Researcher: Frontier LLMs Still Fail at Basic Molecular Cloning Tasks
nlarusstone · x · 2026-10-31
The author amplifies the claim that LLMs remain weak at biology — even basic molecular cloning is beyond frontier models.
They add that better harnesses can boost this performance considerably, suggesting current results are partly limited by tooling.
Related event: Frontier LLMs Fail at Basic Molecular Cloning Tasks(2 posts)→
More from AGI Musings
- Human-in-the-loop theater: approving Claude's 400-line bash scripts with a perfunctory LGTM — generativist · 2026-10-31
- AI safety researchers argue labs' silence on x-risk is a large-scale ethical mistake — DavidSKrueger · 2026-10-31
- The scenario where AI is worth $1T, not $5T, makes bulls and bears sad at once — bookwormengr · 2026-10-31
- DeepMind's Mary Phuong on AI loss-of-control risk sparks debate over freezing frontier AI — GarrisonLovely · 2026-10-31
- Tegmark's "replacement species" AI framing called baseless and dishonest by critic — JosephJacks_ · 2026-10-31
- OpenAI releases 700+ math manuscripts in one batch; Fortnow says it beats 30 years of progress — kimmonismus · 2026-10-31