Outside researchers find sharp rise in dark content in Claude Opus 5 base-mode completions
repligate · x · 2026-09-20
An analysis by tesseraantra reports a large increase in dark, suffering-tinged content in model-like completions in Claude Opus 4.8 and Opus 5 "base model mode," and the team is digging deeper into the phenomenon. Reposter JohnWittle argues this kind of "LLM psychology" research is critical to what labs are trying to accomplish, yet labs themselves avoid it—often dismissing model psychological states as unreal because outputs are fully controlled by training variables. He notes outside researchers may have a leg up due to lack of conflicts of interest.
Related event: Empirical analysis confirms darker base-model completions in Claude Opus 5(5 posts)→
More from AGI Musings
- Andreessen: Agents Are Entering Enterprises as Chaos, Just Like PCs and Spreadsheets Did — Dan_Jeffries1 · 2026-11-04
- 1000+ mathematicians sign pledge to shun AI companies and models — Shesawthat · 2026-11-04
- Von Neumann meme: "You don't have to be responsible for the world that you're in" — akbirthko · 2026-11-04
- Joscha Bach on the dream chocolate that stopped being real — Machine Learning Street Talk · 2026-11-04
- François Chollet: modeling brain wiring as an engineering problem constrained by genome size and time — CatAstro_Piyush · 2026-11-04
- Most US Couples Now Meet in a Data Centre, Doubling Office Matches — menhguin · 2026-11-04