Notes from the team that builds the boring 90%.
Latest from Resources
Your RAG demo is lying to you
A demo answers the questions you tried. Production answers the ones you didn’t.
Why we write evals before prompts
If you can’t measure an answer, you can’t improve it.
The boring 90%: what actually breaks in production AI
Data, retrieval, evals, integrations, guardrails, serving, monitoring and interface.
