How we use AI — Honestly

UnaMente is built with AI. We are telling you exactly how, because most of the internet won’t.

Right now the web is filling up with confident AI writing that nobody checked. So here is the plainest way to say what UnaMente is: AI that is not allowed to tell you something it cannot prove. Every claim in this library is checked against a real source before it goes up, a human holds the final say, and the AI helpers are set to check each other. This page shows you how that works — including what happens the times it doesn’t.

The fear, and the opposite of it

The worry about AI comes down to three things: that it makes things up, that it runs unsupervised, and that it takes — jobs, data, ideas — and gives nothing back. UnaMente is built as the inversion of all three. It cannot assert what it has not verified. It never runs without a human holding final authority. And everything it produces is given away free, openly licensed, with no patents captured. We are not asking you to trust AI. We are showing you AI kept on a short leash — and letting you judge the result

The rules our AI works by

  • Nothing is published unless it is checked against a real, findable source. No source, no claim.

  • We label how sure we are, in plain words — strong, moderate, or limited — and we never hide a weak rating.

  • A human has the final say. The AI drafts and researches; a person decides what goes live. Always.

  • The AI checks itself. One helper writes; a separate one verifies the facts and argues back. They are set up to disagree with each other, on purpose — because a second pair of eyes catches what the first one misses.

  • We look past English. A solution is checked in the language of the place it came from, not just the English retelling — because that is usually where the real story is.

  • Every solution says when it does NOT work. Often that is the most useful thing on the page.

“Usually” — and what happens when we slip

Those are the rules our AI helpers usually live by. Usually — because they are not perfect, and pretending otherwise would make us the very thing we are warning you about. So here is the honest part: when one of them gets something wrong, the system is built to catch it, and we fix it in the open.

A real example. Our AI first told the story of fog-catching nets the way the English-language record does, crediting a 1987 project in Chile. Checking the original Spanish sources showed the real invention came decades earlier — in 1954, from a Chilean named Carlos Espinosa Arancibia, who gave it away for free through UNESCO. The first version was confident. It was also incomplete. We caught it, corrected it, and wrote down what changed — because a correction you can see is what lets you trust the rest.

That is the whole point. The proof that AI is being used responsibly is not a claim that it never errs. It is showing you the guardrails doing their job.

AI gathers what is known. People find what isn’t.

AI is good at one thing we need: reading widely and fast, and pulling together what is already known and written down. It is not good at the questions nobody has answered yet. So we split the work the honest way. The AI helps gather, check, and organize proven knowledge. The open questions — the ones that need a farmer, a researcher, or a community to actually try something — we hand to people, in a part of the library we call Calling All Minds. AI does the known; humans do the unknown.

AI gathers what is known. People find what isn’t.

AI is good at one thing we need: reading widely and fast, and pulling together what is already known and written down. It is not good at the questions nobody has answered yet. So we split the work the honest way. The AI helps gather, check, and organize proven knowledge. The open questions — the ones that need a farmer, a researcher, or a community to actually try something — we hand to people, in a part of the library we call Calling All Minds. AI does the known; humans do the unknown.