The internet loves a tech pivot until it turns out that Silicon Valley’s highest-paid software engineers have been stripped of their career autonomy, forced to record their screens 24/7, and transferred into a unit they openly call a “gulag” to hand-craft coding puzzles for the models meant to replace them.
Inside Meta, the transition to artificial intelligence has ceased to be a polished PR campaign. It is a corporate draft. Wired reporting has exposed a deep rift between Meta’s executive suite and the thousands of engineers who are now training the company’s next-generation coding robot. Morale has deteriorated so quickly that some employees compare their daily routines to the menial labor of digital assembly lines, culminating in a series of highly publicized internal revolts.
Maher Saba and the Hot Mic Town Hall
The frustration within Meta’s newly formed Applied AI unit boiled over in the most public way possible. During a company-wide livestreamed presentation, an employee accidentally left their microphone active. In front of thousands of colleagues, the engineer unleashed a profanity-laced rant aimed directly at the unit’s leadership, ending with a call to tell Maher Saba, the Meta veteran leading the unit, that he is a “piece of shit.”
According to witnesses, presenters on the call covered their faces with their hands, attempting to hide their laughter while the stream cut away.
The outburst was the direct result of a massive restructuring that began in early April 2026. Meta moved roughly 6,500 engineers and product managers into Maher Saba’s Applied AI division. The transfers happened in successive waves, creating a sense of dread as employees watched colleagues vanish from product teams to join the new unit.
Engineers selected for the division were given no choice: accept the transfer or leave the company. This take-it-or-leave-it policy has led employees to describe themselves as “draftees.”
Screen Recording and Soul-Crushing Work
The daily reality for these engineers is a stark departure from the prestigious software development work they previously performed for platforms like Instagram, WhatsApp, and Facebook. Instead of building features for billions of users, the unit is focused on feeding data to Meta’s AI scientists.
The Weekly Grind
- Two coding puzzles per week: Engineers are tasked with generating complex coding challenges to train and evaluate Meta’s frontier models.
- Screen recording: Meta records employees’ desktops continuously to capture the raw coding process. The program was scaled back slightly to allow employees to pause recording for up to 30 minutes daily.
- Menial tasks: Former product builders describe the work as repetitive and “soul-crushing.”
The irony is not lost on the engineers: they are hand-crafting the very datasets that Meta hopes will automate their own jobs. A second employee compared the team to a gulag, describing an environment where developers have no sight into how their work is used and barely interact with other teams. While a lack of meaningful work is far from a historical forced labor camp, the dramatic vocabulary highlights the complete collapse of Silicon Valley corporate prestige.
Chris Cox and the Marathon in a Hailstorm
Management has struggled to contain the internal bleeding. In a recorded meeting obtained by Wired, Meta Chief Product Officer Chris Cox addressed Instagram employees, attempting to validate their frustration. He compared working at Meta in the middle of this AI transition to running a marathon in a hailstorm.
Cox warned against being “over-earnest” about AI, stating that it is “neither God nor the devil.” He noted that the technology is far from perfect and does not even know what day of the week it is.
Yet, the solutions proposed by leadership reflect the standard, tone-deaf corporate playbook. To combat the collapse in morale, Meta announced:
- Hackathons: Intended to rebuild team cohesion by forcing burnt-out developers to code overnight.
- Increased team event budgets: Funding for social gatherings to distract from the screen-recording monitors.
- Manager ratio limits: A plan to reduce manager-to-report ratios, which had ballooned to 50-to-1 inside the Applied AI unit.
The Golden Age of Slop
The organizational chaos is no longer confined to internal messaging channels. It has begun to leak directly onto the consumer-facing platforms. In early June, Facebook users reported widespread outages and lexical errors. The cause was later identified as a deployment that pushed unparseable JSON data directly to production, bypassing standard pipeline checks.
When a multi-billion dollar platform is shipping broken JSON because developers do not have the time to set up active CI/CD pipelines, the slop era has officially arrived.
Zuckerberg has vowed that Meta will not carry out additional mass layoffs in 2026. For the “draftees” sitting in the Applied AI unit, that vow offers little comfort. They remain trapped in a loop of generating puzzles, watching their screens record every keystroke, and waiting for the models they are training to make their roles obsolete.
Sources
- Wired: Reporting on Meta’s internal Applied AI unit turmoil and Chris Cox statements
- Reddit: Leaked transcripts and employee discussions regarding the Applied AI draft
- DownDetector: Outage reports and JSON front-end error tracking
About the Author
The Silicon Skeptic
A former tech engineer turned investigative journalist who tracks corporate hubris, drinks cold brew from a thermos, and is entirely convinced that the AI bubble is just a sophisticated plot to get developers to work under constant surveillance.
