Meta's CTO admits fault: The new AI division's "management is a mess," and 6,500 engineers can now apply to transfer to other positions

Meta's Applied AI new division has about 6,500 engineers. It was established in March this year, but immediately faced numerous issues: employees described the work environment as akin to a "re-education camp." CTO Andrew Bosworth admitted in an internal memo that management was "terrible" and announced that employees could apply to transfer teams.
(Background: Dissatisfaction erupts among Meta's new AI division staff: accusations of a soul-crushing environment similar to a concentration camp, engineers in agony)
(Additional context: Meta invests $115 million to teach you to be an electrician for free, with guaranteed employment: offering five weeks of technical training without prerequisites)

Table of Contents

Toggle

• A division of 6,500 people, management ratio of 50 to 1
• Live stream hijacked, Bosworth has to apologize
• Meta's AI gamble is testing the limits of human resource management

In early June, an internal live stream at Meta was "bombed" by employees, who used profanity to vent their dissatisfaction with the company. The footage was widely circulated in the tech community. Shortly afterward, Andrew Bosworth, Meta's CTO, acknowledged in an internal memo that the management approach when launching the Applied AI division was "terrible."

A division of 6,500 people, management ratio of 50 to 1

Applied AI is a new division announced by Meta in March this year, with the core mission of transforming the company's generative AI models into practical consumer products. Zuckerberg positioned it as Meta's next decade's core engine, pulling about 6,500 engineers and product managers from various business lines to join.

However, from day one, the structure of this division planted a bomb.

The initial management structure adopted an extremely flat design: up to 50 employees per supervisor. This ratio is highly unusual in the software industry; industry norms typically range from 6 to 8 people per supervisor, with giants like Google and Amazon rarely exceeding 12 to 15. A 50-to-1 ratio means each engineer has almost no opportunity for independent career discussions, nor does anyone track their specific contributions.

What further angered employees was the nature of the work itself. Many engineers were assigned to generate puzzles and programming problems used to train Meta AI models, which was far from their expectation of "building products." Some employees began describing the environment as a "gulag" (a reference to the notorious Soviet labor camp); others called themselves "draftees," emphasizing they were forcibly transferred and not voluntarily joined.

Live stream hijacked, Bosworth has to apologize

After the internal live stream was hijacked, it drew widespread media attention. Bosworth issued a memo in this context, using unusually frank language. He wrote:

"We clearly did a terrible job explaining our vision and making everyone see how we support them and their career transitions." He also admitted: "We have broken the trust that your professionalism and contributions will be valued."

Remedial measures followed quickly: the maximum number of direct reports per manager was reduced from 50 to 20, enabling managers to truly track each person's work; at the same time, employees who were transferred could proactively apply to switch to other positions within the company, no longer locked into Applied AI.

Meta's AI gamble is testing the limits of human resource management

The chaos in Applied AI exposes a core unresolved contradiction in tech giants' AI arms race: calling top engineers to "build AI products," yet assigning them to data annotation and training tasks; promoting rapid action to build organizations, but severely overloading management structures. Numbers can grow quickly, but trust cannot.

Additionally, Bosworth left a rather meaningful remark: "AI will not replace your job, but those who know how to use AI might." On the surface, this seems to reassure employees, but in reality, it also tells them that accepting this forced transformation is the only practical choice.