The notebook

Notes along the way.

Small ideas with room to grow. Observations on AI, engineering, and the practice of learning — shared while they’re still useful.

5 field notes · Open for exploration

RAG quality is bounded by what the retriever ranks first.

Knowledge graphs give language models a map, not just pages.

No metric, no progress — a principle for every ML project.

Explaining a concept reveals the gaps in your own model of it.

Broad tool access makes agents harder to trust and evaluate.