Learning in the open
A podcast of what I'm figuring out right now, one idea at a time.
An applied-AI learning log, out loud. Each episode is me working through one idea I want to really own: retrieval, agents, evaluation, fine-tuning, safety. Two voices, plain language, no hand-waving.
Ep 4New
How do you know it's any good?
Evaluation
The least glamorous idea in applied AI, and the one that separates a demo from something you can actually trust. Offline evals, online evals, model-as-judge, and why an eval set is really just acceptance criteria you can run on every commit.
2026-11-21
On the way
Ep 2Coming soon
Answering from real state
Retrieval
How retrieval actually works, and where it quietly breaks.
Ep 3Coming soon
When the model does more than answer
Agents
Tool use, the agent loop, and keeping a second model honest.