David Holtz

🚨 New research alert! For the past few months, I've been a part-time visiting economics researcher at OpenAI. Excited to share the first public piece of work to come out of this, which uses data from Codex to document the ongoing and rapid shift to agentic AI.

Details below 👇

*a nightclub or a lounge or win son bakery

I think we're hitting the point where LessWrong-style casual empiricism is becoming increasingly counterproductive, and potentially harmful, for reasoning about what happens next with AI.

We’re still in a data-poor regime, so speculation and reasoning from first principles are inevitable. But too often, real empirical observations are combined with plausible-sounding causal stories, and the resulting conclusions are being treated as though they follow directly from the data.

I’m hoping that work in this space will move toward more disciplined reasoning about causality, deeper engagement with the technical, social, and organizational processes shaping AI’s development and impacts, and greater transparency about where the evidence ends and the assumptions begin.

just accidentally called them angine de poutine lol

scoop: david holtz is currently riding the subway home from work