Over the years I’ve accumulated a fair amount of robot source code — old projects in zip files, folders with broken .git histories, repos that were once on a long-gone GitLab instance. I’ve been working through these archives recently, using AI tooling to help analyse branches, spot untracked files, and work out what’s already on GitHub vs what needs a new home.

Here’s what’s landed publicly so far:

skittlebot

skittlebot — a robot that uses OpenCV to identify lawn bowl pins by colour and ram them. The oldest commits go back to May 2017, nearly 9 years ago. This one had a messy history: files deleted by a duplicate-remover, genuine new work mixed in with line-ending noise, and a separate local history of 30 commits (including 8 that had never been pushed anywhere). All of that is now on GitHub as the gitlab-history branch, with a clean master kept from the original import. See the skittlebot posts for more on the project.

armBot

armBot — a robot arm mounted on a wheeled base driven by an MD25 motor controller, with history stretching back to September 2014, nearly 12 years. Already public on GitHub but the local archive had diverged. Digging through a reconciliation zip turned up md25.py, an old Python 2 serial-based MD25 motor driver that predated the smbus2 I2C approach — a useful bit of archaeology worth preserving. That’s now on the from_hdd branch. See the armBot posts for more on the hardware.

micropython-music

micropython-music — this one started in 2025 as a teaching session, walking a student through making chip tunes with MicroPython. Once we got going I found myself wanting to explore further, and my own interest was sparked. The code had been sitting untracked in a folder with no commits at all, so it got a proper clean-up, a .gitignore, and is now published for anyone curious about music-making on microcontrollers. See the music posts for related experiments on this site.

More to come as I work through the remaining 42 items in the archive.

The Process

The approach has been to work through each folder in the archive with an AI agent: check if a GitHub repo already exists for it, compare any local branches against the remote, identify untracked files that were never committed, and clean up line-ending issues before pushing. The agent can spot things that are easy to miss manually — like a zip file sitting alongside a repo folder that turns out to contain a file never committed to any branch. It’s not magic; each repo still needs a human decision about what to keep and where it belongs, but it makes the inventory and triage work much faster when you’re staring down 40-odd folders of varying vintage.