Nathaniel Ryan Seals

Independent researcher creating open-source infrastructure primarily for agentic-AI safety, publishing as Cid. Author of SIGIL, an MIT-licensed deterministic control layer for LLM agents with cryptographic prompt integrity, tool-permission enforcement, human in the loop gates, tamper-evident audit chains, and portable proof-of-execution receipts. I also wrote several papers, including a whitepaper on bounded, revocable, auditable delegated authority for autonomous agents. My focus is the deterministic layer around probabilistic models: bounding and accounting for what an autonomous agent can actually do. In model alignment cannot be the whole safety boundary.

Personal Website

https://github.com/mr-gl00m

Projects

No grants recorded.