The AI Treadmill - by Deb Liu - Perspectives
Perspectives
SubscribeSign in
The AI Treadmill<br>Why the people most ahead in AI feel the most behind — and what to do about it
Deb Liu<br>Jun 04, 2026
23
Share
A few weeks ago, I sat down to dinner with a group of executives, founders, and investors in San Francisco, hosted by my friends at MediaMint. It was the kind of dinner that feels completely ordinary here and utterly surreal everywhere else.<br>We went around the table to introduce ourselves and share how we were using AI. Someone was building an agent to manage their household. Another person was running three agents in parallel before leaving the house, so there would be results waiting when they got home. One executive was using NotebookLM to generate custom podcasts about AI news so she could stay current during her commute. Someone else was using AI to summarize long documents that, as it turned out, may have been written largely by AI in the first place.<br>There was real excitement in that room. But underneath it, something else was running quietly, like a current.<br>Fear.<br>Not fear that AI was coming. Everyone at that table knew it was already here. It was the fear of falling behind. The fear that every moment you weren’t building, automating, testing, or optimizing was a moment wasted. Walking down the street and noticing the trees had somehow become an opportunity cost.<br>At some point, I looked around and said what I was thinking: “This is very normal for San Francisco. But this is not normal for the rest of the world.”<br>One investor called it AI psychosis. It’s a dramatic phrase, but I understood exactly what he meant. It’s that feeling that every minute not spent learning, building, or automating is a minute permanently lost. And the strange part is that the people most affected are often the ones who are furthest ahead. Everyone in that room was likely in the top fraction of a percent in their understanding of AI, and yet no one felt calm. No one felt like they had enough time.
Subscribe
The candy store problem
My friend George Lee, who has spent years working inside some of the fastest-moving companies in tech, described it perfectly. He said working at a cutting-edge tech company is like being a kid in a candy store. Everything looks exciting, new, and delicious. The danger isn’t that you try something. The danger is that you overdo it.<br>I found myself doing exactly that.<br>I subscribed to a dozen AI newsletters. I checked LinkedIn and Threads constantly to see what was happening. I tried to follow every new trend and learn every new technique. I told myself I was staying current. I told myself this was what staying relevant looked like.<br>But I didn’t feel smarter. I felt more behind than ever. As if knowing more had somehow made me more aware of how much I didn’t know. The ocean kept getting bigger the further I swam into it.<br>That’s the paradox no one talks about. The more you consume, the more there is to consume. The faster the field moves, the faster you have to run just to stay in place.<br>More efficient, more busy
Someone told me recently that AI had made them significantly more efficient, but also significantly more busy. It was almost as if being able to do more created more to do.<br>I haven’t stopped thinking about that.<br>Because it might be the central tension of this moment. AI can help you do more. But doing more doesn’t automatically mean doing it better. When you clear the pile in front of you, you often discover a longer pile hiding behind it. When you remove the friction of starting something, you also remove the natural stopping points that used to make you ask whether it was worth doing at all.<br>Before AI, if a task took ten hours, you had to make a decision. Is this worth ten hours of my life?<br>Now, if it takes ten minutes, you just do it. And then you do ten more things like it.<br>I think about it like a treadmill. You keep adding speed because you can. But a faster treadmill isn’t taking you somewhere different. It’s taking you to the same place, only more breathlessly.<br>We tell ourselves we’re saving time. But sometimes we’re just filling the saved time with more motion.<br>Share Perspectives<br>The world is not your LinkedIn feed
Here’s something worth remembering: most people are not doing any of this.<br>Most people are not spending their evenings setting up their Mac Minis to run OpenClaw. Most people are not using AI to generate podcasts about AI so they can discuss AI at the next event about AI. Most people are not kicking off multiple agents to run tasks while they walk their dog.<br>Most people are working their jobs, raising kids, caring for aging parents, paying bills, managing a household, and occasionally trying to figure out what everyone keeps talking about. For them, AI is something they’ve heard is important but don’t know where to start. It’s a technology that feels both magical and completely alien.<br>ChatGPT launched in November 2022. That’s less than three years ago.<br>Three years...