A developer has shown how far AI coding agents have come by building a custom watch face for the Pine Time, a $27 smart watch that runs open source firmware. Inspired by a tweet about hacking ESP32 devices with Claude, Mike Kasberg pulled his PineTime out of a drawer and used coding agents to design, build, and iterate on a Casio-style watch face from scratch.
The PineTime turned out to be an ideal target for agentic coding. It is cheap, runs the open source InfiniTime firmware, and is exceptionally well documented. Its InfiniSim simulator gives AI agents a fast feedback loop, letting the model compile, render, and refine the watch face on a desktop before flashing the device. Kasberg noted that while he says Claude, he actually ran much of the work in OpenCode with open-weight models including Kimi K3 and DeepSeek v4 Pro.
The first attempt was rough, with text overlapping and mispositioned, but it produced a solid starting point that the agents iterated on quickly. The story resonates because it highlights two trends founders should watch: the rising quality of open-weight models, and the growing ability of AI agents to work across hardware and software boundaries. What once required deep embedded experience is now within reach of any founder with an idea and a $27 watch.