Alan@bitforth
I'll try to connect the dots for others so they can see what's going on
With TRIBE v2 You upload a video, and without human subjects or brain scanner, the model predicts, vertex by vertex across the cortical surface, exactly which brain regions activate, like emotional processing up, critical reasoning down, prefrontal modulation, etc... all of it.
That's the model, the codebase, the paper and the demo.
Now here's the rest of the context:
Meta has years of Reels data. They know empirically what content makes people stop scrolling, share, get angry, feel good. They know WHAT works.
Now pair that with TRIBE v2 and that gives them WHY it works at the level of neural tissue.
That means that anyone can, theoretically, engineer content for specific brain outcomes before publishing it.
Internally, Meta has tools called Gatekeepers and Quick Promotions. These are distribution tools that help inject content into the feeds of arbitrary population segments at scale. Not your feed specifically (though it's possible via GK Filters), but any population segment they define.
So let's connect it:
1. They know what content produces which neural states.
2. They can predict neural response to new content before publishing.
3. They can distribute that content to any population segment they choose.
So that looks like a pipeline:
First Design, then predict neural outcome, then distribute to target population.
End to end.
Now this doesn't require a human to make a single explicit decision about manipulation. You just optimize for engagement. Engagement is downstream of emotional activation and suppressed critical reasoning, so TRIBE v2 makes that relationship legible and optimizable.
Someone trying to prosecute Meta for this would have a hard case, the manipulation is structural, not explicit intent, which makes it easy to deploy and hard to prove.
Given that beliefs are downstream of emotional state plus repeated exposure. If you can systematically serve content that activates emotion and suppresses the reasoning that would challenge it, across billions of people, over months, you can effectively shape the prior probabilities people assign to claims about reality.
In plain terms: scalable, cheap psyops. Nearly impossible to detect, because it doesn’t force beliefs, it just makes certain ideas FEEL true.
Now add Thiel. Personal friend of Zuck. Early investor in Facebook. Founder of Palantir, whose entire business is population-scale behavioral prediction and targeting for governments and intelligence agencies. The same network of people building the content platform, the brain response simulator, and the population analysis infrastructure.
I'm not saying they meet in a room and plan this. I'm saying the incentives, the tools, and the relationships all point the same direction without anyone having to say a word.
The license says CC BY-NC. So Meta retains commercial rights to the most accurate brain response predictor ever built. The stated goal is accelerating neuroscience. That might even be true, too.
But that's beyond the point because the other capability is compiled into the model itself. Nobody has to explicitly write the other goal down. Someone just has to build it.
