Let's put it this way. M1 Max can build some amount of code in 134 seconds. M3 Max can do it in 70 seconds. Does this sound like a small bump? Source: XcodeBenchmark
The main reason I use fish shell is for the autocompletion feature it offers out of the box. If it had been an option back in the day, I would have tried something like starship
It just feels like “what LLM is better” becomes new “what GPU is better” type of talk. It’s great to find a clear winner, but at the end the gap between the leaders isn’t an order of magnitude.
I think people are missing the context that the prices of even the largest LLMs trend towards $0 in the medium term. Mistral-medium is almost open source, and we are still early days
I recommend reading “Revolution in the valley” by Andy Hertzfeld, who is also the author of this story. The book is a compilation of all stories from folklore.org including more interesting details about development of the Macintosh.
reply