Meta's AI Manifesto Is Long. Its AI Strategy Is Harder to Find.

Zuckerberg published a 6,500-word manifesto on Meta's AI ambitions this week, and the length itself is the first fact worth reading. A strategy document is supposed to commit: capital, timelines, model and product plans, a competitive frame. WIRED's Uncanny Valley panel read it as a document that asserts Meta is going to win the AI race while leaving the specifics of how for the reader to take on faith. (WIRED Uncanny Valley)

The memo does make concrete claims. Zuckerberg argues that personal superintelligence, meaning AI that works for the individual rather than the institution, is the next major platform shift, and that Meta is positioned to lead it. He leans hard on Meta's distribution: more than two billion people already use Meta's apps, which gives the company a runway no frontier lab can match on reach. Those are real points. The question is what a strategy document owes the reader beyond the runway.

A useful AI strategy memo answers four questions. How much capital is being committed, and over what window? Which model capabilities are being prioritized, and on what timeline? How will the product surface those capabilities, and to whom? What is the competitive moat that keeps the lead from being clawed back by a faster lab? The Uncanny Valley panel's read is that Zuckerberg's memo, at 6,500 words, answers none of them with the precision a memo of that length invites. (WIRED Uncanny Valley)

That absence is the story. The frontier labs competing with Meta (OpenAI, Anthropic, Google DeepMind) publish research roadmaps, model release notes, and infrastructure budgets that the press and the market can score against. Meta's own open-weights releases, most notably the Llama family, have historically done the same: here is the model, here is the benchmark, here is what it costs to run. The manifesto reads as a deliberate step away from that accountability surface, replacing specificity with ambition. The result is a document that defends the right to win without describing the path to winning.

Meta is not a startup. It is one of the most-watched AI competitors in the public market, with a multi-billion-dollar GPU buildout, a research organization (FAIR) that has produced real published work, and a product surface (Instagram, WhatsApp, Facebook, Threads) that could plausibly distribute AI features at a scale no consumer AI company can match. A reader trying to evaluate Meta's AI position is owed more than a sweeping vision. The memo's silence on capital intensity, on which models ship next, on how the assistant surfaces into existing apps, and on what Meta's moat is against labs that ship models on a faster cadence is the news.

The pressure on this kind of document is not just editorial. The AI race is moving on a quarterly cadence, with new flagship models from OpenAI, Anthropic, and Google roughly every few months. A manifesto that does not name a model roadmap or a capex figure is, in practice, asking the reader to underwrite a future that the next lab release will either validate or disprove. The Uncanny Valley read is that verbosity, not specificity, is what the memo actually delivers.

There is a portable lesson here, and it will outlast this memo. The next time a major CEO publishes an AI vision document, the question worth bringing is the one the Uncanny Valley panel asked: at the stated length, what is it actually committing to? When a manifesto that long answers the easy questions and leaves the hard ones open, the silence is the headline.