Notebook

This is a lower barrier to entry place for me to put things, where I don’t worry so much about polish.

Training a Small SAE on a Worm

Training a sparse autoencoder on calcium traces of 189 C. Elegans neurons to interpret behaviors

The Village Elder

My vision of how we interact with AI in the future

Updating on AI Agents as part of my workflow

Reflections in light of METR's downlift study

Why does my reward look like *that*?

Millipede Racing

Squirrel

Eigenvectors of the Graph Laplacian hold Spectral Information

Here are some cool proofs from Math 55

DNA as fixed length codings

Can language models keep a secret