About

Why UXMachine exists

I build it on my own, with AI agents. It started out of frustration.

If you have ever commissioned a UX audit, you know how it arrives: a long, well-written document with forty points ordered by severity.

Nowhere does it say where that severity comes from, or which parts someone measured and which occurred to them while writing. Both arrive in the same typeface and the same confident tone, and you are left to tell them apart with nothing to go on.

With a team you fix that by arguing. Someone puts the list on the table, you keep three things, and those three get done.

Without a team there is nobody to argue with. The document gets read once, filed, and there it stays. I have a few of them filed away.

The first thing I tried, and why it did not work

The obvious thing: I gave the page to a model and asked for the report.

It came out well. Well enough for me to believe it.

The writing was not the problem. The problem was that there was no way to know what it had looked at. And if I do not know what it looked at, I cannot defend what it says in front of anyone — not in front of you, and not in front of myself three days later.

How I solved it is in Methodology, which is where it belongs. Here is just the consequence:

I ended up building an instrument instead of a writer.

Why I build it this way

AI pushes you to build. What interests me is not doing the same thing faster: it is that one person can now build something that needed a team three years ago.

UXMachine is both things at once.

A product

For people who build alone and have nobody to show their work to. They have the judgment. What they lack is someone to look at their page and point out what to change.

And my pilot project

I am building it this way—one person and AI agents—to see how far that can go. I do not mention it as a merit: I mention it because it explains why the product turned out the way it did. With a team I would probably have built something bigger and less verifiable.

The most interesting thing about these months has not been what AI let me do. It has been what it forced me to verify.

Three things I did not know when I started

Measuring is far harder than having an opinionAn opinion is free and sounds just as good. Measuring forces you to decide what counts, with what limit, and to live with the number when it does not say what you expected.

An instrument that never fails is not measuringIt has happened to me several times: the problem was not in the product, it was in the tool I was checking it with. A tool that always says everything is fine is the most dangerous one of all, because it lets you relax.

Half the value is in abstainingWriting “we do not know how to see this” costs more than filling the gap with something that sounds good. And it is the only thing that makes the rest worth anything.

Who

My name is Albert Garcia Pujadas. I have spent more than thirty years in digital, technology and marketing, and I write at qtorb.com about AI, the digital economy and what is shifting underneath.

This is not the product of a company with a team behind it. It is just this: me, a few AI agents, and a method I try hard not to break.

If you want to argue with anything here—especially if you think I am getting it wrong—I am on LinkedIn and on X.

Try it on any page you want to look at

The fastest way to find out whether this is any use.

Try it with a URL →How we look at a page →