I build the rules, evals and tools that let AI coding agents work on real systems. I've worked with these agents since 2024, when they first became useful, and I've watched each new generation change what they can do and how they fail. I worked the rules out on embedded hardware, where a confident wrong answer shows up on the scope.

I'm writing Fail Closed: Governing AI Coding Agents on Real Hardware. Chapter 1 is out now as an early release.

Governance and context

• ai-dev-governance. A versioned framework for working with AI coding agents. It requires a plan before code and evidence before a change counts, and it keeps the agent from grading its own work. An embedded profile covers firmware and bench work.
• astaire. A SQLite knowledge store that hands an agent small, token-budgeted context in place of whole files.

Agent tools

• brontes-probe-mcp. An MCP server that lets an AI assistant operate a debug probe. It flashes, halts, reads memory, and streams ITM/SWO trace on Cortex-M targets.
• zydeco-profiler-mcp. Code-size and cycle measurements for A/B comparisons, judged by a decision rule fixed before the numbers come in.

Inference and evals

• kv4-edge-inference. The reproduction package for my study of 4-bit KV-cache quantization on a 4 GB GPU (paper).

How I got here

I build embedded systems because I'm passionate about the process and seeing the thing become reality from imagination. My adventures include taking boards from schematic through fabrication and bring-up, and I've spent enough overheated evenings with a scope on an SPI or I2C bus to trust the wire over the status line.

When AI coding agents turned up at my bench, I started writing rules for them, then wrote more every time an agent got around the last batch. A heartbeat LED I insisted on once caught an agent testing the wrong firmware. The rules grew into the projects above.

My earlier jobs mostly didn't have titles yet. I ran bare-metal hosting across several points of presence, streamed live concerts from the venue before anyone had heard of YouTube, picked up ISO 9001 discipline at BlackBerry, kept remote facilities running in Northern Canada, and sat on the board of a community nonprofit ISP. I live in Seattle.