News
Why our agents classify each turn after they answer
We asked the answering model to file its own paperwork through a mandatory tool, and smaller models showed us the cost. Moving that work to a separate call after the reply fixed it.
engineeringtool-callingstructured-outputgemma4Sovereign AI for public services: running open models on infrastructure you own
Why public institutions run open models like Gemma 4 on their own infrastructure, and how the Bayes Platform keeps citizen data in-house, open-source, and free of lock-in.
sovereigntyopen-sourceHow do you know an AI agent is ready for the public?
A self-reported accuracy rate is not evidence. How review campaigns on the Bayes Platform let the professionals of a public service validate an agent's answers, blind, before it goes live.
evaluationtrustpublic-servicesWhy we test every tool with a mid-range model first
We named our RAG tool poorly, and Gemma 4 told us so. Here's what we learned about designing tool schemas for multi-model platforms.
engineeringtool-callingraggemma4