Scorpions kill more than 3,250 people a year, mostly kids, and the cure is still made by milking venom from thousands of scorpions and injecting it into horses. Anthropic just let a team of investment researchers with zero biology background try to cut the horse out. The horse method is over a century old. You immunize the animal for months, draw its blood, and purify out the antibodies its immune system made. Given early, it works against a narrow set of toxins, and because you're putting horse protein into a child, it can trigger anaphylaxis. Barely anyone makes it, because the people getting stung can't pay enough to cover the factory. Mexico has about a quarter of the world's scorpion stings and relies on a single lab. Last year in Sonora, a 5-year-old was stung at preschool and rushed to a public hospital that had already burned through a shipment delivered days earlier. Staff found more 45 minutes later. She'd already gone into cardiac arrest twice. So Citrini Research walked into Anthropic with a question. Could motivated amateurs with Claude make real progress on a disease the drug industry ignores? An internal Claude model ran for 64 hours off one prompt and four follow-ups. It drove 12 protein design tools, generated 318,484 candidate molecules, ran 553,052 simulated matchups against five scorpion toxins, and picked 180 to build. Then it emailed the order to the lab itself. Two labs made 211 of them. 17 latched onto their target toxin, covering scorpions from Mexico, North Africa and the Middle East. Brazil's yellow scorpion beat every single design. The best two grab a Mexican scorpion toxin at the same spot as an antibody already shown to protect mice. The computer finished in under three days. The lab work stretched past a month, held up by contracts, shipping, and the hassle of sourcing the toxins themselves. One of the outside experts who reviewed the results said binding is no longer the main bottleneck. There's no antivenom yet. None of these molecules has touched a living cell or a mouse. Anascorp, the only scorpion antivenom the FDA has approved, took 35 years to go from a Mexican lab to US approval. The first candidate molecules for a horse-free successor took 64 hours, for a team whose day job is writing investment research.