Technical deep-dives on search, embeddings, RAG, and retrieval systems.
When we switched from OpenAI embeddings to custom fine-tuned models, retrieval quality jumped 35%.
January 5, 2026