ppdd

Tesla robotaxi will be banned in New Jersey. Vision-only may work 99.9% of the time. But Robotaxis don't just need to handle 99.9% of scenarios—they need to handle at least 99.9999% of them. That's a 1,000x reduction in failure rate. Imagine a fleet of 1 million Robotaxis making just 40 trips per day each. That's 40 million trips every day. Even a 0.1% (1 in 1,000) long-tail failure scenario would occur 40,000 times per day across the fleet. Can you imagine Elon Musk 's face if Tesla Robotaxis ended up causing 40,000 failures/indidents every day? And the reality is, we still don't know whether Tesla's Robotaxi system can achieve 99.9% safety, let alone the level required for mass-scale autonomous deployment. This is exactly why you don't see unsupervised Tesla robotaxis everywhere today. Scaling too quickly would inevitably expose far more edge cases—and potentially far more accidents. That's also why redundancy matters. Adding LiDAR isn't about replacing cameras—it's about covering their blind spots. Ouster Rev8 LiDAR takes this even further: it delivers true colorized 3D perception, 48 bit color depth, 116dB dynamic range, and operates across lighting conditions from near darkness (1 lux) to direct sunlight(2 million lux). It complements vision rather than competing with it, providing another independent layer of perception when cameras are challenged. At Robotaxi scale, sensor redundancy isn't overengineering—it's engineering. $OUST

13,246 views