How to Time Travel

The future is already here—it’s just not evenly distributed. William Gibson⁽¹⁾Gibson doesn’t remember when he first said it. The earliest citation Quote Investigator has found is a 1992 San Francisco Examiner article that credits him with “The future has arrived—it’s just not evenly distributed yet.” He said a version of it on NPR’s Talk of the Nation in 1999.

Yesterday I attended part of an event about the future of work. One of the facilitators said two things that I connected for the first time.

First, they kept wishing for a time machine. They’re trying to figure out what’s going to happen to the workforce, and wished they could travel forward and see. To the degree we’re talking about large-scale societal change, that makes sense. As J. C. R. Licklider put it in 1965, people tend to overestimate what can be done in one year and to underestimate what can be done in five or ten.⁽²⁾From Libraries of the Future, where Licklider called it “a modern maxim” and didn’t claim it as his own. The idea is usually attributed to Bill Gates, who wrote in the afterword to the 1996 edition of The Road Ahead that “people often overestimate what will happen in the next two years and underestimate what will happen in ten.” But Licklider’s is the earliest version Quote Investigator has found. It’s been about a year since coding agents emerged and AI models started to get quite intelligent.⁽³⁾I date the emergence of good coding agents to Opus 4.5, which Anthropic released on November 24, 2025, calling it “the best model in the world for coding, agents, and computer use.” Peeking ten years ahead would help avoid our current underestimation.

Next, they quoted Gibson directly. And that made me think. If you want to glimpse the unevenly distributed future, talk to the people who are living in it.

This gap in shared understanding of LLM capability is widening. It’s now less “two groups of people speaking past each other” and more a sharp funnel. Andrej Karpathy, October 11, 2026

Much of what gets presented at events like this is a summary of that uneven distribution. But in any organization there are a few people who are making very good use of these tools to improve their creativity and their productivity. They’re living on the edge of the distribution of the future.

By Karpathy’s napkin math, about six billion people have barely come in contact with an LLM, one to two billion are casual users of free chatbots, and around twenty million use frontier models professionally for math and code.⁽⁴⁾Karpathy’s full post gives the numbers as 6B (~75%), 1–2B (~20%), and 20M (0.2%), and describes what each tier sees. His funnel has a fourth tier: the roughly 5,000 people with internal access to systems the rest of us haven’t seen, who he says are “experiencing the vertigo of the curve going vertical.” They’re living even further into the future, but they’re also even more likely to be too busy, or unable, to discuss what they’re experiencing. That last group is the one we want to hear from. And of that group, it’s likely that only a subset are using these tools well. Those are the people who should be presenting at events like this. Usually, they’re not. I suspect many of them are too busy living in the future.

When you do hear from them, it’s remarkable. I remember meeting Cory Gwin in March at an event organized to discuss AI and computing education. He was on an industry panel. The other members were drawn from the center of the distribution. Cory was not. He was describing multi-agent teams to a room of educators, many of whom had never used a single agent. He gave us a glimpse of the future he was already living in.

Those people are out there. At one of our AI builder meet-ups recently I met Jake Metz, who builds immersive technology at Illinois and is further into the future than I am. It was startling. The things he was thinking about and doing were new to me. He’s further toward that uncertain frontier than I am. And I recently had coffee with John Clark, who teaches architecture at Illinois and uses AI to create and orchestrate his own design tools. His work always blows my mind.

AI won’t take your job—it’s someone using AI who will. Richard Baldwin, World Economic Forum Growth Summit, 2023⁽⁵⁾Baldwin is an economist at the Geneva Graduate Institute. He has since said he isn’t sure he coined the line.

But I don’t think most people actually want a time machine. Because talking with people who are living further into the future can make you uncomfortable. And talking with people who are not yet living in your future can make them uncomfortable.

I feel like I’m at least a few steps into AI’s unevenly distributed future. There is a lot of hesitancy when I try to share what that future is like. People in computing do not want to hear that we’re not going to be interacting with source code. They want to hear that it’s still important for humans to read the code. It’s not. Coding agents are fantastic at reading code, and reading code is an error-prone and inefficient way for humans to validate an artifact. People living a little further into the future are finding better ways. Tell that to people living in the present and they typically don’t like it.

Hearing about the future makes us uncomfortable. It means change, challenge, and competition. Right now a lot of people don’t want to hear from the people who are better at using AI than they are. But if you want to time travel, this is how to do it. Find the people who are living in the unevenly distributed future, and listen to them.

If you’re an educator and you’d like to meet others who are finding the frontier, I built Educators Who Build to help organize meet-ups for faculty using AI to support teaching and learning. It runs the Illinois meet-ups, and you can request a site for your own campus.