That is indeed a fair point: Why do people go to some Elsevier journal to look for an article instead of just going to the corresponding arxiv.org section? Because there is trust in the curation of those articles.

The problem with scientific publications is that there has not been a Spotify, Steam or Netflix disrupting company that provides the same service in a better way.

EDIT: Thinking more about it, I think such a service would be fair to say have a raw collection of journal articles (it could even be based on Arxiv) and charge for the "curation" layer on top of it. Now the only question is how to kickstart that curation layer in a matter that is trusted by scientists. Maybe it could be something more distributed where also curators (reviewers) and writers get some profit.