Agustín LeBron is a former chip designer, quantitative trader at Jane Street, and now founder of a crypto protocol startup. He wrote The Laws of Trading, a book about decision-making under uncertainty that uses trading as a lens. This conversation covers adverse selection, hiring, the economics of finance, technical debt, crypto, and career strategy.
Adverse Selection
Adverse selection is everywhere in hiring. When a company posts a job, the applicants are already a negatively selected pool—top performers are usually retained by their current employers and rarely on the market. The employee also has the final say in accepting an offer, so strong candidates pick from many offers while weaker ones pick from few, systematically disadvantaging employers.
Finance may have more adverse selection than other industries. In fields like aerospace, people have intrinsic motivations (e.g., wanting to go to space). In finance, when someone says “I’ve wanted to be a trader my whole life,” they often really mean “I’ve wanted to make a lot of money my whole life”—a red flag because it signals they don’t understand the actual job.
When you’re on the adverse-selected side (e.g., a young male driver with high insurance premiums), you can sometimes counteract it with monitoring devices (GPS/accelerometer tracking for driving). In employment, the best defense is carefully evaluating your interviewers and the company’s culture—if the interview tasks correlate to the actual job, that’s a strong signal. Don’t let the high-status name of a company override your own assessment of fit.
Brain teaser interview questions are often IQ tests in disguise. Explicit IQ testing is illegal in US employment, but companies route around this by using tests like Wonderlic or brain teasers that correlate with general intelligence. These are often poorly administered, but the underlying logic is that g strongly predicts job performance across industries.
Hiring and Talent
Hire for abilities, not skills. The most common failure mode in tech hiring is hiring for legible skills (e.g., “can you write Python?”) rather than underlying ability and potential. Skills are easy to evaluate but poor predictors of long-term value.
Domain expertise in trading is less important than people think. Retail trading (e.g., Robinhood) is a completely different domain from professional market making. Prior retail trading experience can actually be a negative—it introduces bad habits that must be unlearned, and retail traders are likely losing money on average (adverse selection).
Look for people who enjoy the game for its own sake. A strong signal is someone who is the third-best player in the world at some obscure chess variant—this correlates with willingness to grind and improve at something for intrinsic enjoyment, not just extrinsic reward.
There is a massive arbitrage opportunity in global talent. The smartest 0.1% of high school graduates in countries like India and Nigeria are massively underserved. A bootcamp model that screens for intelligence and grit (perhaps gamified, à la B.F. Skinner’s operant conditioning) and trains them for six months could place them in six-figure Western tech jobs. Companies like Infosys capture most of this arbitrage today.
Hire more junior people and invest in training. Salaries roughly double between year 0 and year 3 because companies refuse to invest in training and instead compete for people who already have two years of experience. This is a collective action problem—if you train people well and give them good work, they tend to stay because switching jobs is costly and risky.
Startups should use a barbell hiring strategy. Hire one or two truly exceptional (90th percentile) people, then fill the rest of the team with solid but not elite (40th percentile) people in situations where they can be effective. Trying to hire all “A-players” as a startup fails because the people you want will go to higher-status, higher-paying companies.
Finance and Trading
Finance at 9% of GDP is hard to evaluate. From the inside, much of it looks like zero-sum competition that should be more efficient. From the outside, it’s hard to argue with GDP growth. The people best positioned to judge (insiders) are biased toward legitimacy by familiarity. At the margin, LeBron would ban leveraged ETFs for retail traders and simplify baroque banking regulations (e.g., Basel III) that feel like job Ponzis.
Trading is labor-intensive. The value is in the seat, not the capital. Small inefficiencies (e.g., pink sheets) will always exist because they’re not worth a large firm’s time—they get competed away by smaller operations or independent traders.
AI will gradually automate trading, but large language models specifically may not be the path. GPT-style models are sample-inefficient compared to human learning and lack true semantic understanding (they’re “clever Hans” systems). The analogy is 20th-century manufacturing: human cognitive load will slowly shift to machines, but the architecture for genuine understanding hasn’t been built yet.
