I don't think Tesla can go and solve phantom braking in specific areas or anything, it's likely more about constantly training the system to better recognize what is and isn't a threat
I have been following this issue for several months, here, on YouTube, and on Reddit. There is plenty of video evidence, and it seems to be happening enough to make people trade in their vehicles. Even reviewers acknowledge the phantom braking problem.
It is quite possible that the problem happens much less or much more mildly than I have been led to believe, but I am uncomfortable betting on it. I just received my M3, and I'm honestly afraid to engage TACC unless there is no one behind me. Needless to say, it won't be a quite as relaxing to have to monitor traffic behind me, but TACC will have to earn my trust before I feel comfortable using it.
Yes, I knew I was potentially taking delivery of an expensive vehicle without usable cruise control!
I do think the problem is fixable, but probably not in the way Tesla is trying to fix it today. I
suspect Tesla plans to solve the problem by fixing FSD. As I understand it, they are using the FSD stack for TACC, so fixing FSD fixes TACC as a freebie. And, if they were really going to delivery FSD in a few months, this might be a viable strategy.
Unfortunately, Tesla has delusional expectations about solving FSD this year. IMO, they might be a decade away from training neural nets to drive a car reliably with vision-only.
The alternative strategy would be for Tesla to write some procedural (non-neural) code to put guard rails on TACC and AEB, and drastically cut down on phantom braking, albeit at the cost of some neural intelligence. They could probably write such a procedural fix in just a few weeks.
The recent NHTSA and WaPo attention is just the beginning. Tesla will have to figure out how much of their reputation and stock price they want to gamble on an FSD-first strategy before implementing a procedural hack for TACC. Maybe the NHTSA inquiry will get them to address the problem more directly.
Like many others, I would welcome a dumb cruise control option, but it is difficult to see how they could implement that without confusing customers or poisoning consumers against autonomous driving.