I would hesitate before entrusting that kind of hardware to an AI agent. It is an interesting idea, but what if it decides full format is necessary? In my personal anecdata of where AI went wrong, at least I was the human in the middle to stop the bad decision. At best, you end up with a Simpson level nuclear plant terminal, where Homer is just trained to press Y each time it sees a prompt.

I guess what I am saying is: some LLM integration may be worthwhile.

Separately, on a more serious note, what makes this offer that different from its competition?