This is where I keep the coding I do while working through AI courses and figuring out how these tools actually work under the hood. Most of it starts the same way: something in the course material doesn't run, I dig into why, fix it, and write down what I learned so the next person isn't stuck on the same thing.
It's a learning lane, not a flagship. The point is an honest record of what I built and what I had to debug — not polished products.
The starter code for the DeepLearning.AI "Claude Code" course doesn't run anymore — the query endpoint just throws a 500. A few reasons: it points at a Claude model Anthropic has retired, the tool-use loop only survives one search before it crashes, and newer models spend hidden "thinking" tokens that blew the tiny reply budget and came back empty. I tracked those down, fixed them, and bundled the course transcripts so it runs the moment you add an API key. It's working on current Claude models up to Sonnet 5 — a future model might need another small update, but the fixes and how to point it at a different model are written up in that repo's README.
More will land here as I work through the courses.
I'm moving from scripting and systems admin into building real apps and working with coding agents. This is where I practice that out in the open: debugging code I didn't write, understanding why agent tool loops break, and leaving the fix behind instead of just quietly getting my own copy working.
Part of my portfolio — the portfolio map has everything else.