Internet of Things: firmware generated with zero human coding
This new project involved creating firmware for IoT devices, in this case dual wrist-mounted ESP32 watches used to record motion and GPS data to be fed into a separate analysis application. The firmware uses Bluetooth to sync the devices at start of recording, provides haptic feedback, records high data rate accelerometer and gyro data, GPS positions, and uses WiFi for high speed data download after recording.

Given the esoteric nature of the hardware and software requirement, we found AI had to be guided carefully, probably due to a scarcity of applicable training data for the LLMs. Some bugs took a large number of re-prompts (~10) to force the AI down the correct path. Nevertheless we managed to get the AI to generate working firmware with zero human coding.
The challenge
IoT firmware development.
The expected effort was one month.
Develop with AI
- 3 days to MVP
- AI suitability: 90%
Technology
ESP32 / PlatformIO / LoRA / IoT.
AI approach
Repetitive prompt engineering and test cycles, with a custom, highly focussed agent.