User stats show Claude admitted 1,897 errors in 10k messages, self-portrait resembles an anglerfish
corozcop · reddit · 2026-09-02
A user analyzed 10,727 messages across 343 sessions with Claude Code, revealing that 33% contained corrections or complaints, with a peak of 187 "blow-up" messages in a single day. The stats show Claude admitted "you're right" 1,897 times and repeated mistakes 249 times. The author argues this pattern erodes user trust and creates unresolved grievances since the AI cannot be held accountable like a person. Finally, Claude was asked to generate a self-portrait based on the chat history, choosing an anglerfish luring jellyfish with its light—symbolizing apologies luring the user back.
More from Fun
- Made by Hand turns game crafting into a real woodworking job, one saw cut at a time — WellsLucasSanto · 2026-10-18
- LLM one-shots a Minecraft 3x3 piston door by building its own Linux test infra — Vjeux · 2026-10-18
- Yacine jokes his datacenter KPI is gigawatt-hours: get paged if not burning enough power — yacineMTB · 2026-10-18
- Siri and Google botch the same euro-dollar conversion, screenshot shows — jdluk87 · 2026-10-18
- Codex took 3 days, Opus 5.5 nailed it in an afternoon — banned user's video protest — dotey · 2026-10-18
- Peter Diamandis-backed AI documentary 'The AI Doc' hits theaters nationwide — irinarish · 2026-10-18