🤚 The Open-Palm Town Hall

At an internal town hall on Thursday, Meta CEO Mark Zuckerberg told his remaining employees something that investors, analysts, and recently-laid-off workers probably did not need an all-hands meeting to figure out: AI agents haven’t progressed as quickly as he’d hoped.

The exact quote, per reporting from Reuters: “The trajectory of the agentic development over at least the last four months hasn’t really accelerated in the way that we expected.” He added that the company’s “bets on the new structure haven’t come to fruition yet.”

For context, here is what that “new structure” involved:

• 8,000 employees laid off — approximately 10% of Meta’s corporate workforce
• 7,000 additional employees reassigned to various AI groups, including one called Agent Transformation
• An expected $125 billion to $145 billion in capital expenditures for 2026, directed at AI infrastructure

Zuckerberg also conceded that the reorganization wasn’t as “clean” as it could have been and that executives had “miscalculated on the timing.”

👐 The Two-Handed Performance Review

Let us take a moment to appreciate the operational choreography on display here. In a single fiscal year, Meta eliminated the equivalent of a mid-sized company’s entire headcount, funneled thousands of survivors into a division called — and we cannot stress this enough — Agent Transformation, committed the GDP of a small nation to server racks, and then gathered everyone in a virtual room to announce that the agents have not, in fact, transformed.

This is the corporate equivalent of demolishing your kitchen, hiring three Michelin-star chefs, buying $145 billion worth of Viking appliances, and then telling your dinner guests that the soufflé needs a few more months.

The admission is notable not because AI agent development is difficult — everyone in the industry knows that agentic AI involves unsolved problems around reliability, tool use, and multi-step reasoning — but because Meta restructured its entire company around the assumption that these problems were about to be solved. The layoffs weren’t described as a cost measure. They were described as a strategic pivot. The pivot pivoted.

Zuckerberg’s assurance that “more substantial benefits” should materialize within three to six months is the kind of timeline that sounds specific enough to be reassuring and vague enough to be meaningless. In Silicon Valley, “three to six months” is the temporal equivalent of “we’ll circle back.”

🌿 The Gentle Awakening

There is something almost poetic about the fact that the company that renamed itself after a virtual world that nobody wanted is now spending more than any company in history on AI agents that also aren’t performing as expected. Meta’s corporate history is becoming a masterclass in premature commitment to technologies that are perpetually six months away from maturity.

The employees who survived the layoffs — the ones now sitting in the town hall listening to their CEO explain that the reason their colleagues were fired hasn’t quite worked out yet — must be experiencing a very specific flavor of existential dread. You were kept because the robots needed supervisors. The robots are behind schedule. What does that make you?

It makes you an employee of a company spending $145 billion on a bet that hasn’t paid off, led by a man who remains “confident” it will, inside a reorganization that was acknowledged to be messy, in an industry where the consensus is shifting from “agents will replace workers” to “agents will eventually assist workers, probably, in certain contexts, with supervision.”

👑 The Gold-Leaf Reckoning

The broader implication is significant. If Meta — a company with virtually unlimited compute, some of the best AI researchers on Earth, and the institutional willingness to fire ten percent of its workforce on a hypothesis — can’t get AI agents to accelerate on schedule, the rest of the industry should be taking very careful notes.

This doesn’t mean AI agents won’t work. It means the timeline between “this technology is impressive in a demo” and “this technology reliably does a job that a human used to do” is longer than the investor presentations suggested. The gap between can and consistently does has always been where fortunes go to die.

Meta will survive this. It has $60 billion in quarterly revenue and a social media monopoly that prints money regardless of what the AI division produces. But the 8,000 people who lost their jobs so that Agent Transformation could be born — they might have some thoughts about a timeline that “hasn’t come to fruition yet.”

“We fired the humans, hired the agents, and now the agents need three to six more months. So if anyone has questions about job security, the answer is: it depends on Q4.” — The Slap of Wisdom HR Analytics Bureau, currently being run by an AI that keeps scheduling its own performance review