Yeah! Don't these people know how much an atom cost nowdays?

I can't see why simple molecular structures would be more difficult to simulate. You can't simulate something complex like martensite at an atom level, because in order to take enough to get a fair mean effect (many local defects etc), you'd have way to many atoms.

Open source software is sorting itself fine without such supervising. Many cases have shown that the journals (i.e. one or two reviewers) aren't particularly good at taking out liars. I don't see how journals would be less biased towards less experienced academics either. What part of the journal system promotes that?