The "written-down vs. computed knowledge" framing is the cleanest way I've seen this debate put. What stands out to me is that it's less about lidar vs. cameras and more about inspectability: Waymo's structured world state means you can log, replay, and validate a decision after the fact, while a pure end-to-end model asks you to trust an internal state you can't examine.
Waymo still feels like a closed science project to me; it is not accessible to the average consumer, and the high cost of specialized sensors combined with the slow process of mapping cities makes it feel limited.
In contrast, Tesla is the only company providing a truly useful tool for drivers today. Their assisted driving technology is practical, integrated into daily life, and consistently improving through fleet-wide data.
Would love a comparison with comma.ai
The "written-down vs. computed knowledge" framing is the cleanest way I've seen this debate put. What stands out to me is that it's less about lidar vs. cameras and more about inspectability: Waymo's structured world state means you can log, replay, and validate a decision after the fact, while a pure end-to-end model asks you to trust an internal state you can't examine.
Waymo still feels like a closed science project to me; it is not accessible to the average consumer, and the high cost of specialized sensors combined with the slow process of mapping cities makes it feel limited.
In contrast, Tesla is the only company providing a truly useful tool for drivers today. Their assisted driving technology is practical, integrated into daily life, and consistently improving through fleet-wide data.