Its all local if that helps?
And each is very specific to a use case, not a "general chat prompt for triggering API calls" but things like "ML model to categorize video clips and assigning tags + names, so you can find it faster" and similar.
I'd also get tired if it was "AI ala Microsoft/Google" where the goal is to get you to write forever with a chat bot somewhere else, but these features are very different from that.
I think they've made lots of great practical choices! 100% in agreement with running local models for these tasks.
My opinion is that, for end users, if you name your feature "AI" to market it, you kind of already failed to read the room. You're writing to VCs while hoping it convinces customers.
Name what the feature does, what it gains them. Call it "smart" if you must imply some black box treatment.
Naming AI as the selling point for everything feels a lot like that Android tablet ad circa 2010:
"Your wife will love the new dual core Tegra™ chipset!"