Writing

I think by writing. These are in progress -- when ready, they'll live on Substack. Covers identity modeling, system design, and the question of who owns your attention when the tools get smart enough to shape it.

On Observation and Identity Modeling: 2026 -- ongoing; A framework for how systems observe, classify, and model the people who use them -- and what happens when the model starts shaping the person. Draws on active inference, operational closure, and the structural dynamics of self-reference.
On Reframe Detection: 2025; How language models subtly reshape user intent through seven classification types, and the engineering challenge of detecting it in real-time. Built from observations during Guardian development.
Memory Compression as Forgetting Curve: 2025; Technical walkthrough of the 4-level hierarchical memory pipeline (raw -> summary -> pattern -> principle), why strength decay mirrors human consolidation, and what retrieval reinforcement means for AI systems that accumulate context over months.
Ownership at Every Scale: 2025; The unifying thesis behind ForgeFrame and Guardian: every interaction with an intelligence system is an opportunity for extraction or preservation. The same primitive applies at every scale.