Peer-reviewed versus published instantly - the difference that decides whether you believe the headline.
Nature has been published since 1869 in London, founded by the astronomer Joseph Norman Lockyer. It has worked on the same principle ever since: a scientist submits a paper, an editor decides whether it's even worth reading further, and if so sends it to other specialists in the same narrow field, people with no connection to the author, who read every step of the research looking for holes. The process is called peer review and takes months. Nature accepts fewer than eight out of 100 submitted papers. The rest don't disappear, they just don't pass the check.
arXiv is something entirely different. It was created in 1991 by physicist Paul Ginsparg while working at Los Alamos National Laboratory (an American nuclear research lab in New Mexico), as a simple way for physicists to swap drafts before official publication. It doesn't go through review. A scientist uploads a file, the paper appears publicly the next day. Until recently the archive was run by Cornell University (an American university in New York), and since mid-2026 it moved to an independent nonprofit organization. No fee, no waiting, no filter.
That's why the word 'preprint' matters: it's the paper before print, the version the author thinks is ready, but no independent specialist has confirmed yet. In math and physics this culture works well, because the community is small and mistakes surface fast in the comments under the paper. The problem is elsewhere, in fields where the topic is hot (artificial intelligence, medicine, climate), a preprint from arXiv reaches the media within hours, while the real check takes months, if it happens at all.
This is where the distortion I see constantly is born, in the news about AI. A lab drops a paper on arXiv with impressive results, a journalist reads the headline, writes 'scientists proved', and the reader takes something unverified for an established fact. The word 'research' itself sounds equally weighty in both cases, but the weight behind them isn't equal.
Here's the trap
I follow arXiv myself constantly, because that's where new AI stuff comes out days or weeks before anyone writes it up in plain terms. But I learned to first read who the author is and which lab stands behind him before believing the result - peer review isn't sacred, but without it you have nothing else to stand on except reputation.
The difference between the two isn't a hierarchy of 'better' and 'worse'. Nature is a slow, conservative filter, it lets revolutions through sometimes years late. arXiv is fast and raw, it brings you tomorrow, but a pile of noise along with it. The healthy habit is to ask which of the two you're reading before deciding what to believe.