Contested questions, worked through claim by claim. We show the sources, the scoring and the arithmetic, so you can disagree with us on the record rather than guess how we got there.
Published assessments, newest first.
whether AI deskills clinicians — Is there real evidence that using AI makes skilled people worse, and who does it happen to?
There is a widely held sense that using AI is making us worse at thinking, and that there is now real evidence for it. The evidence usually offered is weak. The best evidence anyone has is real, and almost nobody outside medicine has met it with its qualifications intact.
Read the assessment → published 30 August 2026 · not revised · 1 correction · reviewed before publication 30 August 2026 · sources checked 3 September 2026 Issue twoa guideline that grades one drug above two similar ones — A guideline grades one of three similar drugs higher. Does that mean the drug is better?
A guideline lists three drugs of the same class as preferred and grades the evidence behind one of them higher. That grade is routinely read as a ranking of the drugs. No trial has ever compared them, and this is what the evidence behind the grade actually shows.
Read the assessment → published 28 August 2026 · revised 13 September 2026 · 20 corrections · reviewed after publication 29 August 2026 · sources checked 12 September 2026 Issue onean mRNA cancer vaccine result — A large trial was announced as a success. What did it actually release, and what will that support?
Merck and Moderna announced that a large trial of a personalised mRNA cancer therapy succeeded in melanoma. The announcement released no numbers from that trial. This is what was actually published, and what it will and will not support.
Read the assessment → published 28 August 2026 · revised 13 September 2026 · 15 corrections · reviewed after publication 8 September 2026 · sources checked 11 September 2026The short version. The long version is in the policy.
What Holds Up is published by CivicScale, which sells healthcare software and takes no outside investment. That is a real conflict, not a theoretical one, and the honest thing is to put it where you will see it rather than in a footer.
No customer, commercial partner or subject of an assessment sees one before it publishes. The full position — what money can and cannot buy, and how often our published assessments turn out to be wrong — is in Who Pays for This.
Assessments are free to read. When the evidence on one of them moves, subscribers hear first — what this is.
Every issue carries a one-click unsubscribe link. We send nothing else, and the list is not shared.