Old-gauges: migrating frameworks with AI
Old-gauges was a fun project from a few years ago. Makers, artists, electronics hobbyists and steampunk enthusiasts can create various artwork, contraptions, desktop ornaments, or other IoT devices containing an ESP32 microcontroller.
Old-gauges is a SaaS service that will feed data to the microcontroller, where it can be displayed on gauges, dials, various displays, nixie tubes, electromechanical devices... whatever the device creator would like to use to display the data.
Typical data feeds are AWS CloudWatch metrics, Bitcoin or stock prices, Weather, How long till Elon's Tesla reaches Mars... it's up to the creator. The old-gauges server takes responsibility for collecting the data and passing it down to the [sometimes wacky] devices, and a web interface allows the owner of the device to decide what they would like to be displayed.
With a new MCP server interface, AI can be used to push data to the server, which also gets passed to devices.
The challenge: migrating between incompatible frameworks
The original software was written using the AWS Amplify, Lambda and DynamoDB development framework. It was a messy, iterative, fragmented architecture and development process. In this redevelopment project we wanted to see how well AI could convert from that piecemeal development framework to a more traditional framework like Next.js.
The result? It took quite a few tokens, but Claude didn't miss a beat.
Redeveloped with AI
- Original effort: 3 weeks
- 2 days to 50% complete with AI
- AI suitability: 100%
Technology
AWS Amplify + Lambda + DynamoDB → Next.js + MySQL + MCP server.
AI framework: GitHub Speckit.