Neuromorphic computing
Main focusOur main focus: computing that works only when something changes, for motion and sensing hardware that is quiet most of the time. Simulation studies, a review of the field and use cases.
What we study, and how we check our work.
Our main focus: computing that works only when something changes, for motion and sensing hardware that is quiet most of the time. Simulation studies, a review of the field and use cases.
We test our models on our own coding benchmark. A new model replaces the one before it only after it scores better there.
How we keep our models and agents from being used to orchestrate cyber attacks: what an AI may do on its own, what it must ask first, how you stop it, and what the evidence says about the threat.
Live instrument readings in code, and hardware tests on every commit. These become our developer tools; none are out yet.
Tree-compressed long-term memory for AI agents, so an agent keeps its context across very long work without bloat. Recent detail is kept as it is, and older memories are summarised in a tree the agent can search and zoom into.
Our first write-ups are on neuromorphic computing: simulation studies with reproducible results, a review of the field and use cases. On AI safety, we publish how Klyroe's controls against misuse work, checked against our code, with sourced reviews of the field.