ChatGPT in Malta makes sense. It is a consumer tool going to a consumer country. Aging population, English-speaking, EU-regulated, small enough to be a clean pilot — a clean deployment story. I am not mad at it. As a press cycle, it works.
But that is not the question I cannot stop thinking about.
The question is: what if Anthropic chose Albania?
Not as a deployment. As a research hub. The country has the inverse profile of Malta in almost every direction that matters for studying what AI actually does to people.
The structural fact most people miss
Malta's median age is over forty. Albania's is in the mid-thirties, and in Tirana it skews younger still. That difference is not trivia. It is the whole story.
If you hand ChatGPT to a country whose median citizen is over forty, you are running a digital-fluency project. A useful one — but the dependent variable is adoption. Can we get this person to use the tool? Can we close the gap between what they already do and what the tool offers? That is the question Malta is going to answer, and the answer will mostly be "yes, with effort, for the simple things."
If you hand a frontier model to a country whose median citizen is in their mid-thirties, with no legacy enterprise software stack to retrofit, no entrenched incumbents to defend their turf, and no regulator yet drawing lines around what the model is allowed to do or be — you are running an entirely different study. The dependent variable is not adoption. It is formation. What do people become when they grow up alongside this technology rather than retrofitting it onto a career?
These are not the same question. They do not have the same value. And the second one, frankly, is the one we actually need answered.
The Tirana scene nobody is writing about
I move between New York and Tirana. The version of Albania most outsiders carry in their heads is the one they got from a movie or a vague memory of the 1990s. That version is gone. What is there now is something I do not see written up anywhere, and it is one of the more interesting things happening in Europe right now.
I sit with founders in two-room offices in Tirana running AI-native businesses that serve clients in New York, Milan, and Berlin. They do not have an "AI strategy." They do not run "AI transformation programs." They have never read a LinkedIn post about prompt engineering. They opened Claude or ChatGPT one day, started shipping things their clients wanted, and never looked up.
Some of them are running four-person teams whose output, measured in client deliverables, beats forty-person teams in capitals where everyone is still debating the rollout plan. They are not doing this because they read a book about it. They are doing it because nobody told them not to.
This is what unwritten ground looks like. Not absence — absence of script. The Albanian software scene has no playbook to live up to, no incumbent vendors to placate, no IT department telling them what they are allowed to install. When you give an unscripted population a frontier tool, they do not use it the way the brochure says. They use it the way it actually works.
That is the experiment Anthropic should be running. Not "can we deploy this." But "what do humans become when they form with this, not after it."
Why Anthropic, specifically, not OpenAI
OpenAI is, increasingly, a consumer product company. That is not a slight. It is a strategic position. They are very good at it. ChatGPT to Malta fits that company perfectly — a national-scale consumer rollout in a tractable English-speaking jurisdiction.
Anthropic is a different kind of organization. Whatever else you say about them, the question their research keeps returning to is not "how do we get more people to use this" but "what is happening to the people who do." They study how models think, how models fail, how models shape the humans around them. They publish papers about it. They organize their company around it.
If you actually care about the second question — what is happening to the humans — you do not run that study in Malta. You run it in a country where the humans are still in the middle of becoming whoever they are going to be. That is not a knock on Malta. It is just a fact about which experiments answer which questions.
Albania is a research-grade environment for exactly the questions Anthropic publicly says it cares about. Young population. Low ambient noise from incumbents. A diaspora that already moves freely between Tirana and the Western capitals, meaning whatever happens there does not stay there. A language situation — Albanian is small, every educated young person uses English daily, multilingual context-switching is the baseline — that makes it a fascinating case for studying how models work across a person's linguistic life rather than within one.
You could not design a cleaner natural experiment.
What a research hub would actually look like
I do not mean an office. Offices are easy. Anthropic could open one tomorrow if it wanted to and most of the value would be PR.
I mean something closer to what Stanford and MIT do with their longitudinal study cohorts. Partner with two or three Albanian universities. Fund a multi-year study group of, say, a thousand students across computer science, medicine, law, and the humanities. Give them access. Watch what happens — not what they say happens, what actually happens, in their work, their writing, their problem-solving, their sense of what their career is.
Pair it with a small research office in Tirana, three or four people, whose job is to surface what the cohort is learning and publish it. Not internal memos. Real papers. The kind of thing that would actually move the field's understanding of what AI does to a generation that grows up with it.
The cost of this is rounding error for a company that has raised what Anthropic has raised. The intellectual return is enormous, because there is no other country on earth where you can run this experiment under conditions this clean.
The line I keep coming back to
The country is not unregulated because it is behind. It is unregulated because it is unwritten. Those are not the same thing, and treating them as the same is the mistake every consultant deck about emerging markets has made for thirty years.
Unwritten is an advantage if you are trying to learn something. It is only a disadvantage if you are trying to sell something the locals are supposed to already want.
ChatGPT to Malta is a press release.
Anthropic to Albania would be a thesis.
Somebody at Anthropic should read that sentence twice.