Citrini Research and Anthropic's Claude AI fails to design multi-toxin scorpion antivenom
Current overview (not historical)
Citrini Research and Anthropic jointly used Claude AI to design scorpion antivenom, generating 320,000 protein candidates narrowed to 211 designs. However, 17 of these proteins only bound a single toxin, failing to achieve multi-toxin binding or drug design as intended. Anthropic described the experiment as never testing whether the designs could block toxin activity, framing the outcome as a failure of the autonomous discovery process rather than a biological breakthrough. The entire process took 64 hours.
Written with AI from the reporting · updated 10/22/2026
Expand the detailed fact route and sources
Story route
A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.
- 1Unclassified fact stationOct 22, 22:06single sourcereport-time onlyrevised
Citrini Research 使用 Claude 设计蝎子抗毒血清但未成功
Read original source ↗ The Verge · AIReplay this moment in the explorerEvidence & revisions · 1 source records
- Subject：
- Citrini Research
- Action：
- 使用 Claude 设计蝎子毒素结合蛋白
- Object：
- Claude 模型
- Occurred (fact side)：
- not recorded
- Published (valid side)：
- Oct 22, 20:21
- First seen (known side)：
- Oct 22, 22:06
Fact stationSingle source · one-sided claimRevisedReport-time onlyCopy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.
Uncertainty marks flag where to check the sources; they are not proven causation.
Report timeline
Explore every side through dated source reports.
- The Verge · AIRelated fact: single sourceFact occurrence time unverifiedSource report revisedCitrini Research and Anthropic test Claude for scorpion antivenom development
Report first recorded Oct 22, 22:06
Citrini Research collaborated with Anthropic to test an unreleased version of Claude in developing scorpion antivenom, aiming to design proteins that bind to scorpion venom toxins. Over 64 hours, Claude generated nearly 320,000 candidates, which were refined to 211 designs. Subsequent testing by Adaptyv Bio and Twist Biosciences found that only 17 proteins could bind to a single toxin among four, failing to achieve multi-toxin binding or drug viability. The experiment cost nearly $200,000 in compute and lab testing.
Read original source ↗
Event discussion
Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.