David Epstein

1 card

1. 01

   Preparing for AI is not a list of tools to learn but four habits trained daily: think before asking the tool, check what you get back, know how to judge quality, and have a field of your own deep enough to judge with.

   Range: Why Generalists Triumph in a Specialized World · 2019

   First write what you think in three lines, then open the model.

   Epstein argues that early specialisation only pays in worlds with stable rules. The world a child of today will work in is not like that, so preparation looks different from "learn to code at eight". The four habits with their exercise. One, your own thinking before the tool: the house rule is that you first write what you think in three lines, and only then open the model. Without that step the tool replaces your thinking instead of extending it, and the difference shows over years. Two, checking: any figure, quotation or claim that comes back is confronted with an independent source. A model produces plausible text, not true text, and the habit of checking is the only defence — the same one that protects against disinformation. Three, taste: being able to say why one version is better than another. Taste is trained by volume and comparison — read a lot, listen a lot, look a lot, plus the question "which of the two and why". Four, depth in a field they chose: you cannot judge an answer in a field where you are nothing. Here Epstein does not contradict specialisation, he only moves its moment: breadth first, then depth chosen knowingly. What you do concretely: one project of their own carried to the end every year, at any age; the obligation to explain out loud what they did; and at least one field where they are better than you.

   “In environments where the rules are stable and feedback is immediate, early specialisation wins. Where the rules change, the winner is whoever tried several things before choosing.”

   Open the card →