Skip to content

Comment on Contract-Centered Iterative Stability v4.7.3

Comments

I’m the author. This is a bounded experiment, not a claim of universal model failure. The result here was that when a rule had to hold across multiple mutation paths, code-only prompting repeatedly fixed the named path and missed the equivalent unnamed one; spec-first runs updated all affected surfaces. The page includes limitations and linked artifacts.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.