Decoding the debate over AI psychosis

This week, Box founder Aaron Levie sparked conversation with a social media post suggesting tech CEOs are "uniquely prone to AI psychosis."

In the latest TechCrunch Equity podcast episode, Kirsten Korosec, Sean O’Kane, and I unpacked Levie’s comment. We noted that he doesn’t reject AI tools but insists CEOs must actually use them to truly understand their value.

This skepticism is relatively mild compared to broader backlash—from graduating college students booing AI mentions, to the negative sentiment around tech layoffs, and a noticeable spike in DuckDuckGo installs after Google announced deeper AI integration in search.

Kirsten highlighted Google’s dilemma: "It’s chasing what it feels it must do to keep up, but that disrupts the core brand attribute people value most, without improving it." She also wondered whether "this anti-AI moment could create opportunities for startups or other business areas."

Read on for a preview of our conversation, edited for brevity and clarity.

Anthony Ha: AI is deeply polarizing, which makes discussion challenging. You can feel conflicted because, at the same time, everyone seems to use and love it, yet no one uses it and everyone hates it. Both truths coexist for large groups.

On the user side, one striking thing we already discussed is Google’s search announcements and how AI is becoming a bigger part of the experience—though it’s interesting to see Google try to walk that back a bit, or at least add nuance: if you want that classic 10 blue links experience, there are still ways to get it. It’s not disappearing entirely.

But many people are unenthusiastic about Google’s direction. For example, DuckDuckGo reported a 30% increase in installs—a huge leap. Of course, DuckDuckGo is much smaller than Google. I don’t think Google is in immediate trouble, but that signals a significant audience dislikes the current AI trajectory.

Sean O’Kane: When I look at leading AI labs and tech companies pushing AI features, I notice a convergence toward Anthropic’s approach: really understanding what you want to offer and sticking to it.

Google, in my view, is still pushing in the opposite direction. They’re trying many things, but being so vague doesn’t help them.

When Google takes the stage at IO to discuss how it will change search, much of what they talk about involves shopping or commercial transactions. Yet, collectively, especially for people who have used Google for two or three decades, it’s seen as an information retrieval system.

Google often struggles with that, reacting defensively to fears they might damage the information retrieval side, responding with, "That’ll still be there—let’s focus on helping you book a flight or something."

Then they shoot themselves in the foot by releasing—stress-testing these systems must be tough—but they put out stuff and run into the same problems they’ve faced for years.

Kirsten Korosec: We published a great article about how Google doesn’t know how to spell its own name. Ask it, "How many P’s are in Google?" and it says two.

This tension exists: Google chases what it feels it must do to keep up, but it messes with what people attach to the brand most, without improving it.

I’m wondering—we’ve seen early signs of people voting with their clicks by switching services. Could this anti-AI moment create opportunities for startups or other business areas we haven’t considered?

Anthony: Absolutely. It’s challenging because opinions vary widely. If you build something tailored for an AI-skeptical group, you might alienate more evangelistic users. But that’s the moment we’re living in.

Look at how DuckDuckGo promotes itself—heavily emphasizing an anti-AI stance. I find that striking because I’ve been moving away from Google myself, trying other search engines. A year ago, when I started, even alternative engines experimented with AI features, emphasizing them because they felt they had to.

Now, they see a lane to say, "We’re not interested in that at all. Or if we do it, we keep it in a separate sandbox that won’t affect your core search experience."

Kirsten: We sometimes unfairly categorize all tech CEOs as force-feeding AI. But at least one—Box founder Aaron Levie—has said, "I think there’s a little bit of psychosis among other tech CEOs around AI."

Levie, a frequent Disrupt guest and friend of TechCrunch, said CEOs are uniquely prone to AI psychosis because they’re sufficiently—and I’m quoting—"distant from the last mile of work that still has to happen to generate most value with AI."

I found that fascinating. I wonder if other CEOs agree. And as part of this shift in thinking about what generates the most value, are they also considering how their workforce changes? That’s our other topic today—not just the AI divide, but how AI is transforming work. We’ve certainly seen the negative side: lots of layoffs.

But we’re also seeing major changes in how people work. In your coverage areas, are you seeing evidence of that? I don’t think it’s limited to the so-called "AI startup sector" or big tech companies.

Sean: Most companies I cover work on physical transportation or adjacent areas. Adoption has been slower there than on the software side, unsurprisingly.

We’re starting to see change. We’ve talked about Mind Robotics, the spin-out from Rivian CEO RJ Scaringe. More AI is being applied to physical infrastructure, manufacturing, robotics, and self-driving.

The software side is really transforming things, especially for people whose jobs directly involve writing code.

Anthony: Part of the question involves both AI adoption in companies and AI-driven layoffs—are they top-down or bottom-up?

Many workforce transformations over the last couple of decades have been at least partly bottom-up: tools that people like using, they bring them in, and eventually executives and IT managers accept them.

But a lot of the belief in AI productivity gains seems to be embraced by executives—or, at startups, by VCs funding them—who love the dream of having a tiny team as effective as a much larger one.

That’s not impossible, but Levie’s point is: if you’re not touching the end work, how would you know? He’s not saying throw out AI tools, but that you must actually use them and understand what they do. You can’t just look at a slide and say, "Yes, incredible efficiency, let’s go."

Kirsten: There’s real evidence that companies are using these tools, directly affecting workers through layoffs and how they work. Both truths are accurate here.