Shlomi Fruchter, mentioned 9 times across podcast episodes and expert conversations analyzed by Sonic.
Gini3 is currently positioned as a research preview, whereas Google DeepMind is developing Veo as a mainstream product.
Despite its advances, Gini3 is still far from perfectly simulating the real world, indicating a large capability gap remains to be closed for future world models.
The current design of the Gini3 model limits its spatial memory to one minute.