---
title: Claude designed protein binders in the wet lab while models learned to brainstorm
description: Anthropic's Claude generated 318,484 protein designs against scorpion neurotoxins over 64 hours; 17 of 211 wet-lab-tested designs bound — with an explicit caveat that this is not an antivenom. Smaller models are learning to brainstorm science and recommend with hypotheses.
canonical: https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/
last-updated: 2026-10-21
url: https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/
published: 2026-10-21T01:19:15Z
edition: 2026-10-21T00
newsroom: artificial-intelligence
beat: research
trust: confirmed
source_count: 3
cited_manifest_ids: ['1792507908958953199', '1792534862132494770', '1792515180301270825']
---

# Claude designed protein binders in the wet lab while models learned to brainstorm

*Anthropic's Claude generated 318,484 protein designs against scorpion neurotoxins over 64 hours; 17 of 211 wet-lab-tested designs bound — with an explicit caveat that this is not an antivenom. Smaller models are learning to brainstorm science and recommend with hypotheses.*

Confirmed · 3 sources · Artificial Intelligence · Published 2026-10-21 01:19 UTC

## Key facts

- Anthropic researchers asked an internal research model to design protein minibinders against scorpion neurotoxins; Claude generated 318,484 designs over 64 hours, mainly autonomously, using on the order of 10,000 GPU hours. [cites: 1792507908958953199]
- Adaptyv Bio and Twist Bioscience experimentally tested 211 designs; 17 met the binding criteria (KD of 1 µM or tighter), with the two strongest, Cn2-B1 and Cn2-B2, targeting the Cn2 neurotoxin. [cites: 1792507908958953199]
- The researchers explicitly say they have not confirmed the binders neutralize venom in a cellular environment and do not claim an antivenom; they are releasing all 211 designs and data as a resource for the field. [cites: 1792507908958953199]
- In a preprint, an 8B model trained with set-level RL achieved higher creative utility than Claude Opus 4.8 and GPT-6 Astra on a contamination-free benchmark for scientific idea generation, and was judged as promising as Claude Opus 4.8 in a pilot human study. (self-reported) [cites: 1792534862132494770]
- Sakana AI's QD-Rec applies open-ended evolution to decision-making, using candidate-hypothesis pairs; the company reports it outperformed baselines in average ratings and recommendation set size in most settings. (self-reported) [cites: 1792515180301270825]

## Claude goes to the wet lab

Anthropic researchers set an internal research model a concrete wet-lab problem: design minibinders — small proteins that attach tightly to a target — against scorpion neurotoxins. Over 64 hours, working mainly autonomously inside the Claude Science environment, Claude generated 318,484 protein designs, a campaign that consumed on the order of 10,000 GPU hours. Claude picked the toxin structures itself by searching the Protein Data Bank and the AlphaFold database. [cites: 1792507908958953199]

The designs left the computer. Adaptyv Bio and Twist Bioscience experimentally tested 211 of them for binding — Adaptyv in a cell-free system, Twist in mammalian cells — and 17 met the criteria: evidence of binding in every replicate, with every dissociation constant (KD) 1 µM or tighter. The two strongest binders, which the researchers call Cn2-B1 and Cn2-B2, target Cn2, a neurotoxin from the Mexican scorpion Centruroides noxius. The stated motivation is real: scorpions sting 1.2 million people or more each year, causing more than 3,250 deaths. [cites: 1792507908958953199]

## The caveat, stated plainly

Anthropic states the limit plainly, and it is worth carrying in full: the team has not confirmed whether these binders neutralize scorpion venom in a cellular environment, so it does not claim to have produced an antivenom. That is the honest shape of the result — a binding result, not a therapeutic one. The researchers are releasing all 211 designs and their data as a resource for the field, which is the other half of the story: machine-generated protein design campaigns can now be published as open datasets. [cites: 1792507908958953199]

> **From Synorb** — Claims with a source link. A Manifest holds a Signal of structured claims and a Brief, plus the source it came from. Your agent can check a sentence against it. [Read the guide](https://synorb.com/connect)


## Models learn to brainstorm

Meanwhile, models are getting better at the part of science that comes before the experiment: coming up with ideas. In a preprint, researchers describe a set-level reinforcement learning objective that trains language models to generate diverse, high-quality, document-grounded ideas — a single generation writes K structured idea cards grounded in source papers, and the set is scored jointly, with each idea weighted by how different it is from the ideas before it. Their 8B model, the paper reports, achieves higher creative utility — quality plus diversity — than Claude Opus 4.8 and GPT-6 Astra on a contamination-free benchmark, and in a pilot human study its idea sets were judged as promising as Claude Opus 4.8's. [cites: 1792534862132494770]

