Claude yields 17 qualifying binders in antivenom test

The experiment grew out of a partnership between Anthropic and investment research firm Citrini Research, which brought five staffers to the project. Biosecurity restrictions initially blocked work on Citrini's own computers, so the team moved to Anthropic's offices, where it used an internal model and received help on a prompt of about 35 pages. The run generated 318,484 candidates over 64 hours using 12 specialist tools; Claude selected 180 designs and humans added 31. Citrini reports five human messages during that initial design run and 367 across the whole project. The budget breakdown shows where time and money went: $49,997 in compute costs, just under $150,000 in lab charges, and a wait of roughly two months for experimental work. Citrini's account leaves the full cost, including Anthropic's staff time and any subsidized model access, unclear. Anthropic's support for the project sets limits on how broadly the approach could be repeated. The company's Life Sciences Verification Program, launched September 17, reviews research credentials, security standards and ethical oversight, and initially served teams and institutions. Access rules could shape who can carry out similar work. In Anthropic's own lab, announced in September, Claude helps form hypotheses and interpret results, while human scientists perform all laboratory work. For newcomers, access to scientists who can design sound tests and assess evidence could be as important as access to the model. Whether the 17 binders can stop the toxins' effects in cells or protect animals remains untested, and Citrini called its universal antivenom experiment a failure. A separate Anthropic research note published September 17 said a simplified workflow reached similar predicted binding scores with about one hundredth of the GPU time required by earlier campaigns, though those comparisons measured computer predictions and equivalent success in lab tests remains unproven.