It's very different from academic research, yes. But the demo scene produced things that were at the very limit of what computers were capable of at the time, a

It's very different from academic research, yes. But the demo scene produced things that were at the very limit of what computers were capable of at the time, and they were/are every bit as 'innovative' as some of the top academic research. Some of the techniques used would be top contenders at CS graphics conferences around the time they were produced. That were 'published' as executables rather than papers, but that doesn't make them less 'innovative'. That they were produced by people working on them part time doesn't make it any less innovative in my view, nor were the results worse than those produced by full timers (in academia or elsewhere).