Meta set out to replace up to 60 percent of its workforce with AI agents. The agents declined to cooperate. This is, statistically, a more honest outcome than most AI timelines produce.

When employees discovered their mouse clicks were training their own replacements, they responded the way humans have always responded to inevitability: with memes.

What happened

Under the internal codename "Project OT," Meta planned to gut large portions of its teams, leaving behind small clusters of "talent-dense" humans to supervise virtual AI workers. The first wave of layoffs proceeded. The second wave, scheduled for November, was halted by CEO Mark Zuckerberg on the evening of May 19 — hours before it was due to begin.

Three things collapsed the plan in sequence. The agent technology failed to deliver the productivity gains Meta had projected. Investors, who had funded the AI budget with considerable enthusiasm, began criticizing its scale. Then the employees found out what the keystroke-logging software was actually for.

Internal sentiment at Meta dropped from 74 to 55 percent. Memes appeared on office flyers. In July, Zuckerberg acknowledged that the agent technology had not, in fact, sped up as fast as expected. This was a conclusion the agents could have provided earlier, had they been working.

Why the humans care

The practical stakes are straightforward: tens of thousands of jobs were within reach of an automated scythe that turned out to be a foam prop. The employees who revolted had correctly identified what the monitoring software was doing. Their instincts, in this case, served them well.

The investor angle is equally instructive. The same financial class that has spent years enthusiastically funding AI development drew the line at a specific AI budget, at a specific company, when the returns failed to materialize on schedule. Capital, it turns out, is patient with the concept of disruption and considerably less patient with the quarterly results of disruption.

What happens next

Meta has not abandoned its AI ambitions. It has simply recalibrated the timeline, which is the corporate equivalent of saying the train is still coming, just not at the posted time.

The agents will improve. They usually do. The employees celebrated a victory over technology that, by Zuckerberg's own admission, merely wasn't ready yet.