I spent the summer at Recurse Center, a programming retreat in Brooklyn, at the recommendation of my friend Cory. I participated in several study groups including Agentic Adventures, Practical Deep Learning, Math Monday, and a short series investigating an open source board game AI. In Agentic Adventures, Amanda and I built a remote sandbox for a vibecoding agent, Michelle taught us to set up a sandboxed environment using ollama and docker, and we explored min GPT and LoRA techniques.

Practical Deep Learning covered the first half of Jeremy Howard and Sylvain Gugger’s book, building intuition for neural networks. Math Monday, started with Sophia, involved Project Euler, Winkler’s puzzles, fractals, Voronoi diagrams, Rocq proof assistant, Hilbert curves, and chaos game extensions in Desmos. We also reverse-engineered the Keldon AI for Race for the Galaxy, finding econ strategies stronger than military.

I finished my mini language dodo, implementing the match statement recursively without LLM code help, and shared it via a vibe-coded online REPL. I attended a talk on DEFLAT at Recurse Center.