Andrew Curran

I said at the time - and repeatedly since - that things changed in November 2025, that the pace sped up, and that we were on a new trajectory. There's no denying now that was true. So, for the record: things changed again in September 2026. We are on a new trajectory.

Gregory Allen talking to Bloomberg about the possibility of passing any strong AI regulation in the next two years:

'The second thing I would say is, in terms of what regulation might actually look like, don't be fooled by the sort of press release style legislation proposals. There are folks who are saying ban superintelligence now. There are folks who are saying ban data center construction nationwide right now. None of that really has any chance of passing. Whatever legislation is going to become law in the next two years has to survive 60 votes to beat a filibuster in the Senate, and potentially has to survive a White House veto.'

youtu.be/AbaaThgpc0c?si…

Anthropic's valuation has now been roughly cut in half from the lofty $2 trillion heights that were being spoken of in financial circles only a month ago. On the plus side, if you are a believer, you can now get in for about half price. The IPO media blitz is starting Monday.

The reason humans have such difficulty judging exponential growth is how rarely examples of it appeared in the ancestral environment. As a result, it was absent from our training data. For us, exponential growth is an out-of-distribution problem.

This is a teaser for Jim Fan's keynote on Monday. The hand movements look next-level. Dr Fan confirmed in the replies that it's real hardware, fully autonomous, and not sped up. It's possible that Project GR00T has made a breakthrough with their hyperbolic time chamber training.