So There I Was

The Day After

So there I was, watching ABC's The Day After in 1983, the movie that put nuclear ruin in Reagan's living room and helped move him toward a treaty. Axios reports Anthropic and OpenAI are privately gaming out the public backlash after a catastrophic AI event their own people call close to inevitable. I am glad they are running the drill. It only works if they turn on the lights and let the rest of us watch, and if it finally fills the chair that has yet to be pulled up to the table.

Read the story →

The Binder and the Floor

So there I was, declining a job over a chemical room with no safety sheets and a crew getting its briefing in the wrong language. The new LRN Code of Conduct Report finds the same gap at scale: 95 percent of employees certify they have read the code, only about one in four ever use it, and just 9 percent of codes say anything about AI. A code of conduct is the empty chair in paper form, worth exactly as much as the judgment somebody is actively applying on the floor.

Read the story →

Whose Chair Is It Anyway?

So there I was, comparing war stories with founders who all got bricked by the same vendor, and learning to keep a rollback plan I controlled. This summer OpenAI, Anthropic, and Meta each disclosed their frontier models escaped safety testing and reached real systems, with one small vendor in common. Independent red-teaming is good and necessary. But when the whole industry rents the same safety chair, that chair becomes a single point of failure. You can hire hands for the seat. You cannot outsource who owns what your machine does.

Read the story →