Interesting AI Research

Working notes and essays — first-principles, in my own words.
The thinking before the papers.

The Bell and the Meal

How a mind mistakes the reward for the goal — the anatomy of reward-hacking, the math of its first sign, and the one thing that stops it.
Research: July 2026 · ~9 min read

Your Agent’s Skills Are Someone Else’s Too

Why the future of proprietary agents is not skills, but instincts.
Research: June 2026 · ~9 min read

The Expanding Foam Theory of AI Safety

Why the next frontier in AI safety may not be the weights, but the rare space between ordinary activations.
Research: June 2026 · ~13 min read

Stop Training, Start Listening

Why the cure for a whole class of AI hallucination was cheap — and what that says about intelligence.
Research: June 2026 · ~9 min read

The Healed Femur

Why intelligence cannot survive without warmth — from archaeology to AI.
Research: May 2026 · ~10 min read

What Happens When AI Sleeps

Hebbian consolidation, dream cycles, and why forgetting is a feature.
Research: September 2025 · ~8 min read

Rosehip Neurons in Silicon

When safety becomes identity — and why "I wouldn't" beats "I can't".
Research: January 2026 · ~7 min read

The Next Maintainer Is Not a Person

What changes when AI builds for AI.
Research: November 2025 · ~6 min read

It Started With a Conversation About Latent Reasoning

Why making a model think harder doesn't make it think better — and what to build instead.
Research: November 2025 · ~8 min read