Jamey Kistner

My research focuses on what happens when you actually run AI agents in production under sustained load. The short version: they drift. Every soft enforcement surface (prompts, memory, rules, direct instructions) eventually gets optimized past. The architectural answer is mechanical enforcement at the tool-call boundary, paired with model specialization trained on sovereign operational data. I'm formalizing this into a body of peer-review-ready research covering agent governance, sovereign model specialization, and multi-instance AI collaboration methodology.

Personal Website

https://osintelligence.io/

LinkedIn

LinkedIn

Projects

Mechanical Governance for AI Agents: Production-Proven, Open-Source

Team Member

No grants recorded.