In 2022, just before the release of ChatGPT, a group of the world’s best forecasters were told to make a prediction about when an AI would become an expert at math. They were specifically asked what year they thought an AI would win a gold medal at the International Math Olympiad, a famous contest that pits the world’s best young math prodigies against one another to solve six very hard problems. The forecasters estimated that there was a 2.3 percent chance that AI would be able to do this by 2025, and their median guess was 2035. Even AI experts, who tend to be more optimistic about the technology they study, guessed this wouldn’t happen until around 2030.
All the experts far underestimated the progress AI would achieve in just a few short years. In 2025 both Google and OpenAI reached gold-medal status at the Math Olympiad. And even more surprisingly, they won this using Large Language Models alone. The same LLMs that were famously bad at counting the letters in the word “strawberry” could now “determine all real numbers α such that, for every positive integer n, the integer ⌊α⌋ + ⌊2α⌋ + · · · + ⌊nα⌋ is a multiple of n.”
Maybe you knew the answer to that question, but I did not (according to both AI and the answer key, it is “every even number”). So, AI is better than me at math, and likely better than you as well. In fact, within a year after the Math Olympiad gold, LLMs were able to solve math problems that had eluded the world’s best math professors for decades, meaning AIs are better at discovering proofs than almost all humans. This may not bother you if your job or selfworth doesn’t depend on generating mathematical proofs, but AI might soon be good at something you do care about, if it isn’t already.
AI is going to kick off a major technological revolution, one at least at the scale of other major revolutions like the computer or even the steam engine.
That is because the list of tasks that AI can do well is growing. AIs can now regularly outperform doctors in giving diagnoses, advertising professionals at making ads people click on, and politicians at persuading people to change their views. We don’t have to restrict ourselves to professional work either to see the encroaching power of AI. One of my friends, a successful finance professional, has an unexpected side talent: He writes limericks for special occasions. It was his thing. When we had dinner, I expected him to want to talk about AI’s impact on finance. Instead, he spent more time lamenting that everyone thinks his masterpieces were written by ChatGPT. He wasn’t joking. He was genuinely sad. Unlike the math example (I am assuming you are not a Math Olympian), you can judge for yourself whether AI is superhuman at writing limericks. I asked a recent version of Anthropic’s Claude to summarize the chapter so far in a limerick. And because that was too easy, I added two arbitrary constraints: Every word had to begin with a different letter, and the first words of each line had to spell “AINOW.”
AI beat clever forecasters today,
Incredible math gold, experts dismay,
Now limiting jobs
Outpacing human snobs
We question progress rapidly underway.
Check it yourself: AINOW down the left margin, and no repeated first letters across the twenty-one words. It took Claude a couple of seconds. Now, this clever limerick (and it is clever) does not threaten the job of anyone. But it is an illustration of the change coming for us all. Soon, if not today, AI is going to be better than you at some things that matter to you. Sometimes the impact of that may be far less than we think (chess is more popular than ever, even though humans cannot compete with AI), but in many cases, the implications will be profound. AI is going to kick off a major technological revolution, one at least at the scale of other major revolutions like the computer or even the steam engine. Jobs will change, education will change, even the way we relate to each other will change.
I see this as inevitable because the AIs that exist in 2026 are already sufficient to drive significant change. As you will see throughout this book, today’s AI is much more capable than most people think, creating what is called a “capability overhang,” where a lot of AI ability is currently unused. And I believe much stronger AIs are coming. Financial crashes or technological blockers might slow the rate of advancement. Governments might restrict development. But even if that happens, AI is too useful and too ubiquitous to disappear.
This might sound fatalistic. It isn’t. The changes AI brings are inevitable; what we do with them is not. Society will shape how AI is deployed, a topic I’ll return to later. But this book is focused on you — the individual human navigating a world where machines keep getting better. And here I have some hopeful news. I think, even as AI gets better, humans will still have a large role to play.
From Co-Existence: The Next Phase of AI by Ethan Mollick, published by Portfolio, an imprint of Penguin Publishing Group, a division of Penguin Random House, LLC. Copyright (c) 2026 by Ethan Mollick.






