Userscripts are fun again

Lately I have been writing userscripts again, mostly because LLMs make them much easier to create.

I am not talking about polished browser extensions or big automation systems. I mean small scripts that fix small problems, like a layout that wastes too much space, a button that is always in the way, or a page that breaks when I cast a video to my TV. These sites work fine for most people, but something about them does not fit my browser, my monitor, or the way I use them.

Userscripts have existed for a long time, but making one always took some effort. I had to inspect the page, understand what the app was doing, write the script, test it, and remember the browser API I needed. It was not difficult, but it was enough work that most of the time I did not do it.

LLMs remove a lot of that work. They can write the boring parts of the code, help me find a selector that works, explain enough of a minified page so I can start, and update the script quickly when the site changes.

LLMs have taken a lot from us, so I like finding the small places where they still leave room for creative work. Some things feel worse now, but small personal tools feel more possible than before.

I have a few of these scripts now. Some adjust video pages, some help with language-learning sites, and some are basically CSS with a small wrapper around it. A few of them watch the app state and fix things when the site starts acting strange. They are not products, and they are not worth publishing as extensions.

That is the part I like the most. A personal script can be ugly, work in only one browser, assume my screen size, break next month, and then get deleted, because it only needs to work for me.

If you have not tried this in a while, try it the next time a website annoys you. Start with one small fix, like hiding something, resizing something, remembering a setting, moving a control, or changing a default.