Machine Learning Engineer working on agent harnesses, evaluation, ML systems, and tools for understanding model capabilities and studying downstream societal impacts.
I have built / am building:
- WikiArena.org — a benchmark for agentic web navigation
- Wikipedia Solver — shortest-path analysis over the Wikipedia graph
- typeGPT.dev — can you type faster than GPT-2?
- blackhole.wgsl/wasm — interactive ASCII black hole raytracer
I like to write and make ascii animations on my blog.



