... be named Quatermass and he say something memorable like "this planet, this orb, going the wrong way 'round, dooms us all as every planet in the universe will follow. It's only a matter of time before we plummet into the sun."

You don't get much from thermoelectric conversion - in my business of wireless sensor networks you see a lot of offerings. The best stuff in the past couple years generates about 50uW/sq cm for a 5C difference. That's good enough for a wireless sensor hugging a tree, perhaps. Available power goes up for more thermal difference, but it's unlikely that anytime soon either BWM, Adui, or even Fnord, for that matter, will be replacing the inexpensive, reliable and robust automotive alternator with a pricey power-producing muffler. Well, maybe Fnord.

Of course they're software-defined radios - just over a narrow class. All those standards are OFDM modulation (well, except for .11b, but other than that...). Once you have an engine that can do the data rates of 802.11n and the OFDM subcarriers of Mobile WiMAX, everything else is just changing over-the-air parameters and frame structures. And in terms of modifiable, it's unlikely the radios could go out of the assigned frequency bands, and given the state of the highly optimized engines in these chips, you're not going to be able to turn a WiFi radio into, say, a broadcast FM receiver or a cellphone. First, the systems aren't even remotely similar; second, the power consumption would probably be awful; and third, there's no money to be made by the chipset manufacturer in allowing that to happen. But a reasonable chipset manufacturer could certainly publish open driver specs that would allow you write drivers to manipulate timing, bandwidth, frequency, packet size, used subcarriers, power levels (as long as it remains Part 15 compliant) - anything that doesn't touch on the transmitter certification itself.