maybe if it was able to do anything useful (like tell me where specific settings that I can’t remember the name of but know what they do are on my phone) people would consider them slightly helpful. But instead of making targeted models that know device specific information the companies insist on making generic models that do almost nothing well.

If the model was properly integrated into the assistant AND the assistant properly integrated into the phone AND the assistant had competent scripting abilities (looking at you Google, filth that broke scripts relying on recursion) then it would probably be helpful for smart home management by being able to correctly answer “are there lights on in rooms I’m not?” and respond with something like “yes, there are 3 lights on. Do you want me to turn them off”. But it seems that the companies want their products to fail. Heck if the assistant could even do a simple on device task like “take a one minute video and send it to friend A” or “strobe the flashlight at 70 BPM” or “does epubfile_on_device mention the cheeto in office” or even just know how itis being ran (Gemini when ran from the Google assistant doesn’t).

edit: I suppose it might be useful to waste someone else’s time.