Briefing
Spiess frames the method as a control loop: the specification is the setpoint, the firmware on the chip is the plant, the tests are the sensor and the agent is the controller. The Embedded-AI-Harness repo has four Claude Code skills, one per phase from defining the product to building it, and each requirement in the spec must say how it will be proven.
The testbench is a Raspberry Pi (a Zero 2 W with a USB hub works) that maps each board to a network serial port, drives reset and boot over GPIO and exposes 70 MCP tools. Firmware is built in GitHub Actions, and the loop flashes the CI artifact. The repository dates from January; this entry is for the October video.
Written by the Boards & Bots curator
From the GitHub API
Repository facts
| Repository | sensorsiot/ |
|---|---|
| Stars | 259 |
| License | MIT Permissive |
| Language | Python |
| Last push | 9 Oct 2026 20 d ago |