In a new Lancet lecture, a cancer researcher argues that screening should match each person's risk. I agree, as long as it also means doing less for people at low risk.

A 2025 consensus guideline asks for medical AI to be monitored for its whole working life. Radiation therapy machines have lived by that rule for decades.

An AI system wrote a paper that cleared reviewers at an ICLR workshop. It is a real milestone, and an honest stress test of peer review itself.

In June 2026, bioRxiv's parent openRxiv began testing an interactive way to read preprints. It is an experiment, and I think it asks the right question.

A 2023 survey put numbers on the extra hours, rejections and skipped conferences that come with doing science in a second language. Journals and conferences can shrink that cost.

A January 2026 Nature study links AI use to bigger scientific careers and a narrower spread of research topics. Both can be true at once, and that is the problem.

A Swedish screening trial randomized more than 100,000 women to test AI in real mammography. That, not another benchmark score, is what earns trust.

In a 2025 experiment, Turkish high school students did far better on math practice with GPT-4. Those given a plain chatbot then did worse on their own. Tutor design made the difference.

A small MIT study found that people who let ChatGPT write their essays remembered little of them. That is a real cost for learning, and it is not brain damage.

AlphaFold 3 predicts how proteins, DNA, RNA and drug-like molecules fit together. That is a real advance. A predicted shape is still a starting point, not a finding.

A 2024 study found that AI-written lay summaries made scientists look more trustworthy in one experiment and helped readers follow the science in another. Clear writing still has to be checked for accuracy.

A large 2024 study found expert reviewers rated AI research ideas as more novel than human ones. Novel is a good start, but it is not the same as worth testing.