What's this about?
A non-AI thread from the chain-of-thought author. His proposal: sports are powerful learning environments, and playing one for thousands of hours conditions your neural network to respond to the incentives that sport rewards — a "cognitive reward shape" that transfers to how you think professionally. The funny part, he notes, is that most people choose their sport for accidental reasons — parents, geography. His example is tennis: hundreds of points per match where a spectacular winner is worth exactly one point, the same as an opponent's unforced error. So you win by minimizing mistakes, playing the percentages, grinding point by point, alone. That maps onto surgery or piloting, where the best appendectomy and the smoothest flight earn no bonus, and the job is consistency with minimal tolerance for error. It maps badly onto starting a company — high variance, team-based, tolerant of failure, occasionally paying exponentially for one creative swing.
What exactly is a "reward shape"?
It's the distribution of payoffs an activity offers. A flat, dense reward shape gives many small equal payoffs — every tennis point is worth one, so consistency wins. A heavy-tailed shape gives mostly nothing with rare enormous payoffs — venture investing, research, startups — where a single hit outweighs a hundred failures, so the correct strategy is taking many uncorrelated shots and surviving the misses. The strategies these demand are genuinely opposed: minimizing variance is right in the first and actively wrong in the second.
Why is this worth noticing?
- It's a legible way to talk about transferable dispositions rather than skills, and it applies well beyond sport — the reward shape of your first job probably conditions you the same way.
- The accidental-choice observation is the sharp part. If the sport you happened to play shapes how you approach professional risk, an important input to your thinking was chosen by your parents.
- It's a change of pace worth noticing, from someone who spent this week arguing about scaling and tool use, and it travelled on its own merits rather than the AI angle.