There’s an article on The Verge that crystallised something I’d been feeling but hadn’t quite found the words for. They call it the personal software revolution. The idea is simple: AI has made it possible for anyone to build software that fits them exactly — not software built for a million users, watered down to offend no one, wrapped in a subscription you half-use.
Reading it felt like someone putting a frame around a picture I’d already been looking at.
Breaking the subscription trap
The subscription trap is real. Every tool you pay for was designed for someone slightly different from you. It has features you’ll never touch and is missing the one thing you actually need. The people who built it made reasonable decisions for the median user. You are not the median user.
For years, the only way out was to hire a developer or learn to code. Most people did neither. They adapted to the software instead of the other way around.
That’s changing.
Imagination (finally) meets code
My team at TalkXO works inside Adobe Creative Cloud every day. Designers, video editors, copywriters — all of them deep inside tools that are powerful but rigid. Over the last year, they’ve started building their own plugins. Small things. A batch-rename script. An auto-export workflow. A custom panel that fits how we actually work, not how Adobe imagined we might.
None of them are engineers. They described what they needed. AI wrote the code. They tested it, broke it, asked again, and shipped it.
We’ve also quietly stopped paying for a few SaaS tools we were using. Not because we were cheap — because the tools weren’t quite right, and building the right version took an afternoon. A custom reporting dashboard. An internal content calendar that matched our workflow. A client-facing tracker that did exactly what we needed and nothing else.
When you can build your own version, the calculus changes. You stop tolerating.
Building for a “market of one”
(Added May 30) A friend of mine is preparing for the GRE. Instead of picking from the usual prep apps — all built for the average test-taker, all with the same flashcard loops and practice test formats — she built her own inside Lovable. An app that matches exactly how she studies: interactive, session-based, structured around her weak areas. The polish on it is surprisingly good. Better than most tools in that category, honestly, because it isn’t trying to serve a thousand different learners. It’s serving one. No tool on the market right now can replicate that — not because they lack the technology, but because they were never trying to solve her specific problem.
Giving old hardware new purpose
The most personal example I have is my old Intel MacBook Pro.
It was sitting idle. Underpowered for modern work, too old to get the good updates. The obvious move was to sell it or forget it.
Instead, I turned it into a home media server. And I didn’t do it from a tutorial — I built exactly what I wanted. Motrix as the downloader, because that’s what I prefer. WebDAV for access across devices, because that’s the protocol that made sense for my setup. A clean UI that runs on my TV, ad-free, no clutter.
The whole thing was written with Claude Opus 4.7. I’d describe what I wanted, it would generate the configuration, I’d test it, refine it, push further. It took a weekend. The result is something no product on the market would have given me — because no product on the market was trying to solve exactly my problem in exactly my context.
That’s the thing about personal software. The specificity is the point.
Fit to fit
Last year I wrote about switching from TeuxDeux to Apple Reminders. The move was less about one app being better and more about fit — offline-first, baked into the OS, one less subscription. That’s a small version of the same instinct. You start asking whether the tool was made for you, or whether you’ve been made to fit the tool.
The answer, most of the time, is the second one.
A quiet revolution
The Verge article frames this as a revolution, and I think that’s right — but it’s a quiet one. It doesn’t look like a launch or a movement. It looks like a designer building a plugin on a Tuesday afternoon. It looks like someone not renewing a subscription because they spent a weekend building a better version. It looks like an old laptop becoming a server that does precisely what one person needs.
The word that keeps coming to mind is fit. Software that fits. Not close enough, not mostly right — actually fits.
For most of computing history, fit was a luxury. You got fit if you were a developer, or if you could pay one. Everyone else got software designed for someone else and learned to live with the gaps.
The gap is closing. The cost of fit is dropping. And for people willing to spend an afternoon describing what they want instead of browsing what exists — the options are genuinely different now.
Taste and clarity
I’m not making a prediction about the future of software development or the fate of SaaS companies. I’m just describing what’s already true for me and the people I work with.
We are building things. Small things, personal things, specific things. Things that wouldn’t exist if we had to justify the engineering cost.
That used to require a skill set. Now it requires clarity about what you actually want.
Turns out that’s the harder part anyway.
For what it’s worth — I have nothing against vibe coding. In the hands of someone technical, someone who owns the architecture, the stack decisions, the structure of the thing — it’s a genuine superpower. The AI handles the tedium, you handle the thinking. For non-technical people, it may not always produce production-grade software, but that almost misses the point. It creates an entirely new category: personal software developers. People who will never ship a SaaS product but can build exactly what they need for themselves. That matters. The caveat is honest though — AI still gets lost. Models are genuinely bad at debugging, and anything with real complexity will eventually land back in your lap. The home server I built hit several walls that no amount of prompting could get through cleanly. You work through it. But that’s a limitation of today, not a verdict on the idea.