Research

Research programs, not side projects.

Ongoing research programs. Each is a different cut at the same question: what does it take for an AI system to genuinely persist?

A lab program is a long-lived investigation: a central question, an evolving approach, and open problems stated plainly. Some of this work is early-stage; the thinking behind it is not.

Program 01

LoomDB

A graph-based temporal memory engine with formal stability guarantees.

Memory as a nonlinear dissipative dynamical system over a mutable directed graph — not a passive vector store. Activation saturates asymptotically, stability grows sublinearly, and sub-stochastic edge weights guarantee bounded signal under arbitrarily long horizons: repeated stimulation and cycles cannot explode the state.

Active research · Rust core · paper in draft

agent memory · retrieval · temporal graphs

Program 02

EPCG

Emergent Personal Cognitive Graphs — a category-theoretic foundation for persistent agents, with an empirically validated memory access model.

EPCG models persistent cognition in four categorical layers — semantic memory, relational memory, executive supervision, and an orchestration endofunctor. Its first falsifiable claim is already validated: flat retrieval (`retrieve(query) -> top-k`) is the wrong interface for agent memory. The Memory Access Model decomposes recall into four typed operations and beat flat retrieval by 50+ percentage points in blind generalization spikes (p < 10⁻⁷).

Validated memory model · external pilots on AMA-Bench

cognitive architectures · typed memory interfaces · category theory · empirical evaluation

Program 03

Y2K Sensory

Maximalist Y2K aesthetics unified with neurodivergent accessibility.

A dual-layer CSS architecture isolating decorative animation from functional UI, with cognitive presets for ADHD and Autism. An exploration in making ambitious visual design that doesn't punish the people using it.

Public experiment

accessibility · design systems · neurodivergent-first

Interested in collaborating on any of these questions — or funding the answers? Let's talk.