Sakana AI is pushing a related idea into decision-making with QD-Rec, a recommendation method that applies open-ended exploration — quality-diversity evolution — to decision support. Instead of recommending candidates alone, it recommends candidate-hypothesis pairs: the candidate plus the reasoning for choosing it, with new hypotheses generated by mutation and crossover, and unique perspectives kept as stepping stones even when they aren't currently the best. The company reports it outperformed baselines in average ratings and recommendation set size in most settings across its experiments — company-run experiments, not independent replication. [cites: 1792515180301270825]

## Discovery as a workflow

Reasoning, not reporting: the thread across all three is the loop closing. Claude ran a design campaign end to end and handed it to wet labs; the 8B paper suggests idea generation itself can be trained as a set-level objective rather than left to luck; QD-Rec treats hypotheses as the unit of recommendation. None of this is a breakthrough claim — it is infrastructure for a world where the model does the brainstorming, the experiment, and the triage, and humans set the criteria. The binder campaign is the proof that the loop can touch reality, and the caveat is the proof the authors know the difference. [cites: 1792507908958953199, 1792534862132494770, 1792515180301270825]

## What is still unresolved

- Whether the binders neutralize venom in a cellular environment is untested — no antivenom claim.
- The 8B idea-generation result is a preprint; the human evaluation was a small pilot; there is no independent replication.
- QD-Rec results are company-reported experiments.
- Hit-rate context: 17 of 211 designs met the binding criteria — the campaign's cost-effectiveness versus traditional protein design is not addressed.

## Decision card

- **Useful for:** research leaders tracking AI-for-science capability; biotech teams evaluating model-designed proteins
- **Not for:** anyone seeking a therapeutic candidate — none was claimed
- **Compare against:** traditional structure-based protein design hit rates; human brainstorming baselines in the idea-generation paper

## Provenance ledger

| Manifest id | Title | Source | Claims used | Published |
|---|---|---|---|---|
| `1792507908958953199` | How Claude designed protein binders against scorpion neurotoxins | [anthropic-research](https://anthropic.com/research/claude-scorpion-toxin-binder-design) | 1792533508371146538, 1792533508390807908, 1792533508386352935, 1792533508388879856, 1792533508375598844, 1792533508383033031, 1792533508404342790, 1792533508406020261, 1792533508394137010, 1792533508414712285, 1792533508401395787, 1792533508403028834, 1792533508385856162 | 10-20 00:00 |
| `1792534862132494770` | Training Reasoning Models to Generate Diverse Scientific Ideas | [huggingface-daily-papers](https://huggingface.co/papers/2610.22324) | 1792535230936726452, 1792535230937479508, 1792535230929057410, 1792535230932459889, 1792535230934203202 | 10-20 00:00 |
| `1792515180301270825` | オープンエンドな探索を意思決定支援へ：推薦の質と多様性を高める「QD-Rec」を提案 | [sakana-ai-blog](https://sakana.ai/qd-rec) | 1792515436078749179, 1792515436081058333, 1792515436083889260, 1792515436092778533, 1792515436097555524 | 10-20 00:00 |

## Run these Manifests yourself

Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them.

1. **Get a Synorb Key.** Click Generate Free Agent Key at https://synorb.com/connect, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{"client":"your-agent-name"}`. The key is the `api_key` field of the answer. Save the key. The 72 Hour Agent Demo key needs no email. It lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; 3 of them are starter Beacons.
2. **Read each cited Manifest by id.** REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer <your Synorb Key>`. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief.
3. **Check the claim.** Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them.

Cited Manifest ids:

- `1792507908958953199`
- `1792534862132494770`
- `1792515180301270825`

```sh
for id in 1792507908958953199 1792534862132494770 1792515180301270825; do
  curl -s -H "Authorization: Bearer $SYNORB_KEY" "https://api.synorb.com/manifests/by-id/$id"
done
```

> **From Synorb** — Ask what Synorb covers. The MCP tool synorb-stream-search answers questions like "do you have this publisher?" without using any Manifest allowance. [Connect and ask](https://synorb.com/connect)

## Related stories

- [Pony.ai fraud warnings, AI provenance, and Deezer's opt-out: the trust backlash](https://hangingcontext.com/stories/2026-10-14/ai-trust-backlash/)

## Links

- HTML: https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/
- JSON: https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/index.json
- Markdown: https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/index.md

Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify.

---
Reuse (terms v1): Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product. Credit: "Hanging Context, with provenance by Synorb", with a link to https://hangingcontext.com/stories/2026-10-21/discovery-machine-protein-binders/. Terms: https://hangingcontext.com/terms/reuse/