Claude designed scorpion toxin binders: 17 of 211 lab-tested designs bound
shoumikgoswami · reddit · 2026-11-01
- The twist: Citrini Research, a financial investment research firm, approached Anthropic to test whether Claude could help motivated non-experts tackle a real health problem — scorpion neurotoxins, hoping binders could one day lead to an antivenom.
- Per Anthropic: an internal research model worked mostly autonomously for 64 hours with just four follow-up prompts, generating 318,484 protein designs and selecting 180; the team added 31 more with Claude's help. Adaptyv Bio and Twist Bioscience lab-tested all 211, and 17 met Anthropic's binder criteria.
- Anthropic notes it hasn't confirmed neutralization in a cellular environment and doesn't claim an antivenom. All 211 designs and data are being released openly.
Related event: Claude Designs Protein Binders, 17 Validated Against Scorpion Toxins(2 posts)→
More from Research
- If your AI scores 100%, your eval is wrong — an AI safety researcher's eval playbook — AI Engineer · 2026-11-01
- Control Vectors experiment: letting an agent manage its own activations — sloppenheimer · 2026-11-01
- FlashRSI open-sources RSI research: 12 algorithms, 30+ benchmarks, OOD gains still elusive — HuaxiuYaoML · 2026-11-01
- Carbon-X: midtrained Qwen3-14B beats every open LLM on DNA benchmarks — LoubnaBenAllal1 · 2026-11-01
- RL fails without midtraining: Qwen3-14B stays at chance on FrontierVEP — LoubnaBenAllal1 · 2026-11-01
- Hugging Face releases FineDNA: 300B+ token dataset to teach LLMs DNA — LoubnaBenAllal1 · 2026-11-01