Articles
Long-form notes on interpretability, human–AI collaboration, and the tools built along the way.
-
Why I keep collapsing my multi-agent systems back into one
Multi-agent setups sound advanced. The bill they charge in production is paid in context loss.
-
The AI that refuses to think for you
An open–source anti–deskilling framework built on a decade of human–AI collaboration research from the Cabitza lab.
-
Your local LLM has feelings. You just can’t see them yet.
Emotions live as geometric directions inside a model’s hidden states.
-
What your calibration metric isn’t telling you
A model that looks well–calibrated globally can be broken exactly where it matters.