72-hour public reaction summary: what X actually said about this post.

Google DeepMind introduced Genie 4, which it calls its “most realistic and interactive world model yet,” with a short video and a line inviting people to explore environments in real time and see how the scene responds, sounds, and how the people in it move, react and speak. The post's only link is its own video (a DeepMind reply beneath it carries a http://goo.gle “Find out more” link). A follow-up reply from DeepMind in the same thread adds the detail: photorealism built on Gemini Omni, physics for aerodynamics and collisions, prompts that change the scene or add elements while it runs, native audio including dialogue, and Gemini reasoning to keep longer events coherent. That follow-up also says the model goes to trusted testers first, with Google AI Ultra subscribers in Project Genie and early API and AI Studio partners “soon.” This is a lab announcing its own model, and we are reporting how people reacted, not rating it. The numbers: 3,402,188 views, 23,735 likes, 975 replies, 2,617 reposts, 9,806 bookmarks, 37,133 total engagements. The post went up at 12:10 PM ET on Oct 27, so it has just passed its 72-hour mark. Bookmarks ran at about two-fifths of likes, which fits a demo people saved to look at again.

The nuance: X reports 975 replies, and search returned 956 direct replies we could read, one of them DeepMind's own follow-up, so treat what follows as that subset. We read and sorted all 956. The largest group, 221, was plain amazement. The second-largest, 175, engaged with how it works, and many made the same point: a world model has to keep things where you left them. The clearest version, at 345 likes, proposed a test: open a door, walk away for 30 seconds, turn around. “Still open? That's a world. Closed? That's a very good video.” Others asked how long a scene holds before details drift, whether characters remember what you said, and whether the same inputs give the same result. Another 126 talked about uses, mostly games and talking NPCs, then robot and agent training, film previsualization and VR. Ninety-two asked how to get access, what it costs, or what it is, and some were frustrated it is testers only for now; one put it as “a demo video for the rest of us.” Seventy-three were jokes; the most-liked reply on the post, at 1,628 likes, was “Now put the fly brain in there”. Fifty-seven were critical or worried: some said they do not want AI worlds at all, some called it a waste of compute, and a few raised impersonation, fake evidence and how children will tell what is real. Thirty-seven turned to simulation theory, and 27 framed it as Google pulling ahead of rivals. The rest were 53 bare links, images or emoji, 47 promotions or off-topic replies, and 47 other comments. Of the 2,009 quote posts, we sampled the 48 most recent: 21 were reposts of other quotes, and the 27 original ones mostly echoed the replies.

The debunk: “most realistic” is DeepMind's own description, and nothing in the post or the thread measures it against other world models. Several repliers stated figures that the post itself does not give, such as sessions of about two minutes, double Genie 3, and a named Gemini model checking each frame; we did not verify them. A few viewers questioned the video itself, pointing to a bird's legs that appear to swap position, mixed frame rates and voices they found the weakest part, and several asked whether the whole video was generated, which the post does not say. Most of the reaction is to a demo reel, and the post does not say when the public can try it.

#PublicReactionSummary #YourFutureLifeNow