I also struggled with this for similar reasons. His favorite AI feature is essentially writing valid js for calculations (his example is literally valid js if you just drop the equals on the last line - you can paste it right into the console on his site and see the answer).
The whole article feels like it suffers from a similar lack of coherence. Ex - I am hardly an apple fanboy (I strongly dislike the company) but the complaints here are basically
The service sometimes has outages
The image gen is not as customizable as he'd like
He's morally opposed to cleanup in photos
Notifications summaries are bad (and how dare I get my texts 5 seconds slower).
---
None of that is really related in any way to the security footprint of the tooling he discusses up front, and it's also hardly distinct from most other current AI offerings, and it's not really a consistent complaint about the tech.
My opinion of Apple is that they do a crappy job with the vast majority of their apps...
They build good hardware, and they abuse their small hardware footprint to make decent device experiences and a decent (but getting worse) OS - but their actual applications are generally mediocre at best (mediocre copies of a previously successful, usually better, app that they will put out of business through shady store practices if I'm being blunt).
---
If anything, the failure here is that Apple marketed a thing that AI can't really do (yet, maybe at all), and most of the things AI can do without being incredibly invasive aren't actually all that useful to most folks. Very useful to a handful of power users in specific circumstances, but otherwise essentially novelty apps.
So... it's not an implementation failure. It's a marketing failure. And this is hardly unique to Apple right now. The only difference is that usually Apple doesn't play their hand until this inflection point with new tech is over, so it's more obvious this time around just how bad the product fit is for general use.
Minor (in terms of how relevant this is to your comment, not in importance) correction: the author is a woman (actually prefers they/them according to their GitHub (https://github.com/Xe).
The whole article feels like it suffers from a similar lack of coherence. Ex - I am hardly an apple fanboy (I strongly dislike the company) but the complaints here are basically
The service sometimes has outages
The image gen is not as customizable as he'd like
He's morally opposed to cleanup in photos
Notifications summaries are bad (and how dare I get my texts 5 seconds slower).
---
None of that is really related in any way to the security footprint of the tooling he discusses up front, and it's also hardly distinct from most other current AI offerings, and it's not really a consistent complaint about the tech.
My opinion of Apple is that they do a crappy job with the vast majority of their apps...
They build good hardware, and they abuse their small hardware footprint to make decent device experiences and a decent (but getting worse) OS - but their actual applications are generally mediocre at best (mediocre copies of a previously successful, usually better, app that they will put out of business through shady store practices if I'm being blunt).
---
If anything, the failure here is that Apple marketed a thing that AI can't really do (yet, maybe at all), and most of the things AI can do without being incredibly invasive aren't actually all that useful to most folks. Very useful to a handful of power users in specific circumstances, but otherwise essentially novelty apps.
So... it's not an implementation failure. It's a marketing failure. And this is hardly unique to Apple right now. The only difference is that usually Apple doesn't play their hand until this inflection point with new tech is over, so it's more obvious this time around just how bad the product fit is for general use.