Ah. Well, I had trouble imagining how it would both read everything that's going on inside the head (millions of neurons firing, often very subtly) without accessing and reading the 'output' via the brainstem. It's the difference of letting a big computer do the calculations itself, and then reading (and interpreting) the output, or trying to follow all the individual calculations and steps it does and try to derive the outcome from that. And then you also need to stimulate very specific parts of the brain, some of them rather deep inside, in a very precise yet quick way, again without tapping into the natural roads that link directly into that area. Though, maybe you could explain it as the mmi reading and stimulating the brain stem/spinal cord input/output. Not sure how it'd do that in a precise manner without actually connecting to those nerves like a normal braincase does though, but if you wanna handwave something like that, and pw lets you, then sure.
And yeah, implanting something like that in the field sounds iffy. Stuff like this would always need quite an adaptation time anyway, to learn the specifics of the host brain (brains can be quite different. Brain plasticity and such) so even if you could install it, you couldn't really use it (without an even bigger helping of handwavium, that is). I have similar concerns for the robolimbs that would be attached in the field. Though there the mmi could actually be handy to explain it, just say it's the mmi logging into the robolimb wirelessly and using it. Maybe talk to kri about it, and ask him how he was gonna have his robolimbs be controlled? Cause if I'm right, and you can't just plug something into a bunch or shoulder nerves and expect it to work, that'd be a great reason for fleshy people to get an mmi.
I'll wait with the other potential uses for later, when I know more about how pw handles these sort of things.