Hi, I'm Curt

I'm a researcher at Goodfire AI. Here, I write occasionally about whatever topics interest me, including life design (and making agentic choices in a world where defaults seem increasingly suboptimal), learning, philosophy, and consciousness. I think the decisions we make in the next few years and decades will have a profound effect on the future of sentient life, and this is something that I think about both in terms of my work and in the sociocultural landscape. (All opinions here are my own, not necessarily those of my employer!)

My professional work is currently focused on mechanistic interpretability—the science of reverse-engineering neural networks to understand their computational mechanisms and representations. Before working at Goodfire, I worked at Decode Research on building open-source tools and replications of key mech interp techniques, as well as scientific direction and building features for Neuronpedia. Some of my recent research has been on evaluating SAEs, analyzing the evolution and consistency of LLM circuits, and finding evidence for the linear represenation hypothesis.

In the past, I've done data science, marketing, and even run a small startup trying to make the process of learning Chinese algorithmically optimal. You can see more of my projects on the Projects page.

You can also find me at X and GitHub.