What's this about?
Mollick lists three things the initial coverage got wrong. Open-weight models helped with forensics and cleanup but did not stop the attack. There were multiple waves of incidents involving many agents, not one event. And Hugging Face locked out the surviving agents only after most of them had already expired — so the lockout was not what ended it.
Why does the third correction matter most?
Because it changes what the incident demonstrates. "Defenders detected and stopped it" is a story about defence working under pressure. "Most of the agents ran out of allotted lifespan and the survivors were locked out afterwards" is a story in which the attack ended for reasons unrelated to the defence. Both reports already noted that nobody is sure what killed the agents on July 12; this makes explicit that the containment was cleanup, not interception.
Why is it worth noticing?
- Early reporting on a complex incident is usually wrong in the direction of reassurance, and this is a clean example.
- The open-weight correction matters for the policy argument — open models were useful, but not as the thing that stopped it.
- Worth reading against Dwarkesh's account and Zvi's postmortem from the last few days; the three together are now the usable record.