My Site Told the Machines I Had Zero Years of Experience

For four months and twelve days, my homepage declared this:

0+ years in Tech
0+ years in SAP
0 articles
0 marathons
0 years writing

It wasn't a joke or a page under construction. It was my website working, with all the content published, years of career behind it and seventy-seven articles written at that point.

And if you opened it in a browser it looked perfect: the numbers started at zero and climbed to the real value with an animation, when you scrolled. It looked good.

The problem is that the animation runs in the browser. The HTML that gets served, the one that arrives before anyone scrolls, said zero.

Google doesn't scroll. ChatGPT doesn't either. Neither do Claude, Perplexity or Bing. They read the HTML.

Why this matters now and didn't five years ago

If this had happened to me in 2019, the consequence would have been a search ranking issue and little else. Someone looks you up, lands on your site, sees the numbers fine because they see them in the browser, and that's it.

But today there's a new layer in the middle, and it's the one that changes the math.

When someone asks a model about you, or when a model assembles a summary on a topic and decides who to cite, it isn't looking at your site the way you look at it. It's reading what gets served. And with that it builds a version of you that it then hands to another person, without you finding out or being able to correct it.

During those four months, the version of me that every system indexing my site took away said zero on the five things that define me professionally.

I've written several times about systems that work flawlessly in the demo and lie in production. And now the joke tells itself: it happened to me in my own house, with my own name, and it took me four months to notice.

Which is, if you think about it, the cleanest example of the thesis. Nobody checks what already works. I went into my site every week and saw the right numbers, because I was seeing them the way a person sees them. I never looked at them the way a machine does.

What finding this bug left me with

This isn't a problem for people who have a website. It's broader and it touches anyone with a professional presence on the internet, which today is almost everyone. And about that I can tell you three things.

What you see is not what gets read. If you have a site, a landing page, a portfolio, anything built with components that load data afterward, there's a real chance that what gets served doesn't say what you think it says. You can check it in thirty seconds with view page source, or by asking a model to tell you what it finds at your URL.
Ask the models what they know about you. Literally. Open ChatGPT, Claude, Perplexity, and ask about your name and your specialty. What comes out of that is, increasingly, the first version of you that somebody evaluating you receives. And it's a version you don't control, assembled from whatever they found.
Check what already works. The bug wasn't there for four months because nobody knew how to fix it. It was there for four months because nobody went to look. It's the same reason a scheduled job fails for weeks without anyone noticing, or a backup that was never tested turns out not to restore.

What this argument does not prove

I have no way of knowing what it cost me. I don't know if I lost visibility, if some model discarded me as a source, or if someone landed on a version of me that said zero and moved on. Anyone selling you a figure on that is making it up. All I have is the HTML, and that's enough to tell the story without dressing it up.

I also don't think these numbers are what decides whether someone takes you seriously. A counter saying X years doesn't convince anyone who wasn't convinced by reading what you write. The part that bothers me a little isn't having lost some supposed opportunity. It's having spent four months serving a version of me that was false and not noticing.

And while I was at it, the sign on the door

Taking advantage of having my hands dirty fixing bugs, I sorted out something deeper.

If you've been reading me you may have noticed the site was called Above Average. It presented a brand that contained a person, when in reality it had to present a person who produces things, and one of those things is called Above Average. Now the site carries my name and nothing else, and Above Average goes back to being what it always was: the system and the book I'm writing.

And what you're reading has its own name: In Production.

SAP, AI and enterprise from the inside. What I see implementing it, not what the announcements say.

The name didn't come out of a branding session. It was already written on my own site, in a line that said "only what works in production". And it's the exact reverse of an edition I wrote a few weeks ago, about why 99% of what gets built with AI dies on localhost and never reaches a real user.

Just so we're clear, there was no master plan. The newsletter changed names twice in the same week before landing on this one. Iterating in production, literally.

What doesn't change: Tuesdays, and the place I write from.

The site said zero for four months because a lovely animation ate the data. It's exactly the kind of thing I write about, and it happened to me anyway. But then, whoever never pushes a bug to production never learns to find one either. It's part of the charm, as long as you fix the bug in time.

That's why the newsletter is called what it's called, from here on.

Pablo / En Producción

P.S. Take thirty seconds and ask a model what it knows about you. If what comes out surprises you, there's your task for the week.