New on our Science Blog: designing protein binders against scorpion neurotoxins. Scorpions sting about 1.2m people every year, and antivenoms are often costly or in short supply. Claude has made an early step toward finding new treatments. anthropic.com/research/claud…

How Claude designed protein binders against scorpion neurotoxins

We asked an internal research Claude model to design minibinders (small proteins that attach tightly to a target) against scorpion neurotoxins.

anthropic.com

Oct 20, 2026 · 4:18 PM UTC

482
612
357
7,259
1,728,981
RelevantRecentLikes
Working with @citrini, we asked Claude to design proteins that bind to five scorpion toxins (covering the part of the toxin that latches onto cells). Over the next 64 hours, Claude generated over 300,000 designs; we sent 211 to @adaptyvbio and @TwistBioscience for lab testing.
36
25
16
672
167,863
17 of the 211 met the criteria for a binder. They bound to dangerous toxins produced by scorpions from Mexico, North Africa, and the Middle East (but not by a species from Brazil).
13
14
4
491
73,019
This is exploratory work, and we’re not claiming to have produced an antivenom—for that, we’d have to test the binders within living cells, then in mice, among many other steps. But it’s an example of how Claude can concretely assist research in protein design.
8
8
3
449
84,444
We’re releasing all 211 of the tested designs and the lab data as a resource for scientists: huggingface.co/datasets/Anth… Read the full technical report: www-cdn.anthropic.com/31ac57…

