Notes on automation, governance, and building systems that maintain themselves.
A working paper introducing GROW — a self-healing governance protocol for autonomous AI agents. Grounded in phase transition theory, dimensional affect models, and workspace context awareness, GROW operationalizes programmatic governance at the infrastructure layer to fill the execution-time gap in the AI governance lifecycle.
Most automation platforms force you into one model: fully manual or fully autonomous. The research says the best approach is somewhere in between — and the infrastructure shouldn't force you to choose.
The first number you hear sets a reference point that everything after is measured against. It's one of the most robust findings in behavioral economics — and it operates whether you notice it or not.
We don't have case studies. We don't have a GitHub portfolio. What we have is three years of private development and a production system that runs our own operations every day.
Loss aversion is one of the most robust findings in behavioral science: pain of loss is roughly 2x more powerful than pleasure of gain. The same principle applies to how you design autonomous systems.