What I Build

Building, toward AI safety

Building things is so much fun. Whether that be building something out of code, building something out of interwoven community, or building something physical. Building with code has gotten so much more fun as AI has gotten better and better. The time between idea and functioning prototype is hours. Ideas can turn into fully fleshed reality so quickly, it's so much fun!

Current Focus

AI Safety

I work full time on AI safety at Trajectory Institute. This feels like the most important problem to be working on right now. I came in through mechanistic interpretability and self-study; what I actually spend my days on now is closer to red teaming, mapping the ecosystem, and figuring out what the field is missing. There's a running record of the output on my AI safety portfolio.

Previous Project: D&D Tools

Before pivoting to AI Safety, I was working on a D&D project, it's a mix of a reddit style forum for sharing content in a specific domain combined with AI to make creating that content as easy as possible. Lots of interesting problems here.

This site

When I get burned out of working on my main project I work on this site! It's a little sandbox. If I want to write about myself I can add a new node. If I want to think about design I can work on a new look for some of the nodes. If I want to think about how to implement a new feature I just add something new like a tree of all 1+1 guests, a dynamic map of neighbors, or how to implement hidden content or one way connections that depend on the user's start/decisions. It's so fucking cool that I can have a weird idea, tell a mostly incomprehensible artificial intelligence to create and push it, iterate it a couple times, and then push it for anyone in the world to see - all of this in easily under an hour. We live in the future! If only it didn't also come with an increased chance of early death for all of us.