The open-source Llama strategy gets praised as democratizing AI, but the licensing terms still restrict commercial use past a certain scale. That's not open source. That's a free tier with a moat. And the "AI for everyone" rhetoric rings hollow when the same company builds custom silicon (MTIA) and 24k GPU clusters that no hobbyist or startup can touch. You get the model weights. They keep the compute advantage.
What's missing is any acknowledgment of the second-order effects. No mention of what happens when AI-generated content floods Facebook and Instagram — which it already is. No plan for moderation at that scale beyond "AI will solve it." No serious engagement with the creator economy they claim to support, beyond tools that let brands pump out more synthetic content cheaper.
The metaverse pivot failed because people didn't want to live in a low-res cartoon office. This manifesto feels like the same playbook: build the infrastructure, assume the use cases will follow, monetize the attention. But LLMs aren't a platform in the same way mobile was. They're a capability layer. The value accrues to whoever owns the distribution and the data — which, surprise, is still Meta.
There's a version of this future where open models actually empower researchers and small teams. This manifesto isn't it. It's a strategy doc dressed up as a mission statement. The most telling line? "We're building for the next decade." Not "we're solving for users today." That tells you everything.
Big Tech is spending way more on AI than the balance sheets 2d ago
Meta is adding Newsmax to its AI training pool 4d ago Meta is patenting AI glasses that can identify people in 4d ago
Alibaba's open source models just crossed 3 billion downloads 4d ago
Since the provided content was only a title 4d ago Meta is playing a double game by releasing Glimmer while keeping 4d ago
Next SpaceX made a play for Cognition and the Devin team said no →