What technological bottlenecks must still be solved to move safely from hands-off to certified eyes-off driving?

The primary challenge is collecting sufficient real-world mileage across rare corner cases to prove statistical safety beyond any doubt. Automotive driving is life-critical, meaning an algorithmic failure impacts public safety rather than just software convenience. To transition from hands-off to eyes-off, the industry requires larger onboard compute capabilities, bigger generative end-to-end foundation models, and massive troves of diverse, high-quality corner-case driving data that demonstrate the vehicle handles extreme road anomalies better than a human driver.