So basically you just think that it's cool.
Completely misunderstood. It is cool, yes, but the point is that in theory you can make a ridiculously small and simple DAC, and a rather cheap one at that, and push all the effort into the final conversion to DSD, done in a computer with a lot of cycles at disposal and ease of modulator upgrade, since it would be just SW.
You're going to have to turn it back into PCM at some point for DSP...
No, no, again you misunderstood. You store stuff in PCM. Process in PCM. Do all your DSP in PCM. You convert to DSD only at the very end, to perform the conversion.
There is the matter of DSD native recordings. I have no problems in having them converted and even distributed as PCM. But even a native DSD recording can just be considered as a rough PCM recording with two extreme values, which then can be filtered to, say, 90Khz, then PCM processed (including volume control) and converted back. This can be done with no loss.