Journals must require source code and data sets to publish

It is understandable that you might not be able to reproduce a paper that does a natural science experiment, given that physics is brutal.

But for papers that have either source code or data sets, academic journals must require that those be made available, or refuse to publish.

Any document without such obvious reproducibility elements is a white paper, not a proper peer reviewed paper.

Big companies like Google are able to publish white papers as peer reviewed papers just due to their reputation, e.g. without giving any source code that is central for the article.

It is insane.

E.g.: AlphaGo is closed source but published as www.nature.com/articles/natnure16961 in 2016 on Nature.