They don’t need to, OpenAI and Anthropic are running at a huge loss to keep it cheaper to use them than to buy the comparable hardware required to replace them. The only way running open source models is cheaper is if you ignore hardware costs, or if you have a very specific and unusual usage pattern (i.e. constant 100% capacity, and not a clock cycle more or less), or if you go with another cloud provider who are also running at a loss.
Imagine if Mazda made an mx-5 EV or Toyota made an 86 EV.
I’m pro-EV in general (pro-trains notwithstanding), but don’t see the point in that. I’d rather they just got the people who designed the MX-5 and GT86 to design EV sports cars on their own terms. An MX-5 EV would just be there to remind you how much more fun the ICE version was to drive.
I wonder if any of Trump’s supporters remember that his original 2016 campaign leant heavily on the idea that he was great at hiring people, and would replace the swamp with a team of the country’s best and most highly qualified people?
We should start using sliced avocados
Yes, though it goes further than that even if you ignore the politics.
Linus is undoubtedly talking about when an LLM is used in a very specific way in his imagination where it’s helping a competent programmer spot some things they might have missed, but the programmer is still diligent and professional and genuinely does take responsibility for ensuring the quality of their commits. Where the delusion comes in is grossly overestimating how many people are actually like that outside of his inner circle of world-level experts. What’s happening in all the other open-source projects is a flood of low quality garbage PRs from meat proxies.
fake: OP is not a frog in a suit
gay: OP did girl kissing