If your agents are doing the execution, the humans you hire are pure judgment — which means a weak hire now costs you something agents can't fix.
Vinod Khosla's gene pool engineering framework insists that a founding team is not a collection of impressive backgrounds but a precisely assembled set of prior risk-retirements. You map your venture's most lethal uncertainties first — technical, regulatory, distribution, whatever actually kills you — and then you find people who have personally navigated each of those specific cliff edges before. The résumé is incidental. The scar tissue is the asset.
The translation is almost uncomfortably direct. If agents are handling research, synthesis, code review, and first-pass decisions, you have fewer human seats, which means each human must carry more risk-retirement weight per dollar spent. A generalist who looks good on paper but has never, say, survived a regulated deployment or rebuilt trust after a model failure is now genuinely unaffordable. Khosla's method becomes a forcing function: audit your agent stack for residual human-judgment chokepoints, name the risks that live there, and hire only the person who has already killed that exact thing once before.
- Map your kill risks before you write a single job description
- treat agent coverage as a filter that surfaces which human judgment gaps remain truly exposed
- the smaller your human team, the less tolerance you have for anyone who hasn't retired the specific risk you're handing them.
