./blog — notes from the build

Notes on cognitive integrity.

What we learn building one governed brain for long-running agents — validity, drift, proof and the failures nobody logs. Written by a human, from production.

agentbrain notes12 Jun 2026·8 min·DE

Warum wir AgentBrain Core mit öffentlichen Research Papers in die Beta bringen

Nicht mit „Wir haben alles gelöst“-Versprechen, sondern mit einem transparenten Research-Footprint: drei öffentliche Working Papers auf SSRN — LongMemEval-Benchmarks, Head-to-Head-Evaluation und angewandtes Property-Management-Research. read
agentbrain notes02 Jun 2026·9 min·EN

Memory without validity is confident lying

AI agents don't just need more memory. They need to know what is still true. On the hidden failure mode where an agent remembers the wrong thing — and delivers it with perfect confidence. read