Define the claim
Separate the testable proposition from the surrounding story, rhetoric and identity cues. If a viral post bundles several claims together, split them before evaluating them.
How we work
Method
A verdict is only useful if you can see how it was reached. Our method is designed so that a motivated reader can reproduce the reasoning without trusting us as an authority.
Find the strongest version
We steelman the argument before testing it. Weak caricatures are easy to defeat and tell the reader very little.
Test the mechanism
Ask what would physically, biologically, historically or statistically have to be true for the claim to work. Then compare that required mechanism with established measurements and observations.
Build the evidence trail
Prefer primary sources, systematic reviews, technical reports, official datasets and standards. Secondary sources can add context, but they should not replace stronger underlying evidence.
Look for disconfirming evidence
Actively search for the strongest serious evidence against our emerging conclusion. If the claim has credible support, present it and explain its strengths and limitations.
Calibrate the conclusion
Use evidence language that matches the strength of the record. Unsupported is not the same as disproven. Inconclusive is not the same as false.
State what would change our mind
Whenever a claim is empirically testable, we state what kind of new evidence would materially change the conclusion. A claim that cannot lose is not a strong empirical claim.
Make updates visible
Material corrections are logged publicly. Evidence changes; our published conclusions must be able to change with it.
Core principle
Don't take our word for it. Follow the evidence.
Governments can be wrong. Companies can mislead. Researchers can make mistakes. Influencers can occasionally be right. The evidentiary standard is the same for everyone.