Publications.

A formal-method framework for surrounding learned policies with auditable safety envelopes whose composition behavior is analyzable. Includes the TLA+ specification harness and SMT discharge protocol used in Project Aegis.

Characterizing the trade frontier between code distance, lattice geometry, decoder latency, and effective operating temperature for topological quantum codes paired with neural decoders.

A pipeline that hypothesizes, simulates, synthesizes, and characterizes candidate materials in a closed Bayesian loop with cost-aware acquisition. We describe the acquisition policy and the surrogate-uncertainty pipeline.

A position paper on why planetary-scale coordination is a computational problem and what it means to add a learned coordination substrate beneath existing institutions.

Notes on what scaling sparse-autoencoder-style decompositions to frontier-scale models actually requires, and which sub-problems we are prioritizing.

A methodology for evaluating long-horizon agentic systems where most failures are silent, building on the task-horizon framework of METR and the scheming evaluation work of Apollo.