A 2024 study reran thousands of notebooks from biomedical papers, and few gave the same answer. Sharing code is the first step. Making it run again is the real work.

A 2023 Nature study found that science is breaking with the past less often. Critics call the yardstick flawed. Either way, it is worth asking what we reward.

Some cancers found by screening would never have made a person sick. This is called overdiagnosis. It is real, it is hard to see, and it does not mean screening is useless.

In 2022, eLife announced it would stop saying yes or no to papers after review. Public reviews and a short verdict on the evidence now travel with the paper instead.

In October 2024 a scholarly society told members that, according to its publisher, its journal content was being made available for AI training. The deal may be lawful. Authors still deserved to hear first.

Each year, a list of the world's top 2% of scientists makes the rounds. Its own authors call it a measure of citation impact and warn against leaning on it.

A 2023 experiment showed ChatGPT makes professional writing faster. Science should be careful about treating that speed as progress.

A 2024 Nature study showed AI models degrade when trained on their own output. Science is not collapsing, but it should start keeping track of what humans wrote.

In January 2023, two leading science journals ruled that ChatGPT cannot be an author. The real reason is what happens after a paper is published.

In 2023, seven popular AI detectors labeled more than half of a set of TOEFL essays by non-native English writers as machine-made. A detector score can start a conversation. It should never end one.

A 2024 dataset of publisher price lists shows what open access can cost authors at list price. Free reading is a real gain, but the fee should not decide who gets published.

In 2023 the whole editorial board of NeuroImage quit over fees and started a rival journal. Elsevier kept NeuroImage running with an interim board. The reputation was harder to keep.