What the piece argues

Mark Zuckerberg has published a long piece proposing a philosophy for how superintelligence should be used. The question at its centre is this: will superintelligence be centralised and restricted to a few institutions, or will it be a tool that empowers everyone?

The framework proposed rests on three principles: individual empowerment as the source of prosperity, invention as the primary purpose of superintelligence, and balance of power as the foundation of safety. Zuckerberg writes that he finds it surprising that the discourse from many developing AI is so filled with doom. He does not understand why anyone who believes AI will eliminate most jobs and much of humanity's relevance would rush to build that future. The sharpest line sits here: the notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic.

The concrete promises

To make the argument tangible, the piece lists several ways Meta says it is building personal superintelligence:

  • A personal agent: everyone will have an exceptionally capable agent that understands them, their goals and everything they care about, working on relationships, health, career, finances and home management. The piece says it will have strong privacy and security options so that no one else can access the information, similar to how encryption works on WhatsApp.
  • Tools for creation: Zuckerberg cites his 8-year-old daughter coding her ideas and producing videos in an evening, and researchers at Meta generating novel crystal structures for augmented reality glasses.
  • Entrepreneurship: as people become able to manifest ideas without raising money or building large teams, the economy is expected to become more entrepreneurial.

The tension in practice

Read on its own the piece is a coherent defence of openness. But TechCrunch's Equity team points out that the model announcements made in the same period complicate the picture.

Meta released Glimmer that week: an open-weight model anyone can download and run on their own hardware. The company's more powerful model, Muse Spark, meanwhile stays locked behind Meta's own APIs. The open model and the most capable model, in other words, are not the same model.

How to read it

That distinction points to where the open-weights debate keeps returning. There is a large practical difference between a company releasing a model and releasing its best model. Opening the second-best model feeds the ecosystem while preserving a competitive edge; that is commercially reasonable, but it leaves a gap between practice and rhetoric.

Zuckerberg's historical argument is also open to challenge. The piece holds that each transformative advance has ended with more people sharing greater prosperity, health and freedom. That optimism rests on the outcomes of past technologies; but the same history also shows that the distribution of benefit was shaped through institutions and regulation rather than arriving on its own. This part of the piece proposes no concrete mechanism for how distribution would be secured.

What we have, in the end, is two separate documents: a philosophical text and a product decision. How closely they align will become clear from which side Meta keeps its most capable model on in future releases.