Nobody can be expert in everything. This is obvious, but the implications are underrated. If you can’t reason from first principles in most domains — and you can’t — then the question becomes: what do you do instead?
The answer, I think, is that you build a library of heuristics. Simple rules that trade away per-case optimisation for something more valuable: not having to think.
Take a mundane example. I had sinusitis last week. Four medications. By day two, the symptoms were mostly gone. So: do I stop the ones I no longer need? I could research each drug, weigh the marginal benefit against the mild side effects, make a reasoned pharmacological judgment. Or I could follow a simple rule: finish the course. The doctor prescribed it, the scope is bounded, the stakes are low. The rule slightly underperforms perfect case-by-case reasoning on any single decision, but it costs nothing to execute.
Now take honesty. Not as a moral stance — as a strategy. Lying has overhead. You have to remember what you told whom. You have to maintain consistency across audiences. You manage the low-grade anxiety of stories intersecting. Every lie is a small piece of state you’re carrying. Honesty collapses all of that. One version of events, no bookkeeping. Short-term, you leave things on the table — the diplomatic dodge, the convenient excuse, the version of the story that makes you look better. Long-term, you save an enormous amount of energy and build a reputation that compounds. I don’t think you need to want to be a good person for this to work. You just need to be lazy enough to prefer the cheaper strategy.
Or take delegation. I use AI coding agents and I don’t read every line they write. I write tests first — define what correct looks like — and let the agent fill in the implementation. If the tests pass, the code is acceptable. People find this uncomfortable. But it’s the same move as trusting the doctor’s prescription. Trusted source, bounded scope, verification at the edges. The heuristic is identical. Only the domain changed.
These three things — finish the course, default to honesty, delegate and verify at the boundaries — look unrelated on the surface. One is medical compliance, one is a life philosophy, one is a software practice. But they’re the same cognitive move: adopt a simple rule so you don’t have to reason about every instance.
The interesting thing is what this suggests about expertise. We tend to think experts are people who reason more carefully. They consider more variables, weigh more trade-offs, think more deeply about each decision. But I don’t think that’s right. Experts reason more carefully about fewer things, because they’ve already resolved most decisions into heuristics. A senior engineer doesn’t agonise over variable naming conventions — they have a rule, they follow it, they spend their reasoning budget on architecture. A good doctor doesn’t rethink the antibiotic course for a routine sinus infection — they prescribe the standard protocol and save their judgment for the ambiguous cases.
The meta-skill isn’t making good decisions. It’s building a library of heuristics that remove the need for most decisions, so that your actual judgment is reserved for the cases where it matters. Every simple rule you adopt — finish the course, be honest, delegate and verify, don’t optimise what you can automate — frees up a slot for genuine reasoning somewhere else.
The failure mode is easy to spot: it’s the person who reasons from first principles about everything. Every medication decision is a research project. Every social interaction is a strategic calculation. Every line of code is reviewed as though it’s novel. This looks rigorous. It’s actually expensive and fragile — you’ll exhaust your reasoning budget long before you reach the decisions that actually need it.
The library metaphor is deliberate. You don’t build a library by writing every book yourself. You curate — pick up heuristics from doctors, from engineers, from people who’ve been burned by the thing you haven’t tried yet. The quality of your library determines the quality of your defaults, and the quality of your defaults determines how much judgment you have left for the hard problems.
So the question isn’t whether you’re smart enough to reason about everything. Nobody is. The question is whether your heuristic library is good enough that you rarely have to.