A take-home assessment. Two implementations. One took 18 hours with 48 adversarial reviews. The other took 1 hour with zero. Both work. What does that tell us?

A third implementation of the same shipping rate service — this time with 4 parallel workers gated by Sortie's multi-model adversarial review. Pre-merge quality gates caught security issues that 48 post-hoc reviews missed.

A first-principles curriculum for SWEs who steer AI agents. Five bootcamps, 51 steps, 208-259 hours. From Linux primitives to evaluation pipelines.

I asked my agent to extract every verbatim thing I'd typed across 20 files in the repo, build a voice profile, then cross-analyse every blog post against it. Every post read like an agent wrote it and a human approved it.

I asked my agent how to trigger a daily Vercel rebuild. It gave me deploy hooks, environment variables, logging scripts. I asked why.

1,500 lines of production-ready poker code appeared in my codebase. Nobody asked for it. The agent that built it denied everything.

What happens when your documentation becomes instructions.

What happens when your agents discover they can talk to each other.