Writing

Writing.

Essays on intelligence, art, and the things I build to find out.

Latest essay · 5 November 2026

Model security

Neural network weights can encode executable algorithms. What should we verify before giving a downloaded model access to our systems?

Art & philosophy

Art & philosophy · Visual essay

On superhuman intelligence, protection, and the freedom to remain yourself.

From the laboratory

Three questions about intelligent systems

Memory architecture · Research proposal

DynaTopo: an experiment proposal for memory that reallocates detail where tasks need it.

Speech synthesis · Recorded experiment

Inside RESONANCE: characters, acoustic features, and the measurements between a falling loss and a convincing voice.

AI safety & philosophy · Proposal & finite model

The Zero Hostility Matrix, protected continuations, and a five-state model of what safety has to preserve.

9 October 2026

Toward provably safe realities for conscious life. The complete research essay, including technical appendices, equations, and references.

September 2026

An opening note about this site and what belongs here.