July 24, 2026
Robot drama has entered the chat
Flux 3 X Mimic: The Next Generation of Video-Action Models
This AI went from making videos to moving robot arms, and commenters are already yelling buyout
TLDR: Black Forest Labs and mimic robotics say their new AI can turn video-learning into real robot actions, and it’s already being tested at Audi. Commenters were split between startup pride, buyout obsession, and being low-key freaked out by a robot that kept retrying a task until it nailed it.
A new AI project called FLUX 3 has people doing the classic internet double-take: wait, the same system that helps make video can also control robots? Black Forest Labs teamed up with mimic robotics and says this model can learn how the world works well enough to guide real machines, with testing already happening at Audi. In plain English: the company thinks teaching an AI to predict video teaches it cause and effect, and that knowledge can then help a robot do physical tasks.
But the real show is in the comments, where the crowd instantly split into hype squad, armchair CEO, and mildly terrified robot-watchers. One person loved the vibe of "European startups" teaming up, which gave the whole thing a bit of underdog pride. Another came in hot with the most boardroom-chaos take imaginable: if Amazon doesn’t buy them, fire the CEO. Subtle! Meanwhile, one commenter zoomed in on a video clip where a robot arm needed three tries to fix a piece of window trim and called it "unnerving" because the machine kept adjusting itself until it got it right. That wasn’t just praise — it was the kind of compliment that sounds one step away from a sci-fi warning.
And then, because no comment section can resist a meme, someone dropped the wonderfully chaotic "u.i. Zugló robot mikor?" energy — basically, cool demo, but when do we get one in my neighborhood? The mood overall: impressed, intrigued, a little spooked, and fully ready to turn a robotics demo into startup gossip.
Key Points
- •Black Forest Labs and mimic robotics developed FLUX-mimic by giving mimic early access to an early version of FLUX 3.
- •The article says FLUX-mimic is a video-action model running robots that have been tested and deployed at Audi.
- •FLUX 3 is described as a single multimodal foundation model jointly trained across images, video, and audio.
- •Black Forest Labs says video prediction accounts for over 95% of FLUX 3’s compute costs and teaches the model physical dynamics such as contact, motion, weight, and cause and effect.
- •In a large-scale training run, adding action prediction initially reduced video-generation ratings by up to 10%, but the model regained prior quality after 3,500 steps while also predicting actions.