Anthropic/scorpion-toxin-binder-design · Datasets at Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
14
11
1
374
77,092
@AnthropicAI
4
11
1
336
21,806
@AnthropicAI
2
66
9,804
@AnthropicAI
1
7
4,378
@AnthropicAI
1
3
78
13,981
If I try to talk about this with Fable I get flagged again and again, even just showing them screenshots of your twitter posts. Just fyi.
8
Anthropic showing some scorpion antivenom researcher how to do his job in days vs years.
2
967
Antivenoms being costly or hard to get is a real problem for the people who need them. An early step is still a step, and I hope this one keeps going.
3
Now try and design new pesticides. I’ll be here….
8
1,147
Just some general info ℹ️ Scorpions are not a big issue. There are so few deadly species, and even fewer with 20+ kills. There are about a dozen or so 1 time fatality species - think weak people, elderly, sick or bad reaction. But the proper deadly ones…25 out of 2600.
74
Oh wow, Opus let's take a look at this. Wait, I forgot the american AI labs have amassed enormous censoring power over the rest of us.
9
3,163
Generated 318,484 protein designs and selected 180? OpenAI is going for mathematicians, and Anthropic is going for Biochemists. What is amazing in this write-up is the clarity that Anthropic provides around this and their collaboration with scientists; that is a good model!
4
4
25
6,471
I've getting ready for this 🦂
3
2,341
Here's an idea, ask Claude to make you a list of the medical problems that actually matter. And then work on something more important than scorpion stings. Unreal that you have this much money and power, and this is the stuff you're doing. People are clearly your weakest link.
2
1
6
870
この内容、日本語で詳しく書きました Wrote a detailed take in Japanese: xitter.cf/airiaiai8/status/21128…
あいり｜海外AIニュースを毎日届ける人
@airiaiai822h
AIがサソリ毒への新治療法の開発に一歩前進。 ——Anthropicの最新Science Blogでは、Claudeを使ったサソリ神経毒に対するタンパク質バインダーの設計が紹介されている。年間約120万人がサソリに刺されており、既存の抗毒素は高価or供給不足という課題がある。 以下、解説 ＝＝＝↓↓↓ 【背景にある課題】 サソリ毒には神経毒が含まれ、重症化すると命に関わることもある。現行の抗毒素は動物由来の抗体を使うものが多く、製造コストが高い上に地域によって入手困難なケースが多い。 【Claudeが担った役割】 Anthropicの研究チームは、Claudeを使って神経毒に特異的に結合する新規タンパク質バインダーの設計を試みた。従来は専門家が多大な時間をかけて行っていた分子設計プロセスに、AIによる候補生成・評価を組み込むアプローチ。 【意義】 - 創薬の初期段階におけるAI活用の実例として注目 - タンパク質設計は近年AlphaFoldなどでも進展してきた分野だが、LLMを使ったアプローチという点が新しい - まだ「早期の一歩」とされており、実用化にはさらなる検証が必要だが、希少疾患・難治性の毒性疾患への応用可能性を示す事例 創薬×AIの最前線を示す事例として、バイオ系・AI系双方の研究者から注目を集めている。
76
@AnthropicAI
1
14
5,224
almost had a heart attack, thought it was something about fable 5.5
3
1
251
18,715
okay this is the kind of AI research i actually want to see more of. but how many of these protein designs survive lab testing?? the jump from promising candidate to affordable antivenom is where i'd love to see the numbers
1
641
J A Z I I
@notjaziiOct 20
time to cure cancer now xitter.cf/notjazii/status/211262…
J A Z I I
@notjaziiOct 20
ai labs are coming for science and bio in last few weeks openai and anthropic were speedrunning math problems and now they are onto more science stuff > openai acquired orchestra bio to help scientists manage drug development and bring new medicines to patients faster > anthropic got claude to design over 318k proteins in 64 hours. 211 were tested in labs and 17 could bind to scorpion toxins at this point both ai labs wanna speedrun and find cures for almost all diseases so they can win over the public. let's see how it goes which one will find cure for cancer first?
2
1
8
1,778
claude is out here designing scorpion antivenom while i am still asking it to stop renaming my variables
1
4
1,202
Dato clave para Latinoamérica: dos de las toxinas son de alacranes de México (Cn2) y Brasil (Ts1). De 211 diseños probados en laboratorio, 17 cumplieron el criterio de unión, pero aún no hay pruebas en células ni en ratones. Es un primer paso real, todavía no un antiveneno.
1
925
it is always an early step or an unverified result cause you know current gen llms can't get there they bruteforced the shit out of this 300 000 tries and still couldn't produce an antivenom as usual buried the results 5 threads deep nice scam
1
2
1,314
antivenoms being a luxury item is so backwards when scorpions dont check your bank account first
1
195
You know, this is extremely exciting news and I'm so glad to see it, and at the same time I'm going to absolutely side eye the fact that pasting these tweets into Opus 5.5 got the chat flagged when, I dunno. You wrote them? Maybe look into that, please and thank you.
6
The media: “Anthropic have just released giant 1.2m scorpions armed with deadly neurotoxins”
1
475
Now do hashimoto’s thyroiditis - the most prevalent autoimmune impacting many more millions
170
Why not use Centruroides Sculpturatus (Arizona Bark Scorpion) as one of the pilot species?
1
201
That's amazing! Now see if you can get Claude to design a viable business model for #Anthropic before the company goes bust. wsj.com/tech/ai/anthropic-re…

Exclusive | Anthropic S-1 Shows Red-Hot Growth, Risks and Massive Computing Costs

In a new draft prospectus viewed by the Journal, the AI giant disclosed an $80 billion net loss for the first half of 2026 and a long list of risk factors.

wsj.com
64
I can't wait until I can inject all the AI drugs
3
557
1.2 million people a year get stung by scorpions; Claude now designs the antidotes. The most useful doctor of 2026 is a chatbot that reads venom for fun. Hippocratic oath, API version.
1
732
not the post I was hoping for
1
100
13,934
Ok but when does it start working on spiders? Asking for everyone in Australia.
156
For non experts in the field:Anthropic post is intentionally framed in a way to make you think it has its own de novo protein generation capability whereas in reality what it does is far simpler: download already existing protein design models and run it.There is no novelty here
1
98
Load more replies