Hello there,

If you’re an avid reader of blogs, you’ll know this medium is basically dead now. Everyone switched to making YouTube videos, complete with cuts and costume changes every few seconds because, I guess, our brains work much faster now.

The YouTube recommendation algorithm, problematic as it is, does turn up some interesting stuff, such this video entitled “Why Work is Starting to Look Medieval” :

It is 15 minutes long, but it does include lots of short snippets and some snipping scissors, so maybe you’ll find it a fun 15 minutes. The key point, I guess, is that before we were wage slaves we used to be craftspeople, more deeply connected to our work and with a sense of purpose. The industrial revolution marked a shift from cottage industry, where craftspeople worked with their own tools in their own house or workshop, to modern capitalism where the owners of the tools are the 1%, and the rest of us are reduced to selling our labour at whatever is the going rate.

Then she posits that, since the invention of the personal computer, influencers and independent content creators have begun to transcend the structures of 20th century capitalism, and are returning to a more traditional relationship with work. Hence, perhaps, why nearly everyone under 18 wants to be a YouTuber. Maybe that’s a stretch.

This message resonated with me after 20 years in the open source software world, and hopefully you can see the link. Software development is a craft. And the Free Software movement has always been in tacit opposition to capitalism, with its implied message that anyone working on a computer should have some ownership of the software tools we use: let me use it, let me improve it, and let me share it.

I’ve read many many takes on AI-generated code this year, and its really only March. I’m guilty one of these myself: AI Predictions for 2026 , in which I made a link between endless immersion in LLM-driven coding and more traditional drug addictions that has now been corroborated by Steve Yegge himself. See his update “The AI Vampire” (which is also something of a critique of capitalism).

I’ve read several takes that the Free Software movement has won now because it is much easier to understand, share and modify programs than ever before. See, for example, this one from Bruce Perens on Linquedin : “The advent of AI and its capability to create software quickly, with human guidance, means that we can probably have almost anything we want as Free Software.” .

I’ve also seen takes that, in fact, the capitalism has won. Such as the (fictional) MALUSCorp : “Our proprietary AI robots independently recreate any open source project from scratch. The result? Legally distinct code with corporate-friendly licensing. No attribution. No copyleft. No problems.”.

One take I haven’t seen is what this means for people who love the craft of building software. Software is a craft, and our tools are the operating system and the compiler. Programmers working on open source, where code serves as reference material and can live in the open for decades, will show much more pride in their code than programmers in academia and industry, whose prototypes or products just need to get the job done. The programmer is a craftsperson, just like the seamstress, the luthier and the blacksmith. But unlike clothes, guitars and horseshoes, the stuff we build is intangible. Perhaps as a result, society sees us less like craftspeople and more like weird, unpopular wizards.

I’ve spent a lot of my career building and testing open source operating systems, as you can see from these 30 different blog posts , which include the blockbuster “Some CMake Tips” , the satisfying “Tracker Meson” , and or largely obsolete “How BuildStream uses OSTree” .

It’s really not that I have some deep-seated desire to rewrite all of the world’s Makefiles. My interest in operating systems and build tools has always came from a desire to democratize these here computers. To free us from being locked into fixed ways of working designed by Apple, Google, Microsoft. Open source tools are great, yes, but I’m more interested in whether someone can access the full power of their computer without needing a university education. This is why I’ve found GNOME interesting over the years: it’s accessible to non-wizards, and the code is right there in the open, for anyone to change. That said, I’ve always wished we GNOME focus more on customizability , and I don’t mean adding more preferences. Look, here’s me in 2009 discovering Nix for the first time and jumping straight to this: “So Nix could give us beautiful support for testing and hacking on bits of GNOME” .

So what happened? Plenty has changed, but I feel that hacking on bits of GNOME hasn’t become meaningfully easier in the intervening 17 years. And perhaps we can put that largely down to the tech industry’s relentless drive to sell us new computers, and our own hunger to do everything faster and better. In the 1980s, an operating system could reasonably get away with running only one program at a time. In the 1990s, you had multitasking but there was still just the one CPU, at least in my PC. I don’t think there was any point in the 2000s when I owned a GPU. In the 2010s, my monitor was small enough that I never worried about fractional scaling. And so on. For every one person working to simplify, there are a hundred more paid to innovate. Nobody gets promoted for simplicity .

I can see a steadily growing interest in tech from people who aren’t necessarily interested in programming. If you’re not tired of videos yet, here’s a harp player discussing the firmware of a digital guitar pedal (cleverly titled “What pedal makers don’t want you to see” ). Here’s another musician discussing STM32 chips and mesh networks under the title “Gadgets For People Who Don’t Trust The Government” . This one does not have costume changes every few seconds.

So we’re at an inflection point.

The billions pumped into the AI bubble come from a desire by rich men to take back control of computing. It’s a feature, not a bug, that you can’t run ChatGPT on a consumer GPU, and that AI companies need absolutely all of the DRAM . They could spend that money on a programme like Outreachy , supporting people to learn and understand today’s software tools … but you don’t consolidate power through education. (The book Careless People, which I recommend last year , will show you how much tech CEOs crave raw power).

In another sense, AI models are a new kind of operating system, exposing the capabilities of a GPU in a radical new interface. The computer now contains a facility that can translate instructions in your native language into any well-known programming language. (Just don’t ask it to generate Whitespace ). By now you must know someone non-technical who has nevertheless automated parts of their job away by prompting ChatGPT to generate Excel macros. This is the future we were aiming for, guys!

I’m no longer sure if the craft I care about is writing software, or getting computers to do things, or both. And I’m really not sure what this craft is going to look like in 10 or 20 years. What topics will be universally understood, what work will be open to individual craftspeople, and what tools will be available only to states and mega-corporations? Will basic computer tools be universally available and understood, like knives and saucepans in a kitchen? Will they require small scale investment and training, like a microbrewery? Or will the whole world come to depend on a few enourmous facilities in China?

And most importantly, will I be able to share my passion for software without feeling like a weird, unpopular wizard any time soon?