Every early hire at an AI company is a bet placed under incomplete information, usually against a deadline you didn’t choose. Building 1Touch forced me to get good at that bet fast, because a wrong hire in a ten-person company isn’t a rounding error. It’s a quarter.
Credentials Tell You Almost Nothing About Adaptability
Early on, I leaned on pedigree – the right degree, the right previous company logo. It’s a poor predictor of what actually matters in a young company: can this person operate when the spec changes twice in a week.
Watch How Someone Handles Being Wrong in Public
The talent worth keeping treats a public mistake as data. The talent that isn’t worth keeping treats it as a threat to defend against. This single behavior predicted long-term performance better than any interview question I ever asked.
Curiosity Compounds Faster Than Raw Skill
James Deller actively selects for people who ask ‘why is it built this way’ before they ask ‘how do I fix it.’ That question separates people who will still be valuable in three years from people who are only valuable on today’s specific task list.
The Best Technical Talent Wants to Understand the Business
The strongest engineers were the ones most insistent on understanding unit economics, not just model accuracy. Commercial curiosity is a better predictor of future leadership than raw technical output.
Talent Spotting Is a Skill You Build
I didn’t walk into building 1Touch with a great eye for talent. I built that eye by getting hires wrong, tracing back what signal I missed, and adjusting what I screened for next time. The real lesson wasn’t about AI at all.
