He's not involved with Comma much anymore and he's 100% right that Nvidia hasn't won the race yet.
I worked at Nvidia on RAPIDS. From what I understand, they've gotten better but it's still a very traditional MS style company. CUDA is amazing and is something akin to x86 with it's ubiquity. It's the fastest and best to work with but is an albatross and it's being used to accelerate the development of it's replacements. I feel his pain with this AMD stuff. I bailed on them in 2018. They do not have the ability to ship real compute tooling.
Also the myriad of ways we can now bridge to the browser with WebGPU and WebASM and Vulkan.... and FPGAs and next year photonic computing is going to start trickling in and then it's going to become. a flood. I think there are two fabs in the US. I plan to ship some test designs to SkyWater in August. With the end of globalization, photonics are going to be where we move quick.
I like Geohot. I really enjoyed his punk ass spirit as a young guy and I think he's grown up to being a decent example of how to be a tech public figure. His soundcloud is dope. Not only does it have some good songs but a few of them are about valley inside baseball that has not ever been reported anywhere but his music.
What are you talking about? They actually shipped a product that has over 90 million miles driven, blowing past all competitors apart from Tesla. You can buy the product today, hook it up to your care and get a better lane assist than pretty much every other car company ships. Plus they're profitable. Compare that to dozens of self-driving startups that are basically vaporware or outright frauds.
You can dislike the guy but you're delusional if you brush off comma.ai as not having produced anything.