Before the current AI focus, I spent years building developer tools and releasing them as open source. Vue File Agent and Rockplate were the most notable, but there were many smaller ones along the way.
Here’s what the experience taught me — lessons that apply just as much today with AI-generated code.
Solve Your Own Problem First
Vue File Agent started because I needed a file upload component that handled large files gracefully in a Vue app. Every existing library fell short in some way. So I built exactly what I needed.
This is the only reliable path to a useful open-source project. Building for an imaginary audience leads to generic, abandoned repos. Building for yourself means you’ll keep maintaining it because you need it.
Documentation Is the Product
Rockplate is a template engine. It’s simple — the README and a few examples are enough to understand it. But I spent as much time on the documentation as on the code. Clear examples, progressive disclosure, edge cases documented.
The response from users was overwhelmingly about the docs. “I understood it in five minutes.” That’s the goal.
Maintenance Compounds
The first year of Vue File Agent was exciting — issues, PRs, feature requests. The second year was maintenance. The third year was the real test: would I still care when it wasn’t novel?
Open source is a long-term commitment that most people underestimate.
How This Applies to AI
The same principles hold with AI-generated code. The tool generates the implementation fast, but the hard work is still: defining the problem clearly, documenting the solution, maintaining it over time.