The point is that it downloads a multi-gig model automatically without providing an easy way to reclaim that storage space if you don't intend to use that feature.
This is very much an anti-user behavior. Certainly having models run on-device is preferred when possible, but if you don't want to use the Apple Intelligence this is an anti-feature. Especially on devices with small storage capacities which are impacted the most by this.
I tried Neostore. All the info it shows is pretty cool, but ultimately I felt the expierence suffered from all the extra information shown compared to the simpler UI in Droid-ify.
I love Neo Store, because in my book complex > simple. Simplicity of interface = process handcuffs most of the time (IMO). Complexity done right permits so much diversity in composability. I would never ever want a 'simpler' After Effects or GIMP or Blender or Siemens NX. GNOME did it back with GNOME 3 and it has been trash in my book ever since! vomitface.gif
Sometim... err more times than not... there should be a "Your IQ must be at least this high" bar for good software. Stop appealing to the lower common denominators!
I'd argue you don't need to run DoT/DoH yourself if it is just for your local network.
Setting up a local resolver, such as a Pi-Hole or Unbound on a firewall can serve unencrypted responses to your devices. Those resolvers in turn can use DoT/DoH with their upstream resolvers to encrypt the requests that go through untrusted networks on the Internet.
This is the best of both worlds, the simplicity of unencrypted DNS while encrypting traffic where it matters.
> I have SMB multichannel enabled, but this real-world speed limitation could also be related to my use of my lower-power Arm NAS (running on Ampere Altra, with 32 fairly-slow CPU cores).
Based on my own testing with a 25 Gb NIC setup on macOS I strongly suspect the issue is not on the NAS side, and the hardware I was working with is by no means underpowered.
The Mac I was working with is one of those rack mount M2 Mac Pros with a 25 Gb PCI-e card. On the NAS side it was a several hundred terabyte all NVMe flash multi-node cluster, each node with Xeon CPUs and 2x 25 Gb bonded NICs. Doing some rough throughput testing from macOS over multi-channel SMB yields throughput rates that are barely faster than the built-in 10 Gb copper NICs on the Mac.
Using iperf I can confirm that I can get near enough to 25 Gb/s raw network throughput on the Mac. And other non-macOS systems see significantly faster throughput to the same NAS. I am not sure exactly what the issue is that causes 25 Gb NICs to underperform so significantly for file sharing tasks, but I am reasonably certain that the issue is in macOS.
This is very much an anti-user behavior. Certainly having models run on-device is preferred when possible, but if you don't want to use the Apple Intelligence this is an anti-feature. Especially on devices with small storage capacities which are impacted the most by this.