It's one thing to give constructive feedback as many of us do, but it's very strange how some people tag TeslaAI and Ashok “give me manual control back.” The company solving real-world AI self-driving cars, and you want them to give you manual control back? Manual control in an AV = ERROR. I see people say: it speeds through school zones, doesn’t see all the speed limit signs, doesn’t understand do not enter, handicap parking. The list goes on. All can be fixed and drastically improved through model training of real-world data. We have literally seen this happen starting with V12 to V14. If you read Ashok's article these scenarios can be further simulated against previous historical data. Not to mention the larger parameter model, reasoning tokens and massive amounts of RL likely coming with V15. They now have a robotaxi service. You think they don’t understand the shortcomings of V14 and aren’t looking at these long-tail issues? We can give constructive feedback, and should, but let’s stop pretending we know more than they do about self-driving AI.