Science is built on prior knowledge, but that knowledge is buried across millions of papers. Priors is building an open public registry of the claims each field rests on: each one weighed against the evidence, graded for certainty by a fixed AI protocol, reviewed by experts, and updated as new evidence arrives.
A living review
This is the review for GLP-1 receptor agonists: 35 claims, ordered from most established to most contested. Its overall grades have been reviewed by three independent experts.
| Claim | Standing | Certainty | Papers |
|---|---|---|---|
| Reduce major adverse cardiovascular events in type 2 diabetes with established cardiovascular diseaseCardiovascular · GLP1-CV-06 | Established | 90% | 27 |
| Tirzepatide outperforms single GLP-1 receptor agonistsMetabolic · GLP1-MET-02 | Established | 89% | 20 |
| Semaglutide 2.4 mg achieves 15% body weight lossMetabolic · GLP1-MET-04 | Established | 88% | 21 |
| Reduce all-cause mortality in type 2 diabetesCardiovascular · GLP1-CV-01 | Likely | 84% | 22 |
| Renal protection and slower chronic kidney disease progressionMetabolic · GLP1-MET-01 | Likely | 84% | 20 |
| Reduce neuroinflammation (microglial suppression)Neurodegeneration · GLP1-NEU-05 | Uncertain | 65% | 19 |
| Reduce opioid and stimulant useAddiction · GLP1-ADD-04 | Doubtful | 45% | 21 |
| Improve cognition in established Alzheimer's diseaseNeurodegeneration · GLP1-NEU-02 | Refuted | 14% | 19 |
8 of 35 claimsSee all 35 claims →
55 fields graded
Every field is open to read. Search or browse to yours, and track it to hear when it has been reviewed and when a grade changes.
What is Priors
Each major claim in a field is stated precisely using PICO: the population, intervention, comparison and outcome it is about. Four AI reviewer roles then grade it from 0 to 100 against one fixed evidence rubric, with five standings from Refuted to Established. Every grade shows the full corpus of papers behind it, claims are re-graded each quarter and when a pivotal trial reports, and every review is open.
Retrieval + audit
Assembles the corpus and audits it: study designs, what was excluded and why, where the evidence is thin.
Weighting
Reads what the body of evidence shows, and how tightly the studies converge on it.
Adversarial
One reviewer whose only job is to attack the finding: the strongest objection, and what would resolve it.
Verdict
Weighs the objection against the claim and sets the certainty. That number is the prior.
Inside one claim
Each claim is graded against the published literature, and every step behind the grade is shown. This is one claim from the living review for GLP-1 receptor agonists.
01
One grade from 0 to 100 on a fixed scale with five standings, so a claim in one field means the same as a claim in another. When a score sits close to a boundary, the review says so.
GLP-1 receptor agonists reduce all-cause mortality in patients with type 2 diabetes compared to placebo or standard care.
84Likely
Well supported, with some gaps or indirect evidence.
02
Each paper in the corpus is read against the claim and classified by what it does to it. The sentence that decides it is quoted, and every DOI comes from the corpus the claim was graded on.
RCT · n=9340 · primary evidence
“The rate of death from any cause was lower in the liraglutide group (381 patients [8.2%]) than in the placebo group (447 [9.6%]) (hazard ratio, 0.85; 95% CI, 0.74 to 0.97; P=0.02).”
RCT · n=3183 · primary evidence
“Death from any cause occurred in 23 of 1591 patients (1.4%) in the oral semaglutide group and 45 of 1592 (2.8%) in the placebo group (hazard ratio, 0.51; 95% CI, 0.31 to 0.84).”
RCT · n=3297 · supporting context
The abstract gives cardiovascular death ('similar in the two groups') but reports no all-cause mortality figure.
03
One reviewer’s only job is to attack the finding. Its challenges are published beside the grade, with what remains uncertain.
Mortality endpoints secondary, not multiplicity-adjusted
In LEADER [1] all-cause mortality is a prespecified exploratory endpoint with no multiplicity adjustment ('No adjustments for multiplicity were performed for the prespecified exploratory outcomes'); PIONEER 6 [4] and SUSTAIN-6 [2] were noninferiority designs not powered for mortality superiority, so the RCT mortality signals are secondary rather than primary findings.
Largest semaglutide trials lack independent mortality benefit
The two largest semaglutide T2D CVOTs do not themselves establish a mortality benefit: SUSTAIN-6 [2] does not report all-cause mortality and found CV death 'similar', and SOUL [6] states secondary outcomes 'did not differ significantly' — weakening a class-wide claim.
04
The panel records the gap that would most move each claim and the study that would close it, so you know what to watch for. Track the field and we will tell you when the evidence changes.
The gap that would most move it
No trial powered primarily for all-cause mortality
The study that would close it: A placebo-controlled RCT, or a prespecified pooled analysis, with all-cause mortality as a primary multiplicity-adjusted endpoint in type 2 diabetes.
Open question
Does a GLP-1 receptor agonist reduce all-cause mortality when that endpoint is the primary, multiplicity-protected outcome?
05
Independent experts check the evidence behind each field; a reviewer corrects the corpus, never the score. Claims are re-graded each quarter, and between releases when a pivotal trial lands, and every change is a new version.
Overall grades reviewed by 3 independent experts · Oct 2026 · release 2026-q4
Borderline: within 4 points of the Established/Likely boundary, so a re-grade could land either side.
Limits
Priors applies a systematic, standardised protocol to the key claims in a field. It is not a replacement for the definitive systematic reviews produced by bodies like Cochrane. Our protocol is built for scale, not depth.
Think of it as a reliable first pass that keeps you current and gives you a foundation to build on. Not the final word, and not a substitute for the review you would write yourself. We don't cover every claim in a field, and when the evidence is thin, Priors says so. That is the point.
A sceptical first reader. Not peer review, and not medical advice.
Release notes for your fields each quarter, an alert when a pivotal trial moves a claim, and word when your field has been reviewed.
Track your field →Free. No account; unsubscribe by replying.