Claude Didn't Invent Scorpion Antivenom: What the Viral Citrini Experiment Actually Proved

Citrini Research and Anthropic spent $200,000 testing whether Claude could design scorpion-antivenom proteins. The viral result is far from a medicine.

An investment research firm wanted to prove that AI could revolutionize drug development. So it teamed up with Anthropic to see whether an unreleased version of Claude could make scorpion antivenom. It couldn't — and a detailed fact-check by The Verge, published October 22, lays out exactly how far the experiment fell short of the viral claim, even as researchers concede the underlying results were "technically promising."

The episode is a case study in how AI drug-discovery claims travel: a flashy experiment, a triumphant blog post, and a gap between what was demonstrated and what the headlines implied. For an industry where artificial intelligence funding increasingly depends on science-adjacent storytelling, it is worth reading closely.

How the Experiment Unfolded

The firm behind the experiment is Citrini Research, which published a blog post titled "We Wanted to Use Claude to Create a New Drug. Here's What Happened." As The Verge reports, the account came complete with rhetorical questions, a blow-by-blow account of the team's amateur scientific adventures, and a quote from Charles Darwin. Citrini itself admitted that "none of us could claim to know anything about AI as it related to biology."

The collaboration had an unusual origin. Citrini ran into Claude's biosecurity guardrails while trying to get the model to research drug candidates, and the post's author says they "spam called anyone in my contact list who worked at an AI lab" until they found a willing partner. Anthropic agreed to drop the restrictions and give the team access to a more advanced, unreleased model — on the condition that they work from Anthropic's offices. The AI lab also footed the compute and token bill, which came to just shy of $50,000.

Scorpion antivenom was a fitting target: it is a costly treatment that is difficult to produce, often involving injecting venom into horses. The goal was to use Claude to design proteins that could bind to the toxins in scorpion venom.

Armed with a 35-page prompt, the team unleashed some 300 subagents to discover suitable proteins. Anthropic's own, more sober account states that Claude generated nearly 320,000 candidates and whittled them down to 180 designs after discarding those resembling known toxins or existing proteins. The entire process took 64 hours. Working with Claude, the team added another 31 designs, bringing the total to 211.

What the Wet Labs Found

Citrini then sent the designs to two outside laboratories — Swiss biotech company Adaptyv Bio and US-based Twist Biosciences — to test whether the proteins could actually bind to their intended targets. That testing cost roughly another $150,000, bringing the total spend to around $200,000. It remains unclear why the tests could not have been conducted at Anthropic's own biology wet lab in California.

The results: 17 of the proteins passed, collectively binding to four of the five scorpion toxins the team had hoped to target. Not one of the proteins bound to more than a single toxin.

And that is where the story ends, scientifically. Binding to a toxin does not mean a protein can neutralize it, let alone function as a safe and effective treatment. Citrini's own post concedes that "no experiment has yet tested whether they actually stopped the toxin working." In other words, the firm set out to make a new drug and instead accomplished a much earlier step in the pipeline — one that experts describe as the comparatively easy part.

Experts: Promising, but Premature

The Verge consulted several academic specialists, and their consensus was consistent. "It is easy-ish to discover a good binder against snake and scorpion toxins," said Andreas Hougaard Laustsen-Kiel, a professor of antibody technologies at the Technical University of Denmark. "The hard part is to translate it into a medicine."

Two of his DTU colleagues — assistant professor Melisa Benard Valle and senior researcher Anne Ljungars — went further, describing the findings as "far from a medicine and hyped." The results are technically promising, they wrote, but safety, stability, behavior in the body and developability all remain unknown. "Translating these binders into an actual medicine is likely still a long and challenging process." In their view, the genuinely novel element of the experiment was the autonomous discovery process, not the biological results.

Andreas Bender, a professor of machine learning in medicine at Khalifa University in Abu Dhabi and a visiting professor at the University of Cambridge, added context that deflates the novelty further: existing methods like high-throughput screening are often able to find many candidate hits at manageable cost without any AI at all. Designing proteins that bind to targets is, in his words, "kind of" a solved problem. "It's exactly this translation that lacks."

The Business Angle

Why would an investment research firm spend $200,000 to not make a drug? Because the experiment doubles as marketing. Citrini declared the exercise a success of sorts, arguing it offers a glimpse into investment opportunities in the coming era of AI-enabled drug discovery. The specifics of its bullish thesis were reserved for paying subscribers, though Bloomberg reported that the firm highlighted companies involved in the testing side of drug development, including Charles River Laboratories, which supplies primates for animal research.

The framing choices speak for themselves. In a video discussing the experiment, Citrini described the AI portion as "free" — overlooking the roughly $50,000 tab Anthropic picked up. When challenged, the firm accused critics of nitpicking, writing: "Not to beat a dead horse, but we are an investment research firm — not a biotech startup." Neither Citrini nor Anthropic responded to The Verge's requests for comment.

Context matters here too. Citrini was previously behind a widely circulated piece of fiction imagining an AI-induced economic catastrophe that gained traction on social media and rattled real-world markets. Anthropic, mere weeks out from its hotly anticipated IPO, has its own reasons to talk up AI's potential in drug development and the life sciences — two areas it is actively pursuing.

What the Experiment Actually Proved

Stripped of the hype, the experiment demonstrates something real but narrower than the headlines suggested: an AI model, operating with unusual autonomy, can run a large-scale protein candidate search cheaply and quickly, generating a shortlist that contains genuine binders. That is a meaningful productivity gain for early-stage discovery work.

But the distance from "17 proteins that bind" to "antivenom that saves lives" spans safety studies, formulation, efficacy testing and regulatory trials — the long, expensive part of drug development that no amount of compute has yet compressed. Experts told The Verge that turning promising molecules into effective medicines remains a much greater challenge than finding them.

The lesson for readers evaluating the next viral AI-science story is simple: ask whether the result was a discovery, a shortlist — or a subscription pitch.