A small X conversation around a post from @chetaslua turned into a quick brag for Meta's multimodal AI push after Alexandr Wang wrote "we are cooking!" and linked a visual reasoning example that, according to the posts around it, Muse Spark 1.1 solved while Fable 5 did not. A separate quote-post from @jhyuxm called the result "pleasantly surprised!" That was enough to get people talking on X about Meta looking stronger in vision models.