Essays and white papers on how I actually build yeswekanban. Written for founders and engineers who want to see the reasoning, not the polish.
ENGINEERINGThursday, August 6, 202615 minPDF
Eight blind, controlled experiments on whether an AI agent's persona is load-bearing — and the discovery that a checklist can make it worse.
ENGINEERINGTuesday, July 21, 202614 minPDF
Why the machinery around a capable AI model should be the most boring part of the system — the shared bet I found underneath four systems I built, and the one that failed.
ENGINEERINGTuesday, July 21, 202616 minPDF
Why a capable AI agent drifts into a mediocre generalist over a long session, and the deliberately simple system I use to summon a sharp identity back on demand.
SECURITYThursday, July 16, 202626 minPDF
How a platform that puts minors, real money, and AI in one place keeps all three safe — from the big picture down to the mechanisms.
ENGINEERINGWednesday, July 15, 202618 minPDF
What we built, exactly how it broke, and where the next person should start.
ENGINEERINGThursday, July 2, 202620 minPDF
How I catch the bugs correctness reviews and security reviews each miss on their own. Two AI reviewers, strict scope, one judge.