NextFin News - Meta Platforms’ newest AI organization is generating internal blowback before it has produced many polished product wins. During a livestreamed employee-only presentation this week, someone interrupted with an expletive-filled rant about “being the company’s bitch” and told presenters to inform a specific Meta AI executive that “he’s a piece of shit,” according to a recording heard by WIRED. The outburst was crude, but the harder fact is organizational: Mark Zuckerberg is trying to build an artificial-intelligence stack at speed, and parts of the company are signaling that the work structure is not keeping up.
WIRED reported that three current employees described broad dissatisfaction inside Meta’s Applied AI team, a unit formed in March to support AI researchers at Meta Superintelligence Labs. Meta has assembled the group to about 6,500 engineers and product managers, yet employees said they were doing repetitive “drudgework” meant to improve AI models. One worker, speaking anonymously because they were not authorized to talk publicly, called the environment “literally the gulag,” saying employees had “zero purpose in life” and only a handful of tasks each week. On the surface this looks like a morale story; the real issue is whether Meta has created a large support layer faster than it has defined where that layer creates value.
Chris Cox, Meta’s chief product officer, effectively confirmed the strain at a separate Instagram-wide meeting this week. In the recording reviewed by WIRED, Cox called the past few months “difficult” and “brutal,” said the environment had been shaped by the “insanity of this company,” and compared shipping features for Instagram’s roughly 2 billion users to “running a marathon in the middle of a hailstorm.” He added that a teammate being replaced while employees were being recorded made the situation feel even stranger. That candor matters because Meta usually presents scale and speed as operating advantages; here, senior leadership is describing them as stress multipliers.
This is not mainly about one angry interruption. It is about whether Meta’s AI push is changing its business model in a disciplined way or merely adding a costly coordination burden. A 6,500-person AI support group should, in theory, help researchers move faster and help product teams turn models into user-facing features. If those employees instead feel underused and detached from outcomes, then the benefit of scale starts to erode: more people do not automatically mean more shipping, better products or stronger pricing power. The real trade-off is between speed of mobilization and clarity of mission, and Meta appears to be leaning hard toward the former.
Zuckerberg’s playbook has long been to move quickly, reorganize often and concentrate resources around the next strategic priority. That can work when the destination is clear and the teams know what success looks like. It works less cleanly when a strategic priority is still being defined in real time, because re-tasking turns into role compression, status anxiety and work that feels tactical rather than inventive. The Applied AI unit described by WIRED looks like the weak side of that model: a sprawling internal service function built to support Meta Superintelligence Labs, but one that employees say lacks purpose and enough meaningful tasks each week. Meta is not short on money, users or distribution. The question is whether it can convert those advantages into a coherent AI production process rather than a large internal holding pattern.
The financial implication is straightforward even if the company can absorb it. Every engineer assigned to unglamorous model-improvement work is an engineer not building product differentiation, developer tools or monetizable features. Meta can afford inefficiency better than smaller AI rivals, but the math doesn’t add up yet if thousands of workers are caught between a strategic pivot and unclear execution. Who benefits is obvious: AI researchers and the core leadership team get more support capacity. Who bears the pressure is also obvious: engineers and product managers inside the 6,500-person layer, plus product groups such as Instagram that still have to ship to roughly 2 billion users while the organization is being reshaped around AI.
What still needs to be verified is whether this is temporary friction or structural waste. The current account comes from anonymous employees and recordings of internal meetings, not from a broad survey of Meta’s workforce or a financial breakdown of what the unit is producing. Whether Meta’s approach works depends on whether this support layer is actually accelerating model quality, product rollout and commercialization at a pace that justifies the disruption. Until that is clearer, Meta’s AI story is not about ambition — it is about execution under strain, and the strain is now visible inside the company’s own meetings.
Explore more exclusive insights at nextfin.ai.
