When Meta paid $14.3 billion for the startup Scale AI and assembled an army of AI engineers, the picture from outside was dazzling - the smartest people, the biggest salaries, the future being built here. From inside, the story is different. Engineers in the new division describe their workplace with one confusing but clear sentence: "literally a gulag."

This is about the Applied AI team, formed about three months ago, with around 6,500 engineers and managers. Many of them were reassigned there without anyone asking - with unexpected emails informing them that their new task is to invent tasks and puzzles used to train AI models. The choice, as they describe it, was simple: accept or quit.

The conditions complete the picture. At first, up to 50 employees reported to a single manager. Over 1,600 workers across the company signed a petition against the tracking of their keystrokes for model training. Work that external freelancers used to do is now pushed onto highly paid engineers - monotonous, exhausting, stripped of meaning.

The justification from the top is almost insulting. Mark Zuckerberg defends the internal employees by claiming they are "significantly more intelligent" than external collaborators - which is precisely why he gives them the very thing the freelancers were avoiding. Overseeing the whole division is Alexandr Wang, the same man who sold Scale AI, now chief AI officer at Meta Superintelligence Labs.

And here's the broader lesson for everyone who believes the "dream job" story at the big tech companies. A high salary doesn't buy meaning - it buys silence. When 6,500 smart people feel like assembly-line factory workers feeding a machine, the question isn't whether AI will replace people. The question is whether it has already turned them into part of the machine - just with better pay and worse meaning.