2 July 2026 · 8 min
When a small model beats a big one
On a narrow task, a fine-tuned 3B model can match a frontier API at a fraction of the cost — and never leak a token.
Read articleField notes on the tools we build with.
On a narrow task, a fine-tuned 3B model can match a frontier API at a fraction of the cost — and never leak a token.
Read articleVector search retrieves what looks similar. For multi-hop questions about relationships, that is not enough.
Read articleThe Act is law. Vague policies will not satisfy an auditor. Here is what to have ready.
Read articleFraud and outages do not respect batch windows. Streaming lets you act while acting still matters.
Read articleWiring agents together is easy. Making them reliable, cheap and safe enough to trust is the actual project.
Read articleA generator that memorises its training set reproduces real people. Measurement is not optional.
Read articleA voice system that sounds flawless in English can fall apart in Finnish. The default pipeline is not enough.
Read articleUptime monitoring says the service is fine. It says nothing about whether the answers still are.
Read articleTeams obsess over the vector database and default the embedding model. It is exactly backwards.
Read articleNot every problem needs a custom model. Sometimes an API call is right; sometimes it is a trap.
Read article