering culture eventually eat the engineering, with a lag.
1 idea
Finding the subproblem
The hard part of dynamic programming is finding the subproblem, not the memoisation.
1 idea
How inflation comes down
Supply shocks fade on their own; policy decides whether expectations and demand let them.
1 idea
How much debt is right
Debt is cheap until it is not. The right amount minimises the cost of capital rather than maximising what a lender will give.
1 idea
Idempotency
In a distributed system every request may arrive twice. Design so the second arrival changes nothing.
1 idea
Inversion
Ask what would make the plan fail, and avoid that, before asking what would make it succeed.
1 idea
Management owns the outcome
Responsibility for a company’s actions sits with its management, not its regulators.
1 idea
Margin arithmetic
Headcount, cost per head and incremental margin on growth, built into a target you can check against guidance.
1 idea
Narrative to numbers
Every forecast row is a sentence about the business. A story that cannot become a number is not a thesis.
1 idea
Negative convexity
When rates fall, mortgage borrowers refinance, so the bond shortens exactly when a longer bond would have gained most.
1 idea
Networking as generosity
Be useful before you are needy. It is the only version that works on someone who owes you nothing.
1 idea
Outputs against outcomes
Shipping is an output. Nobody pays for outputs.
1 idea
Precision against recall
Which error you can afford is a business decision wearing a statistical costume.
1 idea
Private credit
Loans made by funds rather than banks or markets: larger, less liquid, with thinner coverage and lower recoveries.
1 idea
Reversible and irreversible decisions
A decision that can be undone deserves a light process; one that cannot deserves a heavy one. Treating them alike slows everything.
1 idea
Risk against uncertainty
Low risk and high uncertainty are different things, and the gap between them is where returns come from.
1 idea
Risk as loss of purchasing power
Risk is the chance of losing buying power over the holding period, not the volatility of the price along the way.
1 idea
Segregation of duties
One person holding both a trade and its settlement is the whole control failure.
1 idea
Signals are small and short-lived
A signal that works is small, decays, and is expensive to find.
1 idea
Speed and stability together
The teams that ship most often also fail least. The trade-off between the two is a myth the data does not support.
1 idea
Stock-based compensation
An operating expense that costs no cash and dilutes you anyway.
1 idea
Sum of the parts
Valuing each business a company owns separately, then asking why the whole trades for less.
1 idea
Synergies
The extra value two businesses are claimed to create together. Worth something only once valued rather than asserted.
1 idea
Terminal value
The value of the cash flows beyond the forecast, usually most of a DCF and most of its risk.
1 idea
The acquired brand’s market share
The cleanest test of an acquisition is what happened to the acquired brand’s share of its market afterwards.
1 idea
The cold start problem
A network begins in one atomic market, campus by campus or city by city, and most die before they tip.
1 idea
The cost of protection
A hedge is insurance with a premium. Over a long bull market the premium is the whole story.
1 idea
The least bad option
There is often no good option, only the least bad one taken quickly.
1 idea
The speed of a change
A right decision taken too fast reads as a wrong one. The mistake is often the pace, not the direction.
1 idea
The warning in the footnote
The disclosure that flagged a failure was usually there years early, in a footnote.
1 idea
Two-part intrinsic value
Investments less debt, plus normalised earnings times a conservative multiple, per share.
1 idea
Uninterrupted blocks
Preparation compounds in long blocks of focus and not otherwise.
1 idea
Weighting opinions by record
Weight an opinion by the record of the person holding it, not by how loudly it is held.
1 idea
What the curve signals
An inverted curve says markets expect rate cuts. The near-term part of the curve carries the signal, not the long end.
1 idea
Winners sorted faster
Markets now separate winners from losers faster, and profits concentrate among fewer companies.
1 idea