The shape of judgment in technical hiring
The founders I know who hire well share an unglamorous habit: they pay attention to small things in the work. Not to credentials, not to confidence in interviews, but to the texture of how someone explains a decision they already made.
In the early years of a company, every hire is load-bearing. The wrong third engineer doesn’t just slow you down — they bend the shape of the team. They shift what gets argued about in code review, which problems feel hard, which problems feel beneath the team’s attention. You can recover from a bad hire at fifty engineers; at five, you re-found the company.
The interview is the wrong instrument
Most interviews measure performance in interviews. They reward fluency, composure, and the ability to recover from being put on the spot. None of these are useless. None of them are the thing.
The thing is closer to this: when this person was last surprised by something they built, what did they do next?
When this person was last surprised by something they built, what did they do next?
What I look for
A few rough heuristics, none of them sufficient on their own:
- They tell stories about being wrong. Specifically wrong. With dates, names of systems, the shape of the bug. Vague self-deprecation is its opposite — a learned move.
- They argue with their own examples. Mid-sentence, they catch the counter-case. This is what taste sounds like under pressure.
- They are bored by easy questions in their own field, and curious about easy questions outside it. The combination is rare and almost always means a working mind.
None of this is provable in a forty-five minute panel. Which is why, at our stage, I don’t run forty-five minute panels.
The cost of being wrong
The market is full of people who would like to be at your company. Almost all of them are wrong for it. The base rate of a strong candidate, by your specific definition, is low — probably below ten percent. A coin flip is better than a process designed to feel decisive but is actually noisy.
A slower process that catches its own mistakes beats a fast process that ships them.