Hacker News Reader

For a big result like this, enough experts read the paper to catch glaring mistakes. Sort of a de facto peer review

Exactly, Quanta would have the experts they interview go over the paper first if they didn't already go over it before and evaluate it before making any comments.

Plus arXiv papers are freely accessible ;)

Reporters often rely on social input to decide what to cover—when someone they trust tells them, “you should check this out.” Then there is an exploratory phase where they gather some basic facts and reach out to contacts to see if it is in fact newsworthy, and maybe to start gathering quotes to add context and fill out the story. At some point they say “this is real” and start working with a editor on a schedule to publish it.

Whether a paper is on a preprint serve is not a big deal. If a lot of experts are excited about it, that’s a story anyway.

And if it falls apart later due to some surprise flaw… well that is also a story!

Publications don't check facts. They just check formatting and interestingness and plausibility. Peer review is totally separate, not part of publication's pseudo "peer review".