Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
83 78 88 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 2 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 0 promises resolved, 5 needed for a score.
All types combined: 83 (likely 78 to 88, high confidence), from 86 claims: 86 factual claims. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Daniela Amodei's claims
38 claims
“In the lab, Claude designed new protein binders from scratch.”
Daniela Amodei, Anthropic co-founder and president, foreword 21 October 2026 Anthropic 2026 Public Benefit Report: Taking stock of our progress (foreword by Daniela Amodei) (Anthropic)
In laboratory work, Anthropic's Claude designed new protein binders from scratch.
Mostly true True in substance, wrong in a detail.
The second judge would not resolve this claim; the verdict shown is the first judge's.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
TRUE: on or before 2026-10-21, Anthropic or its collaborators reported (in a publication, preprint or official announcement with technical detail) that Claude generated new protein binder designs de novo, meaning not by changing known binders, and that the designs were made and tested in a wet lab, with at least some confirmed to bind. MOSTLY TRUE: the same, with a minor detail different (for example Claude ran or directed specialised design tools such as RFdiffusion as an agent rather than writing the sequences itself, while still producing the new binders). MIXED: Claude's role in the design was substantial but shared with human scientists to an extent that makes 'Claude designed' only partly accurate, or only some of the binders were new. MOSTLY FALSE: the binders were only designed and evaluated computationally with no lab work, or Claude only assisted human designers in a minor way, or the binders were adapted from existing ones rather than made from scratch. FALSE: no such work happened; evidence shows Claude was not involved in designing protein binders, or the reported work concerns something else.
Evidence
- Claude accelerates protein design and analytical chemistry Anthropic · 18 August 2026 Own record
“Our external evaluators, Adaptyv Bio and Twist Bioscience, independently produced and tested Claude’s designs in the lab, finding that of the 15 targets we designed against, Claude successfully designed binders against 14 of them.”
- Case study: Benchmarking Claude’s protein designs in the wet lab Adaptyv Bio · 20 August 2026
“Overall, Claude designed proteins that bound 14 out of 15 targets, with 1 target (GDF-8 mature) being excluded due to low-quality measurements in our setup”
- Case study: Benchmarking Claude’s protein designs in the wet lab Adaptyv Bio · 20 August 2026
“This Anthropic campaign was an open-loop experiment. Claude designed proteins, Adaptyv tested them experimentally, and here we have presented the results.”
- How Claude designed protein binders against scorpion neurotoxins Anthropic · 20 October 2026 Own record
“Claude’s design campaign produced 17 protein binders that each bind to one of four scorpion neurotoxins that are known to be dangerous to humans.”
Reasoning
Anthropic reported on 2026-08-18 that outside evaluators produced and tested Claude's designs in the lab, and that Claude designed binders against 14 of 15 targets (E1). Adaptyv Bio, which is independent of Anthropic, confirmed that it tested the proteins Claude designed and that they bound 14 of 15 targets (E2, E3). A later Anthropic report describes 17 binders Claude designed against scorpion neurotoxins (E4). This supports the core claim that Claude's designs were made and confirmed to bind in a wet lab, but the quotes do not say outright that the binders were made de novo rather than adapted from known binders. They also do not say whether Claude wrote the sequences itself or ran specialised design tools, so the verdict is mostly true rather than true.
Cite this claim
Daniela Amodei, “In laboratory work, Anthropic's Claude designed new protein binders from scratch.” (said 21 October 2026). Verdict: mostly true. Reliability API, claim c_0e4df8465376, method 0.1.0+28da144b. https://epistacklabs.com/claims/c_0e4df8465376
Claims without a verdict
Is a verdict here wrong, or do you speak for Daniela Amodei? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.