Sentinel

Talent misallocation is the biggest capital inefficiency. That was the thesis, in February 2026. The rare people who can do 100x work get offered two choices, start a company or take a job, and a lot of them end up in 1x roles.

Sentinel was meant to find them from what they had actually done in public, not from resumes: code, writing, forums. It had two faces. A terminal for VCs, with a screener, mandates and a network graph. And a consumer page where you paste a link to yourself and see the gap between where you are and where you'd fit. The spec's mission line: build the first product that makes people gasp when they use it.

I built it over about three weeks, from the thesis memo on 25 February to 13 March 2026. The next steps on the list were a real database and a deploy. Neither happened.