A trading strategy that predicts correctly 55% of the time can still lose money in production because of adverse selection—you’ll execute all the bad trades and only a fraction of the good ones—and because successful strategies signal to competitors who arbitrage them away.
Jane Street’s success doesn’t create a negative feedback loop because they provide a genuine service: liquidity. They trade against pension funds and hedge funds that need to make macro bets in foreign markets, not against other market makers. There are real gains from trade.
Jane Street’s culture attracts the rationality/EA community because of its emphasis on intellectual honesty, “shut up and multiply” thinking, and collegiality. The overlap is also functional: trading requires modeling other agents and their incentives, which is a more complex problem than debugging static systems.
Technical Debt
Technical debt should be understood in financial terms. Startups should take on more technical debt because their implied interest rate is high—if the company fails, the debt is non-recourse; if it succeeds, they rewrite from scratch. The MVP model is exactly this.
Large companies spend most of their engineering time servicing accumulated technical debt—upgrading deprecated libraries, migrating platforms, maintaining decades-old code. Moving to a new platform is like refinancing debt.
Software development is fundamentally sociology. The hardest part is organizing teams, creating processes, managing culture, and building conventions around complexity. The technical problems are secondary to the human ones.
Jane Street uses OCaml because its strong static type system makes impossible states unrepresentable, minimizing the risk of code that incinerates money. In finance, correctness matters more than speed of iteration.
Crypto
LeBron is building a crypto protocol with novel cryptographic guarantees against things traders dislike (details kept vague to preserve edge).
He’s skeptical of automated market makers (AMMs) like Uniswap. At least half of Uniswap V2 liquidity providers lose money. The fees are set by fiat (e.g., 0.3%) and aren’t adaptive to market conditions. Outsized returns should come from knowing something others don’t, not from passively throwing money into a pool. This contrasts with passive stock investing, where you’re providing real risk capital to companies.
Crypto’s value as a Schelling point for experimentation. Even if crypto goes to zero, it has already coordinated a wealth transfer from “olds” (VC LPs, institutional investors) to young builders willing to try bold new things. That’s a net positive.
Success for crypto looks like integration with traditional finance, not replacement. Goldman Sachs and JPMorgan absorb the best crypto ideas; crypto companies become incredibly successful within a hybrid system. The “laser eyes” narrative (crypto replaces everything) is wrong.
In a successful crypto future: Western Union and airport currency exchanges are gone; NFTs handle credentialing (transcripts, identity) without intermediaries; the plumbing of finance is “crypto-fied” (not everything on-chain, but blockchain’s useful properties are leveraged where appropriate); and entrenched financial oligopolies face competition (e.g., the CFTC review of FTX’s futures margining proposal).
Career Advice
Life is long, not short. You have many opportunities to learn and try different things. LeBron’s approach is “sequential excellence”—spend ~6 years getting deeply excellent at one thing, then pivot to something different. This works because deep expertise is undervalued, and cross-domain transfer is surprisingly high.
Engineering training was directly useful in trading, which was directly useful in consulting, which is directly useful in his startup. Each career built on the last in non-obvious ways.
The average startup founder is older than popular culture suggests. Young founders get the press, but many successful founders have broad experience across multiple domains.
For someone in their early 20s: Don’t rush. Spend five years becoming world-class at something, then pivot if you want. You have time. The key is self-awareness about whether you’re an “evolutionary” discoverer (pushing boundaries in an existing field) or a “revolutionary” one (inventing a new field, like Shannon or Einstein)—the latter benefits from broad exposure to many domains.
Book Recommendations
A Deepness in the Sky by Vernor Vinge—a science fiction novel that is secretly about software engineering and managing complexity over millennia.
Red-Blooded Risk by Aaron Brown—about risk-taking in general; all of Brown’s books are excellent. The Poker Face of Wall Street is what got LeBron interested in finance.
Kolyma Stories by Varlam Shalamov—short stories about life in Soviet gulags; possibly the most revealing book about human nature LeBron has ever read (but very depressing).
