It started, as many bad ideas do, with a laptop that wouldn’t let me do what I wanted with it. The machine in question: a low-spec Chromebook, running ChromeOS, and utterly convinced that it knew better than I did. It wasn’t wrong about a lot of things. But it was wrong about this.

I wanted to install EndeavourOS, an Arch-based Linux distribution and set up a clean desktop environment on hardware that most people would’ve already binned. What followed was one of the stranger afternoons I’ve had in front of a computer, and also one of the more satisfying ones.

The write protection problem

Here’s something Chromebooks don’t advertise on the box: they come with firmware write protection enabled by default. Google does this for security reasons, and fair enough, for most users, it makes the device harder to brick. But it also means that if you want to replace the firmware entirely, which you do if you’re installing a full Linux distro, you first have to tell the laptop to stop protecting itself.

On newer models, this is done via a software flag in developer mode. On my machine, older, cheaper, less cooperative, the write protection was enforced in hardware. Which meant I had to open the laptop, locate the write-protect screw or battery connector, and physically disconnect it.

I undid all the screw and pried off a back panel that had clearly never been opened before, and unplugged the battery. It felt absurdly ceremonial. Like cutting a ribbon, except the ribbon was a small white connector and the ceremony was taking place on my kitchen table at 11pm.

Where the AI came in

I’ll be honest: I went into this knowing roughly what I wanted to do and roughly how to do it, and not much more than that. The actual step-by-step, which firmware replacement tool to use, what flags to set in developer mode, which kernel parameters to pass, why my trackpad wasn’t working after the first boot that’s where I leaned on Claude.

What surprised me wasn’t that it could answer the questions. It’s that it could answer them one at a time, in sequence, as new problems surfaced. It didn’t give me a ten-page guide upfront and leave me to figure out where I was in it. It let the conversation be iterative more like pair programming with someone patient than reading documentation alone.

The process went roughly like this: first, enable developer mode, which involves a factory reset, so save anything you need. Then disconnect the battery to disable hardware write protection and power the board via the charger alone. From there, open the crosh shell and run MrChromebox’s firmware utility script to flash the UEFI firmware. Then boot from USB with EndeavourOS and walk through the installer, followed by installing audio and trackpad fixes specific to the Chromebook’s hardware. Then spend too long choosing a pink colour scheme for XFCE and regret nothing.

The AI assistance wasn’t magic. It was more like having someone to ask when the documentation ran out, or when I hit an error message that looked like noise but turned out to be meaningful. At one point I got a kernel panic on first boot, pasted the relevant line to Claude, and got back a clear explanation and a fix within seconds. That kind of turnaround, on a niche hardware problem at midnight, used to require either deep personal expertise or posting to a forum and waiting.

What I actually learned

There’s a version of this story where the AI did it for me and I just followed instructions. That’s not quite what happened. Following instructions well is itself a skill, knowing which instructions to trust, when to deviate, and how to debug when something breaks. The AI scaffolded the learning; it didn’t replace it.

I now understand, concretely, things I only vaguely understood before: how firmware sits between hardware and OS, what a UEFI boot sequence looks like, why Arch-based distros expect you to make more decisions than Ubuntu does. I could explain to someone else how to do this, with caveats, on similar hardware. That feels like knowledge, not just a completed task.

A note on independence

ChromeOS is fine. For most people it’s genuinely good, fast (sometimes), secure, updated automatically, never gets in the way. But it also means running on Google’s terms. The default assumption is that your computing happens in a browser, your files live in Drive, and the device is a window rather than a machine.

Linux doesn’t make that assumption. The tradeoff is that you own more of the complexity, but you also own more of the machine. I can install arbitrary software, configure the system at the level of individual config files, and make it look exactly the way I want. Small freedoms, maybe, but they’re mine, and they compound.

There’s something worth saying about the politics of it too. Proprietary operating systems aren’t neutral. They make decisions about what you can and can’t do on your own hardware, usually in the interest of the manufacturer. Learning to work around those constraints where doing so is legal and affects only your own device feels like a skill worth having.

The sustainability angle

The Chromebook I used for this had, in hardware terms, years of life left in it. It was slow on ChromeOS because ChromeOS had outgrown it the overhead of a modern browser-first OS on older internals is real. On a lightweight Linux desktop with XFCE, the same machine runs well. Snappy enough for writing, coding, and daily use.

Without this, it would probably have ended up in a drawer, then a bin, then a landfill, contributing to an e-waste problem that is genuinely bad and getting worse. The carbon footprint of manufacturing a laptop is substantial; extending its useful life by several years is not a trivial act. But I do think there’s a habit of mind worth cultivating: before replacing a device, ask whether the device is actually the problem. Often the answer is the software, and often the software can be changed.

The laptop is on my desk right now, running EndeavourOS, with a wallpaper I like and a terminal emulator I chose and a colour scheme that took embarrassingly long to get right. It’s mine in a way it wasn’t before not because I own it, but because I understand it. That’s the part no one tells you about doing something like this for the first time. The thing you end up with isn’t just a working computer. It’s a slightly different relationship to the idea of what a computer is, and what you’re allowed to do with one.

Worth the kitchen-table surgery. Absolutely worth the kernel panic at midnight.