[P] I tested Meta’s brain-response model on posts. It predicted the Elon one almost perfectly.

A developer has created an experimental interface to test Meta's open-source 'Tribe v2' brain predictive foundation model on real social media content. The tool allows users to feed in text and visualize the AI's estimated 'brain-response footprint'—a prediction of how a human brain might engage with the content. Crucially, in a test on an Elon Musk post, the model flagged it as highly engaging and viral-like based solely on the text, having zero access to actual engagement metrics like likes, reposts, or metadata. This suggests the model is identifying latent virality signals within language itself, not just mirroring existing popularity.

Further experiments solidified its potential. The creator's own chess-related content was 'demolished' by the model, receiving a low predicted engagement score. More revealingly, when comparing space-related content framed as 'UFO' speculation versus 'astrophysics' science, the model produced completely different predicted neural response patterns despite the similar subject matter. This indicates a sensitivity to narrative framing that goes beyond basic topic classification. The project, showcased in a short video, moves the technology from a research paper into a tangible, interactive demo, forcing a concrete discussion about its implications.

The core revelation is that this represents a fundamentally different class of feedback than traditional analytics. It's not sentiment analysis rebranded; it's a proxy for pre-conscious neural engagement. This capability makes it a powerful research tool for understanding communication but also a potentially dangerous optimization engine. Creators, marketers, or bad actors could theoretically use such models to engineer content specifically crafted to hijack attention at a neurological level, prioritizing engagement over truth or quality. The open-source nature of Meta's model democratizes this powerful capability, making these questions urgent.