I would say that such a platform would run our own binaries. The more I think about 6502 emulation the more I think it'd be pretty tricky due to the external parts (ANTIC, POKEY, GTIA, etc.)

I wonder... I don't know their respective specs, but how hard would it be to emulate each of these chips in a -separate- 8052? I mean, we replace the interconnections between all the chips near-exactly, but instead of dedicated chips we use "generic" '51s. So you run a 6502 program through the abstraction layer on '51 "CPU", it makes some call to ANTIC. Okay, data is sent over the bus just as if ANTIC was there, on the other end. But there's another '51 on the other end, which does exactly what ANTIC was doing, and responds in the same way ANTIC would, emulating that chip. The "physical" (video) output might differ a bit, but the "logical" (response to others computer internals) would be exact copy of ANTIC response. And then you send a special opcode that switches it to '51 native mode, at which it leaves emulation and starts doing things -our- way, never caring for the abstraction of emulation and just performing at maximum speed, or "generic" opcode at which it shuts the display down and starts executing program from common memory, just like a CPU, doing some paralell processing, until it switches back to some other mode.

BTW, Atari vs Commodore. I used to be a fanatic Atari advocate in the times of the holy war. But now I've read the C64 manual and in the perspective I must admit Commodore was superior. Atari was more of "toy", "game console", hiding internals from the user, making all the "heavy wizardry" more difficult. Commodore was more "open architecture" allowing really deep hacks more easily.