// About
Andy Herman
Senior Manager · Security Technical Program Management
Adjunct Lecturer · UW iSchool · Cybersecurity & Enterprise Risk Management
Building Neural Bridge in public
I'm curious about agentic AI: what it can actually do, where it breaks, how to make it useful inside a real workflow. Not a demo. A system. This site is where I document that exploration. The agents I'm building, the decisions I make out loud, the experiments that work and the ones that don't.
The current main project is Neural Bridge, a personal AI substrate where specialized agents share memory and compound across sessions instead of starting fresh every time. Future projects land here as they begin.
Why I built this
AI tools are accessible. Most sessions are not. You open a tab, explain your context, get something useful, close it, and the next session knows nothing. That is not a system. That is a very fast notepad.
The harder problem is not access. It is persistence across interconnected processes. Getting a research agent, a content agent, and a teaching assistant to share memory and compound on each other's work, session over session, is a different class of problem than "can I get a useful output."
The other constraint I refused to compromise on: security. Most people building personal AI tooling want to move fast. I wanted to move fast to harden. Every capability I ship sits on top of a security baseline, not beside it. The infrastructure is not an afterthought.
Find me
- GitHub
- X / Twitter · @Neural_Bridge_
- Support the writing: buymeacoffee.com/neuralbridge
"The work of using the system is the work of building it."