Comment by Ostatnigrosh

It's a bit more complicated than that isn't it? Every time you remove a bottleneck things advance until a new one appears, right? Previously ideas were cheap, good ideas were hard to come by, accurately distinguishing between good and bad without an implementations was an incredibly rare skill that many people wrongly believed themselves to have (the same as how everyone naturally thinks that their preferred language is the most powerful, etc, etc), and implementation was expensive.

Now you can distinguish sufficiently simple ideas by means of slop, but accurately determining what will eventually mesh well with the wider world is still extremely difficult.

There's also a complexity ceiling. It's been raised but it's still there. For example it will presumably be a while yet before a feature complete project on the scale of LLVM can be quickly slopped together.

It's also cheaper than ever to clone a piece of software once someone else has done the hard work of figuring out the general shape it should take. So I'm not sure how viable services without strong network effects will be going forward.