Welcome to the SAIPAL blog

This is the blog of SAIPAL, an independent lab working on the safety and security of AI systems. The formal work goes in the research section. This is everything else: lab notes, progress updates, notes on papers I’m reading, and arguments I’m not sure about yet.

Posts here are deliberately less finished than the research. If I waited until everything was defensible I’d never publish anything, and the whole point of this place is thinking in public.

There’s an RSS feed if you want posts as they land.

If something here is wrong, tell me: contact@saipal.net. Being correctable in public is most of the value of writing in public.