We are living in interesting times: there are now zillions of messiahs, not just one.

More seriously, this take seems to ignore something important: Apple Silicon is already very good at running local LLMs. If the future interface is agentic and on-device, Apple may be better positioned than this argument suggests.

I have the opposite opinion of Apple. It's not a software moat, to me. Their software is inferior in every way to the Linux systems I usually use, I find macOS quite unpleasant (shit package management, so bad people consider Homebrew an upgrade, bizarre mix-and-match of UNIX and old Mac conventions, unpleasant virtualization, weak CLI tools, and so forth), but I recently bought some Apple hardware anyway, because it's notably better hardware. And, with the RAM crisis happening, it's not just better, it's a better value. Though maybe the latter will change as Apple's RAM contracts at the old prices run out and they have to ramp up prices, too.

Apple has the best hardware for running local AI. It's such an extreme difference that a mid-range iPad or iPhone smokes most desktop or laptop PCs. Google is, AFAIK, the closest competitor with their Pixel phones/tablets, and they're also significantly behind.

But, software does seem like it won't be a moat for much longer. 18 months seems optimistic or pessimistic, based on the limits of what agents can do, even with the best models, without expert guidance and a lot of steps. You're not making much useful software anytime soon with nothing but a vague prompt.