Zuckerberg's AI manifesto promises AI will free up your time. Two days later, his own CTO told employees not to expect any of it back.

Meta CEO Mark Zuckerberg published a roughly 6,500-word essay on Aug. 10 titled "The Future Is for Everyone," arguing that AI's central danger isn't the technology's capability but who controls it. His answer is what he calls personal superintelligence: AI distributed to individuals rather than concentrated in a handful of labs, companies, or governments, built around three stated principles — individual empowerment, invention as AI's primary purpose, and a balance of power as the actual foundation of safety.

It is, on its face, one of the more direct public statements any frontier-lab leader has made this year on who should hold power over AI rather than just how capable it should get. That framing also has a direct line to Meta's own product decisions: Muse Glimmer, the open-weight agent model Meta released the same week, was itself framed by Zuckerberg as a response to Chinese open-weight competition and a deliberate break from the closed-model consensus at OpenAI and Google — the manifesto is the philosophical case for a product strategy Meta had already started shipping.

"Most other labs are focused on building AI for companies, governments, or other institutions, so if those labs lead, the balance of power will favor larger institutions over individuals," Zuckerberg wrote, in the essay's most direct shot at the rest of the frontier-lab field. Meta backed the argument with concrete commitments: a $1 billion "Future Is for Everyone Fund" for communities near its data centers, a new board-level structure to review model safety compliance, and a resumed push on open-weight models — the same week Meta released Muse Glimmer, a 30-billion-parameter open-weight agent model.

The product vision underneath the philosophy is concrete, if unproven. Zuckerberg describes agents that come to "understand you, your goals, and everything you care about," working across a user's devices, paired with what he calls a fully private mode he says even Meta itself cannot access. None of that exists yet as a shipped, independently testable product — it describes what Meta intends to build, not something a reporter or reviewer has been able to check against the claim.

The quote the manifesto didn't have to answer

The manifesto's pitch to ordinary users is that personal AI frees up time for "the things you enjoy." Two days after it published, Meta CTO Andrew Bosworth — one of Zuckerberg's longest-serving lieutenants and the executive most directly responsible for Meta's AI and infrastructure build-out — gave a different answer in an internal Q&A, after an employee asked whether AI-driven productivity gains would translate into more time off. According to Futurism's report on the exchange, Bosworth said he treats his own gains from AI as more work, not less, and characterized the question itself as poor career strategy for the employee who asked it.

"I get an extra hour. You know what I do with it? I put it into that." — Andrew Bosworth, Meta CTO, in an internal Q&A reported by Futurism

Whether the messenger undercuts the message

Meta has been here before. Tech critic Mike Masnick, writing at Techdirt, traced a similar arc through the company's history with open standards: Meta built Threads on the open ActivityPub protocol while the product needed the credibility of interoperability, then — in Masnick's account — moved integration work into "maintenance mode" once Threads had scale, with Meta's own leadership later saying the open standard wasn't actually a priority. Masnick's essay also points to Meta's broader open-source record, including projects like PyTorch and React, as evidence the company's relationship with openness has always been selective rather than principled — genuine when it serves Meta's competitive position, expendable when it stops. Masnick's argument isn't that Zuckerberg is wrong about the danger of concentrated AI power; he calls that point "obviously true" regardless of who's making it. It's that Meta's own record makes it, in his account, an unreliable messenger for delivering the outcome the essay promises.

The gap between the two statements is really a gap between two different questions about AI-driven productivity. The manifesto answers who should control the technology. Bosworth's remark answers who captures the value once it works — and inside Meta, on his own account, the answer is the company's output targets, not an employee's calendar. Those are separable questions, and a reader can find the manifesto's concentration-of-power argument sound while still noticing that Meta hasn't yet said anything public about the second question at all.

None of this makes the manifesto's central argument wrong. The risk of AI power concentrating in a handful of institutions is real, and Meta's open-weight releases this year are a genuine, checkable counterweight to that risk, whatever the company's motives for making them. What it means is that a single week produced two Meta statements about the same technology, aimed at two different audiences, that don't describe the same relationship between the company, its AI, and the people who use it. Readers got the essay promising more time for what they enjoy. Meta's own employees, for now, got the other answer.