I thought to check for VW's plans for the batteries needed for 1 million cars and sure enough:
Slide 9, "Battery supply secured": I wonder if Diess will be sued when this turns out to not work like he had envisioned?
Even more interesting is this little unexpected tidbit:
Slide 10, "Level 5 autonomous driving will take up to 1 billion lines of code": Trying to get to L5 by writing that much code - and doing it with sufficiently few detects - will not be possible on any relevant time scale.
For comparison, the Linux Kernel that runs (on) every Tesla is only on the order of 10^7 lines, with 99% of that being drivers and portability(*). So between the actual Linux kernel and Diess L5 strategy you have 4 orders of magnitude in lines of code that have to be correct - and the complexity of the software to be validated does not just grow linearly with the number of lines of code.
Clearly, VW's lack of understanding of software goes way beyond being unable to write what they need for their ID.3 .
Contrast that with Tesla's FSD approach, which now is a matter of tagging relevant training data and applying that to their Neural Network.
For Tesla, this is awesome news.
PS. (*) Linux runs on an astounding number of platforms - and includes the source code for drivers for all supported peripherals. Naturally, any actual system (like a Tesla) will need only a minute subset of this vast majority of the code.
