Meta's AI overhaul strains its engineering culture, insiders say
Meta's aggressive pivot toward AI is taking a visible toll on the engineering culture it was long known for, according to interviews with current and former employees gathered by WIRED, accounts collected by tech writer Gergely Orosz in June 2026, and a TechCrunch report that branded Meta's months-old AI unit "a soul-crushing gulag."
Inside Applied AI. Much of the discontent centres on Applied AI (AAI), a unit of roughly 6,500 engineers and product managers formed in March 2026 to support researchers at Meta Superintelligence LabsDealroom has a profile for this one. Try Dealroom →. Three current employees told WIRED there is widespread dissatisfaction with how the team was assembled and with the work assigned to it — generating coding problems and puzzles to train and evaluate frontier models, with some staff asked to finish two tasks a week. One described the assignments as "mechanical and not creative" and a poor use of skills; another said "most people find the work soul-crushing"; a third called it "literally the gulag," adding: "You have zero purpose in life all of a sudden, you barely interact with anyone, you just have these tasks every week." Engineers selected for the unit must either join or leave the company — an unusual ultimatum for highly valued technical staff that has led some to call themselves "draftees." The team has grown in waves since early April; as one early member put it, "It's crazy to watch people experience the shock of it as each wave comes in."
The frustration boiled over publicly when someone interrupted a livestreamed, employee-only presentation — open to thousands of staff — with an expletive-filled outburst about "being the company's bitch," asking the call's leaders to relay an insult to a specific Meta AI executive. One presenter reportedly covered their face with their hands; the meeting's leaders asked everyone to mute and moved on, as employees commented on the stream about the "spicy" start.
Force-reassignments, layoffs and monitoring. Per engineers Orosz has spoken with, Meta leadership has force-reassigned some of its strongest developers to full-time AI data-labelling work; carried out a further round of layoffs (around 10%, or roughly 8,000 staff, on top of earlier cuts that ran to roughly 25% in two steps); and begun recording the screens of US-based developers around the clock. More than 1,600 employees have signed a petition demanding Meta halt an initiative to monitor US workers' clicks and keystrokes to generate AI training data; the company has since allowed staff to pause collection for up to 30 minutes and request exemptions.
Orosz stresses that none of this was forced by external pressure or business necessity — Meta has been posting record revenue and record profits — calling it "the biggest self-inflicted eng culture destruction" he has seen in a matter of weeks. The strain extends beyond AAI: at an Instagram all-hands, chief product officer Chris Cox described a "difficult" and "brutal" environment created by the "insanity of this company," likening the past few months to "running a marathon in the middle of a hailstorm and then, like, your teammate gets replaced and then we're recording you." Cox also urged a reality check on the technology itself: "It is neither god, nor is it the devil... it's nowhere near as good as you think it is, and it is nowhere near as bad as you think it is." In a Friday memo, chief executive Mark ZuckerbergDealroom has a profile for this one. Try Dealroom → acknowledged the changes had caused distress, admitting "we've made mistakes and will almost certainly make more."
Walking it back. Zuckerberg has reportedly moved to steady the company: vowing no further mass layoffs in 2026, capping the number of employees per manager (a ratio that had ballooned to 50-to-one on teams such as AAI), raising team-event budgets, planning a hackathon, and restoring assigned desks in many offices by year-end. He framed AAI as a waypoint rather than a destination, writing that the work "lets very talented people contribute" while Meta creates other roles for them "over the coming months," and — unlike some rival labs — cast "automating work" as not Meta's primary focus, with a "north star" of being "the best place for the most talented people in the world to make an impact."
The strain also ties to Meta's broader AI push. In June 2025 Meta took a roughly 49% stake in data-labelling company Scale for about $14.3bn and installed founder Alexandr Wang to lead its new superintelligence effort — a deal that helps explain the redeployment of internal engineers onto labelling tasks. Orosz also links Meta's heavier reliance on AI for code and reviews to a run of recent reliability problems, including a SEV-0 outage and a high-profile account-takeover incident. As he puts it, with the layoffs and forced reassignments, "hiring from Meta has never been easier."
Read more: WIRED · TechCrunch · Gergely Orosz on X · thread
