I found it interesting that the OP defaulted to using an AI agent for his voice recording software rather than doing a Google search. Perhaps a sign of things to come? I would've chosen Google, but maybe I'll be falling behind in the future.

Aside from getting an LLM up and running on a device, what's stopping AI from creating an operating system? I admittedly don't know much about operating system development, but aren't most operating systems written primarily in C?

I guess what I meant by that is it would be interesting if the AI prompt itself were the OS, and all software would be generated via prompting the agent. No downloads, just a "What do you need?" prompt with the AI generating everything on the fly.

Perhaps becoming so fast that you wouldn't even notice it thinking. Just: "I need to edit a document that was sent to my email" The AI would then retrieve the email, download the document and generate its own text editor to display the document in. All within a few milliseconds.

Call it AIOS