Blog
Experiments and lessons from building AI products, written as I learn them.
2026 September
- I taught a free, local model to make my AI's decisions as well as a paid API
September 30, 2026 — AI, Fine-tuning, Decision Models
Distilling Jev into Laya: what fine-tuning is, how I did it without a cluster, and the parts that didn't work.
- Where a decision model earns its place in an AI product (and where a regex still wins)
September 29, 2026 — AI, Decision Models, Product Engineering
I tested Jev at seven decision points in my assistant. Here's the four-question test I ended up with.
- Jev lost to my reranker. Then it did something a reranker can't.
September 28, 2026 — AI, RAG, Decision Models
I tested a new kind of AI model in my document search to decide whether to ship it. It lost at the job I bought it for, and won at one I didn't know I had.
2026 August
- Cognitive Debt: The Bill Nobody Is Tracking
August 10, 2026 — AI, Engineering Practice, Developer Experience
AI writes the code faster than anyone can understand it. The research on what that costs — and what teams are starting to do about it.
2026 July
- BM25: Complete Reference Summary
July 27, 2026 — Search, Information Retrieval, NLP
History, people, mechanics, and significance of BM25 — the ranking formula behind lexical search, from Okapi and TREC to today's hybrid search systems.
- Word2Vec: Complete Reference Summary
July 27, 2026 — NLP, Embeddings, Machine Learning
History, people, mechanics, and significance of Word2Vec — the method that turned words into geometry, and the direct ancestor of today's semantic search embeddings.