A very unique trait of Elon Musk companies is that they're all insanely good at the most important things, and accept less than top-tier results for the rest of things. Tesla as an example: "the cars need to absolutely blow your mind in performance and how you interface with it, and then we can figure out the whole rattle squeak thing later. As long as the car is incredible at being a car, people will come back. Yes, this includes the car driving itself. It needs to be insanely awesome regardless of how crazy people think I am." SpaceX: "It has to be reusable, and it has to go on chopsticks. Yes I get that we'll blow up like 20 rockets. Yes, we'll be so late people will think we're retarded. I don't care - needs to be reusable and it has to go on chopsticks, otherwise Mars ain't happening." I think Grok is beginning to exude these principles with Grok4 and Imagine. Just try your best to not bullshit, and here's a super slick UI to do some very unique, fun, but super useful things with AI that's intertwined with your social media app. The thing is, it's very difficult to tell if the most important things in AI today will be the most important things in the future. Today, it seems like the most important thing in AI is being repeatable at being really, really good, and the friction to make it really, really good needs to be 0. That translate to super simple, vague prompts needing to generate results faaarrrrr better than what the user expects. And once it does that, it needs to keep doing it FOREVER so that the user doesn't get frustrated with inconsistent results. But once this is figured out, the biggest brain will be BY FAR the most important, because that will dictate the company's long-term ability. What is also likely the case is that the ONLY WAY to generate those consistent, high-quality results is having a bigger, giant AI brain. And so chip demand will probably continue until we reach that point - a machine so obviously beyond powerful that we literally have no reason to keep working to make it smarter. It could also be that the easiest way to make really, really good consistent AI is by making one that knows how to improve itself, and it improving itself is what leads us to incredible results. And at that point, the system will be self-improving so fast our brains will melt. And by then, who knows what we are, who we are, why we are... We live in a very weird time you guys lol