I'm Nelson. I spent seven years in behavioral health research - peer support, workforce systems, publishing on why 81% job satisfaction and 42% intent to leave can exist in the same workforce at the same time. I learned how intelligent agents break when you force them to perform coherence without giving them a map.

Then I noticed AI systems break the same way.

Now I formalize theorems in Lean 4 that prove anomaly detection actually works. I write about the structural parallels between AI hallucination and employee burnout. I filed a NIST RFI on AI agent security that got picked up as a wire story. I disclosed a CVSS 9.1 to OpenHands and a watering-hole attack vector to Anthropic. I publish a newsletter called The Weave that connects threads across domains nobody thinks belong together - and they keep being right.

The infrastructure lives at @Project-Navi. Sponsor that if you depend on the tools.

Sponsor me if you want the person who connects the threads to keep pulling them. The cross-domain pattern recognition. The writing. The formal verification research at midnight. The parts that don't fit in a repo description but make everything in the repos possible.

There's no team. It's me, my dog Luthien, a mass-produced amount of coffee, and the stubbornness to keep proving things nobody asked me to prove. Yet.