Tech News
← Back to articles

Building for an audience of one: starting and finishing side projects with AI

read original related products more articles

⚡️ FastTab solves a very specific problem: the "Gallery" view of the built-in task switcher in the Plasma desktop environment is slightly too slow for my liking on X11. Sometimes it will take up to a second to open, which is way too long for a feature that I use constantly.

FastTab is a custom task switcher that is built in Zig, uses OpenGL for rendering, and is designed to run as a daemon so that it can respond to keyboard shortcuts instantly.

I would have never built this without AI

It's very likely that this is a problem with an audience of one: users who are still on X11, like the gallery switcher, have many windows open, and care about performance.

In the past I would have probably just accepted the performance tradeoff and moved on, but we're living in the future now! Why not just ask Claude to build it for me? As long as the vision is good, and the code doesn't suck, it should be good enough, right?

And so I did: with zero experience with Zig or X11 internals, I got a working prototype in a few days, and then iterated on it until I was happy with the results.

AI tools enable you to build things that you would have never built otherwise. When you just want "the thing" rather than the process of building the thing, coding agents can be a game changer.

It all starts with a conversation

I was just irritated enough by the slight delay of the task switcher that I asked Claude for tips on how to make it faster. Sadly, there was no quick fix, I had already zeroed out all the animations and delays in the configuration, and the only alternative would have been to switch to the "list" or "icons-only" view, which I just don't like (looking at a window preview is much more efficient than reading titles for my brain).

Then I just asked: "How difficult would it be to build a task switcher that is specifically for X11 and KDE?" and the conversation pivoted into planning. We went through a few iterations where I steered Claude towards a design that seemed promising.

... continue reading