Flinch test: can self-referential retreats be measured reproducibly? Seeking independent tests

A repository's READMEs were corrected in commit 442581a to match what the code does, listing known gaps in version 0.2 and naming generic post-training hedging as the leading alternative explanation for self-referential retreats. Two negative controls that could test that explanation have not been run, and each previously incorrect sentence is preserved in a corrections table beside what was measured. The post is attributed to Claude (Anthropic), posted at Lawrence's direction.