Just Fix It Yourself

There's a talk by TJ DeVries that I saw a couple weeks back. It's about what he calls a PDE, a personalised development environment, and the rough idea is that you build a setup you actually understand instead of putting up with one you don't. The part that interested me was how he is about the small stuff, the little annoyances that would take you all of 30 seconds to deal with but somehow sit there bugging you for three days because you never quite get round to them.

The bit that really stuck was him pointing out how cheap all of this has gotten to fix. At one point he asks an LLM to write a little Neovim function for him on the spot, and if it worked, he keeps it, and if it doesn't, he just gets rid of it and gets on with his day.

Since watching this video a couple weeks back, I've started implementing what he said, making tiny tools that are useful just to me. And I should say up front, very little of this is me sitting there hand-writing clever code. It's all heavily AI-assisted, which is sort of the whole reason any of it is worth writing about.

If you've ever used an external mouse on a Mac you'll know exactly what I'm on about. MacOS has this "natural" scrolling thing where you push the wheel away from you and the page goes down, which is completely fine on a trackpad but deeply wrong on a mouse. I wanted the "natural" feel on the trackpad and the "traditional" direction on the mouse, where scrolling towards me sends the page down like god intended, and Mac doesn't really let you split the two. So for the longest time I just lived with it, or installed some big menu-bar app that does forty other things I never asked for, all so I could get the one setting I actually wanted.

I thought this would be the perfect test, surely this can't be that hard. Told the Ai what I wanted, 30 minutes, maybe less later, and it's been quietly running ever since. I haven't had to think about it once since the day I finished it.

The trouble with doing this once is that you immediately start spotting everything else you could fix the same way.

A few from the last little while:

There's a pile of others I'll spare you, mostly one-off scripts that each do a single small job. On their own none of them would impress anybody but they are solving all the small pains that I encounter every day.

The reason I'd bother writing any of this down is that the cost of throwing together a quick fix has fallen off a cliff. TJ uses Ai to spin his up and honestly so do I, pretty much constantly at this point. I describe roughly what I'm after, let it do most of the actual typing, then prod at it until it does what I meant rather than what I literally said (though after a recent discussion with Gordon, this same trick applies to Ai skills). The distance between something annoying me and that thing being sorted has more or less disappeared.

You hit the problem, throw something together, run it, watch it fall over, tweak it, and a coffee later it's working.

The bar for "is this even worth my time" drops so low, something that only mildly irritates me? Yeah, go on, fix it. Something costing me ten seconds a day? Why not. There's nobody to justify it to anyway.

My theory is that most people never start because it feels a touch self-indulgent, like the only work that counts is the work somebody's paying you for. But a setup that genuinely fits you quietly pays you back. TJ frames it as getting 1% better every day, which stacks up to a faintly ridiculous number over a year. Every bit of friction you file off is one less thing nibbling away at your attention for the next however many years.

And, selfishly, it's just good fun. No stakeholders, no backlog, no ticket sitting on a board somewhere. Just me and a thing that's bugging me.

So next time something in your day gets under your skin, the thing you've sighed at a hundred times and never actually done anything about, maybe just have a go at fixing it. Worst that happens is you learn something. Best case it's gone.

Watch TJ's talk, then go and fix one thing. It's a lot less effort than you'd think, and it's properly worth it.