LLMs are most confident where they should hedge — biology lacks the checker code has
anshulkundaje · x · 2026-09-26
Stanford professor Anshul Kundaje amplifies a discussion: LLMs are most confident exactly where they should hedge. Code and math have executable checkers (tests, proofs) that beat the bluff out of models; open-ended biology never did, making hallucination far harder to constrain.
More from Models
- Perceptron launches Mk1.5, one embodied AI model for drones, quadrupeds and smart glasses — code_star · 2026-09-26
- Tester: Astra is "autistic" at parsing human sentiment; Fable and Opus run circles around it — teortaxesTex · 2026-09-26
- User generates an anime-style fight scene entirely with code using Claude Opus 5.5 — EricBuess · 2026-09-26
- Dev verdict on Opus 5.5: the first good Opus since 4.8 — holdenmatt · 2026-09-26
- Opus 5.5 code review demo draws attention — tlakomy · 2026-09-26
- One prompt, zero libraries: Opus 5.5 builds polished vanilla JS motion graphics in Claude Code — EricBuess · 2026-09-26