What is closed-loop intelligence?
Closed-loop intelligence is AI that looks at what happened after it acted, checks the result, and uses that to do better next time.
Read the note ↗04 / NOTESInsights
Ideas for intelligence
that learns by doing.
Short, clear explainers on the ideas behind our work: how AI can act, check itself, and learn.
Why AI that works on its own must check its work
AI can only work on its own if it knows whether its actions worked. Independent checks make results visible, reviewable, and worth learning from.
Read the note ↗Why AI agents need goals, not just prompts
A goal gives an AI agent a lasting objective, a record of progress, and a clear definition of success, even across long tasks.
Read the note ↗FIELD NOTEA note on scope
These notes explain where our research is heading. They describe goals and open questions. They do not claim that AI can already work on its own for long periods.