I was only able to watch 70% of the video because of streaming problems..
Anyway, I have a couple of questions. Why was it decided to use byte-code as opposed to assembly instructions generated by the host machine.
The problem is, you have a high speed processor on the host system and very low speeds on the actual device, so why not shift the load to the big machine by having it generate the 'final code' that is executed as opposed to having the device do (byte-code > Assembly)?
If it was me, I would have written a Micro-kernel that has the ability to start executing a single set of instructions but supports some basic form of multi-threading. Such a system would mean you need minimal memory and CPU power and thus the device should consume minimal power.
This is a very odd question, but at the start of the video you have the interview.. there is a guy behind him working on his laptop.. does he have a web-cam type device stuck to the top of his laptop screen? It just looks like he does..