You Should Really Build a Thing Created: Thu 26 Mar 2026 · Updated: Sun 13 Sep 2026

Last year I published Build a Thing - a short rambling about building software for oneself without external accountability. It brings joy and gives room for exploration into things that would never come up in the day job.

Well, shit got real since then. Yes, I mean the AI bogeyman coming for all our jobs.

Now, more than ever, I implore you to build a thing. Something pointless, something that fixes a specific problem, something that looks good. Use AI, use multiple AIs, shove agents in a loop and see what they come up with (spoiler: lots of spaghetti and architectural inconsistency). It really doesn’t matter - the key is exposure and experiment. AIs might stagnate, they might be 100x more powerful or 100x more expensive next week - you need to keep up to date with it. Building something for only yourself is the perfect environment to do that.

You want to know why this matters so much? A family member has just finished computer science at university. When he started everyone wanted to be in software and everyone needed software engineers. He now comes out of university with a small mountain of debt into the worst possible graduate market in a long time.

At the very same time the senior end of the market is hotting up as it’s becoming clear AI is not effective without some greybeard oversight. It sucks, it’s unfair, blame the boomers! I hear you, but that’s irrelevant - that’s just not how the world works. We could talk about it all day but it doesn’t achieve anything. Get out there. Build shit, throw it out, repeat.

I can’t say it enough: practice, practice, practice! The ones who can effectively wield AI and grasp code/systems will land the jobs that remain.