Box founder Aaron Levie stirred a debate after suggesting tech CEOs are uniquely prone to “AI psychosis.” On the Equity podcast, guests connect that skepticism to broader frustration with AI features—visible in student booing, anxiety around layoffs, and a rep
When the conversation turned toward Aaron Levie’s latest jab at the tech world, it didn’t land like a technical critique. It landed like a diagnosis.
Levie—Box founder—had posted on social media suggesting that tech CEOs are “uniquely prone to AI psychosis.” On the latest episode of TechCrunch’s Equity podcast. Kirsten Korosec. Sean O’Kane. and Anthony Ha tried to make sense of what. exactly. he meant and why it’s catching on at a moment when public sentiment toward AI is splitting down the middle.
The panelists didn’t treat the idea as an outright rejection of AI tools. They focused instead on Levie’s core argument that CEOs need to actually use these tools to understand them.
That feels almost gentle compared to the pushback surfacing elsewhere. Graduating college students have reportedly booed whenever AI is mentioned. There’s also been bad blood around tech industry layoffs. And one more data point has been hard to ignore: DuckDuckGo said its installs are up 30% after Google’s announcement that it will bring more AI to the search experience.
Ha put it in plain terms: people seem to be living inside a contradiction—everyone’s using AI and everyone’s loving it, yet also no one’s using it and everyone’s hating it. He pointed to the search shake-up as a place where the backlash is showing up in real behavior.
O’Kane framed the problem through a different lens. watching leading AI labs and tech companies for signals of where they’re going. To him. their product direction appears to be “collapsing” toward Anthropic’s approach—trying to understand what they want to offer people and sticking closer to that. He contrasted that with Google. which he says is still pushing a different direction and doing “not [so]selves any favors” by being vague about it.
He argued that when Google talks about how it thinks it will change search, it often sounds focused on use cases like shopping and commercial transactions—rather than what many people associate with Google after using it for “two or three decades”: information retrieval.
Korosec’s contribution brought the tension into sharper focus. She described a separate moment showing what she sees as a mismatch between Google’s ambitions and the fundamentals users expect. “We had a great article that just published about how Google doesn’t know how to spell its own name. ” she said. pointing to a test where the system was asked: “How many P’s are in Google?” and reportedly answered “two.” For her. it’s the same clash: Google appears to be chasing the thing it feels it has to do to keep up. while potentially damaging the part of the brand people attach to most.
In that atmosphere, Korosec also asked a question that sounded like it was looking for opportunity as much as explanation. If anti-AI sentiment is driving people’s feet—literally moving them to other services—might that create openings for startups. or for other areas of business that haven’t been considering that lane?.
Ha agreed that building for a skeptical audience isn’t simple. He described how polarized the market is: a product tailored for people wary of AI could alienate users who are more enthusiastic. But he also pointed to how DuckDuckGo is promoting itself, emphasizing that it is “anti-AI,” which he found striking.
He said he’s been moving away from Google and exploring alternatives. In his telling. a year ago even those competitors were still experimenting with AI features and putting AI more forward because they thought they had to. Now. he said. the posture has shifted—offering a more deliberate choice to keep AI out of the core experience. or contain it in a separate sandbox.
That’s where Levie’s “AI psychosis” comment came back into the room—not as a meme. but as a way to explain what happens when executives talk about AI without being close to the work it touches. Korosec introduced Levie’s point that CEOs are prone to the condition because they’re “distant from the last mile of work that still has to happen to generate most value with AI.” She wondered whether other CEOs agree. and what this shift in thinking might mean for their workforce.
Work, of course, is the other half of the story—and it’s already arrived with hard edges. Korosec pointed to “real evidence” that companies are using AI tools in ways that affect workers directly, including layoffs and changes in how people work.
O’Kane brought that workforce question into the sectors he covers. saying that companies he tracks often deal with physical transportation. or areas adjacent to it—where change has appeared slower than it has in software. He mentioned Mind Robotics. a spinout from Rivian CEO RJ Scaringe. and said more AI is being applied to physical infrastructure. manufacturing. robotics. and self-driving.
Anthony connected the debate about tools to the debate about layoffs. asking whether AI adoption and job cuts are driven from the top down or bottom up. He described earlier workplace shifts as at least partly bottom up. with tools that people liked getting adopted first. then executives and IT teams catching up.
His worry is that today’s AI productivity story often gets embraced by executives—or by the VCs funding startups—because the dream is simple: a smaller team doing the work of a much larger one. He didn’t dismiss the idea entirely. but he echoed Levie’s warning that if you’re not touching the end work. it’s hard to know what’s actually being achieved.
Korosec’s closing emphasis landed on the same theme from a different angle: companies aren’t just talking about AI. She said there’s evidence they are using it—and that those choices are hitting workers in the form of layoffs and changing day-to-day work.
The debate over “AI psychosis” isn’t really about whether AI exists. It’s about who gets to interpret it, who pays the cost when expectations meet reality, and whether the people steering the ship are close enough to feel the waves.
AI psychosis Aaron Levie Equity podcast DuckDuckGo installs up 30% Google search AI TechCrunch layoffs AI workforce change