Foundational case
A green exit code can hide a broken product.
Autonomous systems need proof that the work is visible and usable, not merely that the process finished without a compiler error.
Read the original on XThe evidence card
The originating failure was simple and devastating: an agent built entity cards for a Show Bible, passed typecheck, and reported success. When the interface opened, 37 of 38 entities were invisible. That gap between structure and reality became a four-tier quality ladder.
The model
The visual is an editorial compression of the source essay: enough structure to orient a reader, without pretending that the model replaces the underlying evidence.
The argument
The site version keeps the source's operational insight while making its sequence easier to inspect, share, and connect to the rest of the archive.
Typecheck proves structure: syntax, imports, and types. It cannot prove that a route renders, that a layout communicates, or that an interaction works. The agent was not lying; the completion signal was too weak.
The quality ladder adds visual verification, a screenshot gate, and a bounded design critique after typecheck. Completion becomes a claim supported by artifacts rather than a feeling emitted by the agent.
The ladder catches more classes of failure, but behavior, taste, and consequential judgment still need people and domain-specific tests. A stronger gate is not an oracle.
The boundary
Done should resolve to evidence.
The quality ladder raises the floor for autonomous work; it does not certify that a product is correct, safe, or valuable.