Meta CTO Calls AI Reorg 'Atrocious,' Promises Management Fixes. Facebook Also Rolls Out AI Features Nobody Asked For.

Since the internal revolt inside Meta's Applied AI unit went public in May, the company has cycled through internal acknowledgments, a hackathon backlash, a 1,600-person petition, and a viral livestream meltdown. As of June 15, 2026, the executive most directly responsible for the division has put his name on an admission: it was mishandled.

Bosworth Names the Problem

Chief Technology Officer Andrew Bosworth posted an internal memo Monday that Wired obtained and reported on. His assessment was blunt. "We obviously did an atrocious job explaining the vision, giving people a clear picture of how we would support them and their careers in the shift, and painting a picture of how it would change over time," Bosworth wrote.

The Applied AI division was formed in March. It consolidated roughly 6,500 engineers and product managers into a single unit tasked with improving Meta's generative AI models. Workers described the resulting experience as menial. One called it "a gulag" — a quote that surfaced in prior Wired reporting and has since defined the public narrative around this reorganization.

Bosworth's memo did not dispute the characterization of the rollout. He acknowledged that the rapid restructuring "shook up the management structure that was providing you stability" while a "boom/bust cycle of hiring left entire teams in the lurch."

What He's Actually Promising

Bosworth outlined several specific structural changes, according to Wired.

Managers will be capped at roughly 20 direct reports each. The company will try to limit how frequently employees are reshuffled to new managers during future reorganizations. Managers will be evaluated primarily on managing people, not on their own independent technical output. Workers will have access to "AI coaching" tools, though Bosworth framed those as optional.

The memo also mentioned snacks. Wired noted that Bosworth described plans to "rekindle" a more cheerful internal culture through better communication, career growth, and snacks. That last item may signal whether leadership views office perks as meaningful reform. Given that the complaints center on career stagnation, surveillance, and menial work assignments, the snack mention will likely not go unnoticed internally.

Meta declined to comment to Wired on the record.

The Strongest Defense of the Reorg

Meta's leadership would argue what some engineers outside the controversy might agree with: AI product development genuinely requires large-scale organizational pivots that are inherently disruptive. Zuckerberg has said publicly that Meta needs to compete directly with OpenAI, Google DeepMind, and Anthropic. Building a centralized AI division from 6,500 people pulled from different teams is not a bureaucratic preference — it is an attempt to stop fragmented development from losing ground to better-coordinated competitors.

Bosworth himself acknowledged in the memo that the strategy behind the reorg was sound. His admission was specifically about the execution: the communication, the career-path clarity, and the management continuity. That distinction matters. A badly communicated good idea is still a bad implementation, but it is a different problem than a bad idea.

The open question is whether the structural fixes Bosworth announced will hold once the next strategic pivot arrives. At Meta, those pivots have come fast.

Meanwhile: Facebook Gets an AI Suite That Looks Like Everyone Else's

Separately on June 15, Meta announced a set of new AI tools for Facebook users, reported by Engadget. The headline feature is "AI Mode," a chatbot that answers questions. Meta's own example: asking it for nearby summer vacation spots. That is the same use-case demo Google, OpenAI, Anthropic, and X have all used when launching similar products.

Meta's claimed differentiator is that AI Mode pulls data from within its own ecosystem — Facebook Groups, Reels — to surface "real perspectives and experience rather than a generic list of search results." That is a plausible edge if it works as described, since Meta's social graph has information no other AI company can replicate. Whether the actual output justifies that claim remains to be seen.

The update also includes photo-editing tools: collage cutout templates, transition effects for video montages, and photo presets that let users virtually change their clothing and accessories. Meta's pitch for that last feature centers on sports fans wanting to "rep your fandom and virtually wear a team jersey." Engadget noted the obvious: "Nothing says true fandom like a fake jersey."

All of this runs on Muse Spark, the technology Zuckerberg announced last week. The features are live now on mobile Facebook, according to Engadget. No web version has been announced.

The Gap Between Inside and Outside

Meta is rolling out consumer AI features that look polished and broadly available while its internal AI engineering workforce is in documented revolt over exactly the work that produces those features. Bosworth's memo addresses the people. The Facebook AI launch addresses the product. Neither addresses whether the two are reconciled — whether the 6,500 engineers who have spent months feeling their expertise is undervalued are the same people shipping the tools Facebook users will see this week.

Bosworth said Meta will focus more on AI coding tools as a strategic priority, per Wired. That is a direct market play against GitHub Copilot and Google's internal tooling. How the restructured Applied AI unit fits into that priority and whether the management changes announced Monday are enough to stop talent erosion will determine whether any of these product launches have legs.