What if the Emirates did not need to build their own Stanford?

Ben Horowitz recently described a conversation with Dan Boneh, a professor of computer science and cryptography at Stanford, about his students' use of AI. Boneh told him that he essentially had two options: ban AI, which would not work anyway, or make the problems so difficult that his students could no longer solve them without it. He chose the second option and now has students who can solve problems that no previous generation of students could have solved at their age.

The reversal is brilliant because for the past few years, we have spent a lot of time wondering how to adapt education to AI, but we often ask the question backwards. We look for ways to keep students doing the same things as before, now that they have a tool that can do them much faster. We ban ChatGPT during certain exams, try to detect its use in assignments, and change the rules to preserve the exercise. But if a student can do in ten minutes what used to take three hours, we can also simply give them something much harder to do. That is what we do everywhere else: when computers become more powerful, we do not ask them to keep doing the same calculations while politely waiting as long as they would have taken in 1995. We increase the size of the problems.

Obviously, this does not mean there is nothing left to learn. It is actually quite easy to become very stupid with a very intelligent machine. The more the tool can produce on your behalf, the more important it becomes to know whether what it produces is correct, interesting, or completely wrong. But it changes enormously what you can ask of an eighteen-year-old. A student can now write code without being an excellent developer, analyze huge amounts of data, produce images, translate texts, quickly learn whatever they are missing to move forward, and go from an idea to something that works in the same day. Not so long ago, some of those things would have required several people or skills that took years to acquire. And yet a large part of our education system is still organized around a fairly old idea of what a student can do.

This is the context in which Ben Horowitz, Gagan Biyani, and Erik Torenberg are launching the Horowitz Andreessen Academy in San Francisco. The first class is due to arrive in September 2027. a16z incubated the project with $35 million at launch, alongside Anthropic, NVIDIA, OpenAI, Palantir, Google, Meta, Stripe, Coinbase, Replit, Anduril, and other companies that are expected to contribute to the teaching, provide access to their tools, and host students. But the idea is not simply to create a university with more AI. They are instead taking apart something we have grown used to considering indivisible.

A university is a fairly strange thing. We ask it to teach you things, certify that you have learned them, give you a network, introduce you to people your age, help you find a job, give you time to work out what you want to do with your life, and finally produce a degree that someone who does not know you will use to decide whether to hire you. We ended up putting all those functions into the same four-year product. HAA keeps some of them but tries to assemble them differently: students live together in San Francisco, learn from people who practice what they teach, work with the tools available today, and spend a substantial part of their time on projects and inside companies.

This changes, in particular, the boundary between knowing something and having already tried to do it. You can learn an enormous amount about starting a company without ever trying to sell a product to someone who does not want it. You can study management and then discover that four people not doing what you expected of them are a rather different subject. You can study negotiation for weeks before finding yourself, for the first time, facing someone who can simply say no and leave. AI dramatically reduces the cost of those attempts: at eighteen, you can now try to build something without a team of ten, have an idea in the morning, make a first version during the day, and discover that evening that nobody wants it. You have failed, but you have learned something that might previously have taken several weeks and much more money to learn.

That does not mean turning every eighteen-year-old into a startup founder and treating a book with suspicion if nothing has been produced by the end. There is immense value in learning things that serve no immediate purpose. A university also gives you the right to spend six months becoming obsessed with the Byzantine Empire, Persian poetry, or a mathematical problem whose usefulness nobody around you understands. Part of education consists precisely in not yet knowing what the things you learn will be useful for, and it would be rather sad to replace four years of study with four years of hackathons. But between that and asking a twenty-year-old to write, for the fiftieth time, an essay a machine can now produce in thirty seconds, there are an enormous number of things to try.

This is what makes HAA interesting as an experiment, rather than yet another attempt to "disrupt" the university. Nobody yet knows whether the model works, since the first student has not set foot there. They are simply taking several functions historically brought together in the same institution and trying to see what happens when time is distributed differently between learning, building, working, and living with other students. And the experiment becomes even more interesting once you take it out of San Francisco.

I live in the Emirates, and education is in a rather particular situation here. Historically, for many families who want their children to attend the best universities, that has also meant watching them leave at eighteen. London, the United States, sometimes elsewhere. It is heartbreaking for families, and the government has long sought to attract world-class universities and develop institutions that can ensure academic excellence does not necessarily mean leaving the country. Dubai now has dozens of private higher-education institutions, the vast majority of them international campuses. The effort to bring the world's universities to the Emirates is already well underway.

The logical next step would be to keep going: bring in better universities, build better campuses, attract better professors, and progress until an Emirati student can find at home an institution comparable to those they currently leave for. Except there is a rather amusing advantage to developing some infrastructure later than everyone else: sometimes you can skip an entire generation. Countries that had few landlines did not necessarily need to reproduce the entire European telephone network before moving to mobile phones. Certain banking practices developed directly on phones in places where the traditional banking network was much less extensive. Being behind is sometimes an opportunity to avoid building what others are already replacing.

The Emirates have that kind of window with universities. The government wants world-class institutions, there is capital, there are companies, entrepreneurs, a huge variety of nationalities, and a rather distinctive ability to bring people in from abroad quickly when the country decides a subject matters. They could spend the next twenty years trying to build universities that increasingly resemble the best American universities, and that would already be an achievement. But it would be a shame to devote twenty years solely to catching up with a model at the very moment when those who have its best versions are themselves starting to experiment with something else.

And a school of this kind in the Emirates would have no reason to be exclusively Emirati. You can put someone from Abu Dhabi, Mumbai, Beirut, Paris, Lagos, or Shanghai in the same class and have them live in a place where those worlds already cross paths every day. We could continue attracting excellent traditional universities while building much smaller, experimental institutions alongside them, designed from the outset around the tools students have today and problems difficult enough to make them use those tools.

People often talk about building the Stanford of the Middle East. I think that is aiming too low. The really interesting thing would be for a young Emirati admitted to Stanford to hesitate about leaving.

We will see which majlis I can tell that to…