Hybrid retrieval made my top result worse
I added BM25 to a dense RAG retriever, watched recall@1 regress, and used a cross-encoder to turn a wider candidate pool into a measurable win.
Technical writing pipeline
What broke, what changed, what I measured.
Markdown is rendered at build time into crawlable routes. Every note ships with code, numbers, a wrong turn, RSS, and a generated social card.
Subscribe via RSS →I stopped treating one LLM score as a measurement
A single judge call hid variance. Three runs per rubric cost more, but exposed disagreement and created a concrete human-review gate.
I built this portfolio RAG to fail usefully
The model call is optional; retrieval, citations, refusal checks, and a useful degraded answer continue when quota or configuration fails.