I’m an independent founder working on Nexus Spine, a verification-focused system for AI-agent authorization, consent, revocation, and audit evidence.

The core problem I’m focused on is delegated authority under composition: as AI agents call tools, hand off tasks, act for users, or operate across vendors and enterprise systems, local permissions can appear valid while the overall authority chain becomes stale, overbroad, revoked, or unsafe.

My recent work has focused on formal-methods-backed invariants and conformance thinking: STOP / revocation dominance, non-amplifying delegation, fail-closed behavior, freshness checks, deterministic tests, and portable evidence receipts.

I’m especially interested in making agent authorization safety testable rather than rhetorical, starting with public-safe specs, synthetic failure examples, conformance tests, and evidence receipt prototypes.