Average is Over with AI

Reflections following a visit to an Alpha School

That either made you very good if you had what it took to be good, or you gave up

—Tony Sheridan¹, describing what it was like playing in Hamburg, where the Beatles played in 1960, for seven hours a night.

The Beatles thrived on pressure. Much of the pressure was internal—they put pressure on themselves. They were determined to make the most of their abilities.

I believe that going forward such a combination of a decent amount of natural talent plus much internal pressure is going to produce massive winners in the age of AI. Everyone else will be left behind to some extent.

Tyler Cowen has seen this coming for a long time. I asked Claude to extract some key ideas from Cowen’s 2013 book Average is Over. Claude wrote,

The central thesis is that the economy increasingly rewards people who can work effectively with intelligent machines, while those who cannot are being left behind.

The key factors he identifies include:

Technological complementarity: The crucial dividing line is whether your skills complement computers or compete with them. Those who can leverage machine intelligence—using it to enhance their judgment, creativity, or decision-making—thrive. Those whose jobs involve tasks machines can do better face declining wages and opportunities.

The rise of machine intelligence: Cowen draws heavily on chess as a model, showing how “freestyle” chess teams (humans working with computers) outperform both humans alone and computers alone. He sees this as a template for the broader economy—the winners are those who can collaborate with algorithms.

Conscientiousness and self-motivation: Because working with machines often requires discipline, focus, and the ability to learn continuously, personality traits like conscientiousness become more economically valuable.

Geographic and social sorting: High earners cluster in expensive cities with good amenities, while others move to lower-cost areas. This creates divergent communities with different life prospects.

The erosion of middle-skill jobs: Routine cognitive and manual tasks are automated, hollowing out the middle class and leaving a barbell economy of high-skill, high-pay work alongside low-skill service jobs.

On January 29, I visited the Alpha School in Austin, which is just a few blocks from where I am teaching at UATX. It was the afternoon, and the high school students seemed to be lounging around. Most of them were on computers, and some of them were engaged in conversation, but there was nothing like the structure of an ordinary classroom. Alpha has a vision to radically remake school.

I cannot say that I learned much from the visit, because I already had read a great deal about Alpha. But I was stimulated to think about Alpha, UATX, and Tyler’s analysis.

Alpha comes with a very pronounced “average is over” mindset. There is determination to be outstanding, not just ordinary. If nothing else, that mindset fits well with our current environment.

I am going to assume that you know the gist of the Alpha School story. If not, you can ask an AI to find the most useful links, including this anonymous review on Scott Alexander’s substack.

Also, the essays in Austin Scholar's substack provide her perspective on the school. Most of all, they reveal her extraordinary level of pride in what she has done and ambition in what she believes she can accomplish.

A recent piece by CNN says,

They emphasize learning through AI rather than human teachers. There’s no homework and no textbooks — just software that students use each morning to learn, with human “guides” for motivation and classroom support.

I would like to see UATX be as bravely radical for remaking college as Alpha is for remaking K-12. Personally, I would rather be a guide suggesting topics and recommending readings than a lecturer or a grader.

The CNN story says that in one of the Alpha schools in Brownsville Texas, the attempt to prove that the Alpha model would work for disadvantaged children encountered a few stumbling blocks.

Some students were “so stressed out” and yet the school was “propping them up as, like, the model” and “evidence why Alpha works,” Jessica Lopez, one of the most vocal parents, told CNN this year, when speaking of the experience. She withdrew her two daughters from the school in 2024.

…Alpha has denied any link between its schools and anxiety and noted the Brownsville school has changed from a homeschool-hybrid model that was “confusing and ineffective” and bears “no resemblance to the school model today.”

“The entire school experience, pedagogy and approach to learning have transformed dramatically,” including its apps, goals, workshops and other parts of the school system, the company said. “There is no accurate or fair way” to hold the Brownsville experience from those early years “as representative of the current Alpha student or school experience,” the school said

My theory of the Alpha School is that it selects for students with: (a) the ability to learn from one-on-one tutoring by AI on a computer screen²; and (b) the internal pressure to be high achievers. Students and parents who visit can see what the most enthusiastic students are like, and they can decide how well their personalities suit Alpha.

The AI-tutored learning takes place in the morning. The internal pressure to be a high achiever is provided an outlet in the afternoon, when students work on projects that they have selected for themselves.

My sense is that the near future belongs to the sorts of people who thrive at Alpha. Having the ability to work with AI and the ambition that comes from internal pressure will enable some people to accomplish far more than they could have previously. Those who resist AI and/or fail to set high expectations for themselves will have to settle for lower income and status. The next several years are going to find us living out “average is over.”

1. From the documentary The Compleat Beatles. Sheridan met the Beatles in Hamburg. Their first recording was as a backup band for Sheridan’s recording of “My Bonnie Lies Over the Ocean.”

2. Note that Alpha’s software is “old” in that it predates the emergence of ChatGPT, although I imagine that one could now replicate and improve on Alpha’s software relatively easily using the latest AI models.