I started Separatrix to make it more likely that (pre-superintelligence) AIs and humans cooperate to avoid catastrophic outcomes, like the development of misaligned ASI. I think that we don't need to solve the hard problems of alignment to cooperate with aligned-enough capable-enough AIs.

Previously I developed a free open-source API-level AI Control platform, Luthien and contributed to METR's RE-Bench (one of the main eval sets that define the METR AI task time horizons chart) while visiting through MATS. Before that I worked on AI and software at Meta, Amazon, and a handful of startups.

I've written some things, one of them was about humanity killing Smallpox. Later in an unrelated incident I deployed an AI that temporarily deleted the Declaration of Independence from Facebook.

Seattle based.