> cannot imagine a personal computing usage which can justify a 10k machine
For me, the privacy pitch wins. I have a friend visiting, however, who spends like $2,400 with Anthropic every year. That's a solid ROI even if the thing becomes obsolete after a couple years. (I'm still on my 2020 MacBook Pro. I love it and will be sad when I have to replace it.)
In addition to privacy I’d like to be able to burn as many tokens as the hardware will let me 24/7 without getting a surprise bill at the end of the month. I don’t care if it is slower that the cloud, I’m not in a hurry.
> That's a solid ROI even if the thing becomes obsolete after a couple years.
How can that be a solid return on investment? There's no model you can run locally to have frontier model level performance. Also who spends 2.4k yearly for personal AI usage, like what's the usecase? If your friend is spending that money for his business then it's not personal computing.
> There's no model you can run locally to have frontier model level performance
I'm betting he doesn't need a frontier model. Sonnet, today, is likely good for 80% of his tasks, which largely involve repretitive, tedious work.
> Also who spends 2.4k yearly for personal AI usage, like what's the usecase? If your friend is spending that money for his business then it's not personal computing
I have a friend who is a teacher and spends that much yearly for ChatGPT and Claude. It saves him many hours a year in time assembling documents and creating posters, flyers, and personalized learning materials for his students.
For me, the privacy pitch wins. I have a friend visiting, however, who spends like $2,400 with Anthropic every year. That's a solid ROI even if the thing becomes obsolete after a couple years. (I'm still on my 2020 MacBook Pro. I love it and will be sad when I have to replace it.)