Why I open-sourced the tool I built "just for myself"

The story of why I open-sourced DeepSeek Phone Harness, a tool I created for personal use, begins with a late-night decision. At 1 a.m., I was in bed, unable to get up, and needed to send a task to my computer agent. This pure selfishness turned into a useful pattern, a phone remote for a desktop agent. After three reasons prompted me to share it publicly, the third reason surprised me the most.

The first reason was a hard-won knowledge: a bug that caused tasks to hang forever. This bug was concealed in the separate agent-question and approval channels of the DSH agent. It was invisible until discovered, taking two days to uncover. This knowledge should be shared, so others don't have to pay that tuition.

The second reason was the pattern itself: using a phone as a remote for a desktop agent. This pattern is useful, but it currently exists in isolated, platform-specific implementations. By making a zero-dependency, MIT-licensed reference implementation, I lowered the barrier for others. They could clone it, read around 4000 lines, and build their own tool for DeepSeek Harness or any agent runtime with a similar event model.

The third reason, the most surprising, changed how I write code. Open-sourcing the tool forced me to make improvements I wouldn't have otherwise. I removed secrets from history, wrote real docs and onboarding, and structured the code for easy understanding by a stranger. After open-sourcing, the tool improved in one week more than it did in the previous week. This experience convinced me to keep building in public.

My request is simple: nothing, but if you take one thing, please document and share when you fix a nasty bug nobody warned you about. This is the spirit of this project. You can find the repository here: https://github.com/2903077918-lgtm/DeepSeek-phone-harness. Have you open-sourced something you built for yourself? What changed? I'd genuinely like to know.