A corollary to Gell-Mann Amnesia: I don't know shit about drug development, but I know rather a lot about the cybersecurity industry. Most of what Anthropic has said about cyber isn't wrong exactly, but it's been breathless and so crudely expressed that it fails to capture the real path that developments are likely to take. For example, Anthropic said that AI might create a giant botnet (no shit, so what) whilst not really talking about agentic offensive frameworks (uh-oh) at all.

I have to assume that they're probably doing the same with medicine.