Meta's A.I. Employees Are Miserable

The May 8 NYT piece anchors a story that surrounding coverage had already reported in pieces: Meta is laying off 8,000 people starting May 20, 2026 — roughly 600 of them inside Alexandr Wang's Superintelligence Labs (SSL). The framing on the inside is mission drift; the framing on the outside is that the alexandr-wang experiment isn't working.

Wang took over as Chief AI Officer four months before this article ran, as part of Meta's ~$14.3B acqui-hire of Scale AI. In two months SSL lost at least eight named researchers. Meta's reported AI-employee retention rate sits at 64% versus Google DeepMind's 78%. The math is telling — the more interesting Meta hires don't stay long enough for the org chart to settle.

In March 2026 andrew-bosworth (CTO) created a parallel Applied AI Engineering organization under Reality Labs VP Maher Saba, reporting to him rather than Wang. Engineers across Meta have been reshuffled into Saba's org instead of SSL. Insiders describe the split as deliberate — not a redundancy, a hedge. The signal to staff is that the company itself doesn't believe SSL is the AI org of record.

Wang's stated rationale for the SSL layoffs, in his memo: "Fewer conversations will be required to make a decision," and remaining staff will "have more scope + impact." The framing is indistinguishable from any 2026 efficiency-mandate memo, which is part of what makes employees miserable: there is no positive vision, only headcount math wrapped in superintelligence vocabulary. Yann LeCun publicly called Wang "inexperienced" for a $300M-package CAO role at 29.

This sits alongside two other 2026 Meta morale items: the Fortune report that Meta would start tracking employee screens and keystrokes to train internal AI tools, and Zuckerberg's broader $135B AI capex plan that locks the company into a spend cycle regardless of what SSL produces. The NYT article's contribution is to put numbers and named sources on what the rest of the coverage had only implied — that personal superintelligence is, internally, a slogan no one believes.

Connects to the broader ai-bubble-pale-horses checklist: senior departures, executive theater outpacing technical product, hiring in panic followed by layoffs at the same orgs. See also ceo-ai-psychosis for the structural sycophancy that makes these big-bet org changes feel productive without being productive, and subprime-ai-crisis for the financial overhang.

Note: NYT blocks automated fetches; the underlying source file (nyt-meta-ai-employees-miserable) reconstructs the article's claims from corroborating coverage rather than the article itself.