1.10x: Uruguay and the AI Adoption Paradox

Estimated read: 6 minutes. Enough for a coffee.

Back in the 2000s, living in Venezuela, I was in my 8th semester of Computer Engineering taking an elective called Neural Networks. The professor was teaching us perceptrons, backpropagation, and activation functions on a chalkboard. TensorFlow didn't exist. PyTorch didn't exist. ChatGPT didn't exist. "Artificial intelligence" was a theoretical chapter that most of my classmates and I considered irrelevant to the real world at that moment.

I didn't understand half of what they were teaching me. Thanks to Ronald Pérez and Marcel Aponte we managed to put together a decent project and somehow passed that elective, but something stayed with me: the idea that a machine could learn patterns without someone explicitly programming every rule. That seemed to me the closest thing to magic I'd seen in my entire degree. The downside: I never used it again until it went mainstream.

Today I live in Uruguay again. I use Claude, Gemini, and other AI companions every day to ship code, solve SAP architecture problems, orchestrate agents, send emails, plan my training routine for the next marathon, and even ask for recipes when Max and Mía request a dessert. What we now take for granted — waiting a few seconds for a near-perfect answer — used to take days or weeks to get. And this week two data points crossed that forced me to pause, connect the dots, and do a kind of retrospective on all of this.

The Map Nobody Is Reading

Anthropic published its Claude AI usage index by country. It's not a vanity map — it measures adoption intensity adjusted by economically active population. A score of 1x means Claude usage in a country is exactly proportional to the size of its global workforce. Any number above 1 implies accelerated adoption.

Israel leads with 4.90x. The United States marks 3.69x. Singapore 4.19x. Detached as always, understandable if you look at their R&D GDP and everything that implies.

But what matters to me today is this: Uruguay registers 1.10x. One of the few South American countries above 1. We're adopting AI at a higher rate than our economic size would suggest. That's not a coincidence.

For perspective: Brazil, with all its scale, marks 0.70x. Mexico 0.44x. Argentina 0.75x. Chile at 1.04x is close but doesn't quite get there. Uruguay surpasses almost all of them.

The Laboratory

The same week, Microsoft announced the first AI for Good Lab in South America. Where? At LATU, Montevideo. Not in São Paulo. Not in Buenos Aires. Not in Santiago. In Uruguay.

Juan M. Lavista Ferres, Microsoft's Corporate VP and Uruguayan, said it plainly: Uruguay is positioned as the country with the best performance in South America in AI adoption and use. The lab operates in co-governance with the government (MIEM, ANDE, ANII, LATU) with an investment of US$800k annually. For a country of this size, that's a more than considerable bet.

This isn't a consolation prize for a small country. It's a signal: market size stopped being the variable that defines who leads. The variable now is adoption speed and ecosystem quality. Uruguay has 3.5 million inhabitants and is competing head-to-head with countries of 200 million. It's hard not to make the cliché comparison to the national sport: la garra charrúa knows how to push forward and leave everything on the field. And it seems like we're doing it with AI almost without realizing it.

The Paradox

But here's the fine print that sometimes doesn't get read.

Last week I wrote about Éric Sadin, Palantir, and the cognitive desert of the SAP consultant. The central argument revolved around an Anthropic and Stanford paper that demonstrated that programmers who delegated everything to AI dropped their conceptual understanding 17% and their debugging skills were destroyed. AI gave them the output but robbed them of the competence.

See the tension? Uruguay adopts AI faster than any neighbor. But adopting fast without judgment is exactly the cognitive vegetalization Sadin warns about. Being the country that uses Claude most means nothing if the majority copies prompts without understanding what they're executing.

And this is where Microsoft's lab can make a real difference. The AI for Good Lab isn't a commercial arm: it's a philanthropic initiative that carries over 200 projects worldwide, from detecting pancreatic cancer with machine learning to monitoring Amazon deforestation with satellite images, to predictive models for food insecurity and accessibility tools for people with disabilities.

They work with the UN, the Red Cross, Harvard, Stanford, Johns Hopkins. Uruguay's hub will operate by receiving proposals, evaluating whether the problem can be solved with AI, verifying there's data and domain experts, and prioritizing projects by real social impact. If that orientation stays genuine (without becoming another corporate showcase), then it's not just a lab. It's a vaccine against superficial adoption.

The difference between being a passive node of the system and being a real orchestrator isn't in usage volume. It's in cognitive friction. In the effort to understand why the agent made that decision. In auditing the output instead of pasting it blindly.

The Point

When I took Neural Networks, AI was an abstract academic concept. Today it's the tool that defines who is productive and who is replaceable. But the lesson didn't change: those who understand the fundamentals orchestrate. Those who don't are orchestrated.

Uruguay has the opportunity. The data confirms it. Microsoft validated it with real resources. But the opportunity doesn't activate itself. It activates through professionals and people around these projects who know how to use these tools without delegating the thinking.

And in a context where tech companies are leaving the country and some alarmist headlines blame AI, this news should be read as a necessary counterweight. It's one of the few initiatives (perhaps the only one) where you see the government sit down to work with someone real from the industry, without commercial orientation, to solve problems that matter. Health, education, sustainability. Not another pitch deck. Something that generates concrete reasons for professionals who want to leave because the market doesn't retain them to find a different story: that of a country that, instead of watching the train leave, got on board.

Because the reality is this: countries that fall behind in AI adoption, use, and development will see the gap grow every year. It's not a prophecy — it's a mathematical function. And within those countries, professionals who don't understand the fundamentals will end up on the wrong side of the curve. This is more savage than the arrival of the Internet.

The question is no longer whether you'll use AI. It's whether you'll use it as a cognitive wheelchair or as an exoskeleton. One atrophies you. The other amplifies you. But both require you to put in the effort.

That's being Above Average. Not using more AI than average. Using AI better than average. Uruguay already proved it can be above average in adoption. Now it needs to prove it can be above average in judgment.

Shall we begin?

---

Pablo / En Producción