[haywalk]

I’d disagree with this. As local models and compute improve, I can definitely see something like this becoming doable offline in the next decades. In the meantime, at the very least, I would never use something like this unless I could bring my own API/cloud provider so that my data isn’t being handled by the vendor. Also teleoperation is a huge no go for me.

[jurgenburgen]

I get that the voice interface probably uses an LLM but surely the rest of the device isn’t based on LLM tech?

[Schiendelman]

LBM likely - and then that creates rules for a local system to follow.