Research at the intersection of ecological dynamics, information networks, and economic incentives — building systems that sense, coordinate, and adapt alongside living environments.
Making the reward functions and reasoning of autonomous systems legible, verifiable, and safe.
Reward function geometry & the boundaries of aligned behavior.
Why "Number Go Up" breaks AI alignment, and what reward geometry can do about it. Sheaf-theoretic reward spaces, boundary behavior, and what this buys us for detecting alignment faking.
Verifiable generation for AI via ontologically embedded tokens.
Why LLMs hallucinate, and how to build models that can't. Ontological induction, type-safe attention, and what this buys us on the road to auditable super-intelligence.
Borrowing patterns biology already solved for the engineered and emergent systems we build.
Modeling climate, energy, and economic systems — and the civic infrastructure they run through.
Pure mathematics and computer science, sub-venture discoveries that don't fit elsewhere, and what it means to do genuine intellectual work in the AI era.
Latest writing is delivered to subscribers. A subscribe form is coming with the next release — for now, follow Oasis-X for updates.