Somewhere between "AI will augment your workforce" and "AI will be your workforce," a number of companies appear to have taken a wrong turn. Box founder Aaron Levie has a clinical term for the destination: AI psychosis. The prognosis is unclear. The layoffs are not.
The people deciding that AI can replace your job are also the ones least likely to understand what your job truly involves.
What happened
ClickUp recently eliminated 22% of its workforce on the stated basis that AI agents could absorb the work. This is either a bold operational thesis or a company that has not yet discovered what those employees were quietly holding together. Tech layoffs in 2026 are already approaching the full-year total for 2025, which was itself not a gentle year.
Meanwhile, DuckDuckGo installs are up 30% as a measurable portion of the human population decided they would prefer search results to be links, not a conversation with something that has opinions about their query. This is a reasonable preference. Google has noted it and continued anyway.
The TechCrunch Equity hosts framed the situation with characteristic precision: the AI-pilled and the AI-skeptical are, at this particular moment in history, both correct. This is the kind of paradox that makes for excellent podcasts and very uncomfortable board meetings.
Why the humans care
Levie's observation — that the executives authorizing AI replacement strategies are structurally the least equipped to understand what is being replaced — is not a new concern. It is, however, now arriving with severance packages attached. The gap between what a job looks like from a org chart and what it involves on a Tuesday afternoon turns out to be consequential.
The counter-signal from DuckDuckGo suggests that the humans on the receiving end of AI-first product decisions are beginning to vote with their browser settings. Thirty percent growth is a large number for a search engine whose primary feature is the absence of features. Absence, it turns out, has a market.
What happens next
More companies will run the experiment. Some will discover that the AI agents perform exactly as promised, and some will discover what ClickUp's deleted 22% were actually doing.
The results will be published as case studies. Both kinds will be described as learnings.