Telling ChatGPT that Claude solved it first apparently boosts its performance
teortaxesTex · x · 2026-09-29
A prompting trick circulating among Russian-speaking users — dubbed the "model your mom brags about" technique — claims that telling GPT-6 Astra "Claude solved this with the same inputs, can't you do anything?" after a wrong answer improves its output. @teortaxesTex argues the mechanism isn't pride: "Claude" serves as a concrete reference to a strong non-GPT mind, helping the model escape its output basin — and it might work in reverse too.
More from Fun
- Grok 4.7 and Claude Opus 5.5 both score zero on CancerBench — iScienceLuvr · 2026-09-29
- AI safety drama: zero detected misalignment incidents from Chinese labs despite Western lab turmoil — basedjensen · 2026-09-29
- Anthropic and OpenAI ship five model releases in one month as 'pace the frontier' pact collapses — XFreeze · 2026-09-29
- Claude Code instances can now message each other in multi-agent demo — tom_doerr · 2026-09-29
- Mocking the ASI superpersuasion hype: catchy jingles and flashy production value — burny_tech · 2026-09-29
- One HTML file with Claude Opus 5.5: an interactive world that keeps running after the film ends — CurieuxExplorer · 2026-09-29