This is a pretty interesting $NATIX update. 5,500 hours of real world driving data from the NATIX Network helped Valeo train VATIX 9B from scratch. These were some of the results: • Image quality was up 60%+ • Video consistency was nearly 70% better than other open source models They also ran 200+ experiments and got within 3.6% of the final 9B performance when predicting the result beforehand. That’s what makes $NATIX interesting. They’re building a global camera network for Physical AI, and we’re now seeing that real world data being used to train actual models.