B.A. in Philosophy, Bennington College. Then film: screenwriting, and production-office work (APOC) on major studio features. In 2023, self-taught AI systems engineering; full-time on it since.
The research program that grew out of that work has two threads. The first is agent context substrates: what structure a tool-using agent's context needs to have—flat, tree, or graph—and what it costs when the prevailing defaults withhold it. That thread is the substrate trilogy and the tools that back it (papers, tools). The second is graded oversight of autonomous systems: the acceptance problem—who or what is entitled to accept an autonomous system's work as done, and under what conditions that authority can be progressively and reversibly delegated. A manuscript on that thread is in preparation, and its position is stated as a research area: compositional alignment—alignment as a property of composition, enforced at the joints between parts rather than presumed inside any component.
The Sanctuary System Research Institute is an independent research effort run as an open commons: the papers are published free, the tools are MIT-licensed, and claims are stated with their receipts—repository links, test counts, dates—or with their honest status when a receipt does not yet exist.