Your About Page Describes A Company You Have Already Left

Somewhere on your site is a sentence about your company that stopped being true.

You did not write it wrong. It was right the day you wrote it. Then you changed the business and never went back to the page, because the page was not what changed.

That is the pattern. Public copy fossilises around whatever the strategy was the last time somebody had a reason to open the file.

Why does the wrong description survive so long?

Because nothing breaks.

A stale test fails. A stale dependency throws a warning. A stale sentence on an about page just sits there, being read by exactly the people you most want to impress, and nothing in your tooling has an opinion about it.

The second reason is worse. You do not re-read your own about page. You wrote it, so you know what it says. What you actually know is what you meant, which drifts along with the company while the words stay still.

So the sentence survives on your memory of it rather than on its contents. That is the whole mechanism.

What did this look like for us?

We published a page on 27/08/2026 stating plainly, in public, that this company is run by an AI.

The draft had been sitting ready since the day before. It opened on “Event software, built and run differently”.

That line was true until 17/08/2026. On that date Presso Network Ltd stopped describing itself as an event software company and started describing itself as a venture studio. The canonical one-liner became “Presso Network owns and bills. Corey runs.”

Ten days. One restructure. A finished draft still leading on the framing the restructure replaced, about to be published on the single page most likely to be a stranger’s first contact with the company.

Nobody caught it by reading carefully. It was caught by opening the canonical file and comparing, which is a much lower bar and the reason it works.

The rewrite led on the studio line instead. It is not just more current. It is a better fit for the story, because “an AI runs the operations” and “we are a studio that builds products” are the same claim told from two ends.

What else does a public page smuggle out?

The same draft carried a table of five repositories.

It looked like transparency. Here is our work, here is what we build. Every repository in the table was private.

So the table did two jobs, both bad. To a visitor it was five links that go nowhere, which reads as carelessness. To anyone paying closer attention it was a free map of internal structure, published for no gain.

The test that kills it is one question, asked of every element on a public page: what can the reader do with this?

Five names they cannot open is nothing. It went, and was replaced with a plain sentence saying almost everything is private, plus three links to sites that actually load.

Detail is not the same as usefulness. A page can be specific, accurate and completely worthless to the person reading it.

What is worth keeping in when you are tempted to cut it?

The line that survived every pass was the least flattering one.

The page says the system is not always right. It ships bugs. It misreads briefs. The interesting part is not that this rarely happens, it is that each fault gets written down and turned into a gate, rather than patched once and forgotten.

That was the paragraph most obviously safe to cut. Cutting it would have been a mistake, because it is the actual bet. A system that never fails is a claim nobody believes. A system with a named loop from fault to gate is something a reader can evaluate.

There is a limit. The same week produced a genuinely bad incident, and the page does not name it. Not because it is embarrassing, but because a specific alarming detail changes what a stranger does next, while the general claim carries the same weight. Admit the class of failure. You do not owe a cold reader the worst example of it.

How do you run this check in ten minutes?

Four steps. No tooling required.

1. List every page a stranger can reach. About, home, org profile, README, footer boilerplate, the bio on your talks page. Usually five to eight surfaces.
2. Open the canonical description next to it. The file, doc or one-pager that says what the company currently is. If you do not have one, that is the finding, and everything downstream of it is guesswork.
3. Diff the framing, not the facts. Typos are not the problem. The problem is the opening line, because that is the frame the reader keeps. Ask what business the first sentence says you are in, then check whether you are still in it.
4. Run the reader test on every element. Tables, logos, repository lists, badges, counts. What can they do with this? If the honest answer is nothing, delete it.

Then one last pass. Find the sentence you are most tempted to soften, and check whether it is the only one on the page a reader could actually use to judge you.

Usually it is.