The method
This is not an opinion blog. Every text follows a fixed discipline, designed so that what is put forward can be checked rather than taken on trust.
The principle
An argument can be very convincing and completely wrong. The force of conviction it carries says nothing about its soundness; what does is its logical validity and the truth of its premises. The whole task is to separate these two things: what wins us over, and what establishes the truth.
The rules
- Steel-man first. Before any criticism, the argument is presented in its strongest form, citing its best defenders, never a version that is easy to knock down.
- Two families, two tools. A subjective argument (a feeling) is tested by asking whether the feeling would be there anyway. A factual argument (a claim about the world or a text) is checked head-on, against the facts.
- The symmetry test. An argument that, if valid, would prove the opposing position just as well establishes neither. We note precisely what this test settles and what it does not.
- Every claim is traced. No affirmative sentence without an identifiable source. Sources are primary or academic references, with a verified identifier (DOI) and a check on authorship, not blogs or loose quotations.
- Falsifiability. For every fact put forward, we state what would count against it. Our own explanations are held to the same standard as the ones we criticise.
- Symmetrical scepticism. Sources that suit the thesis are examined as strictly as those that get in its way. A strong claim resting on a single source is flagged as such.
- The tone follows the proof. No absolute formulas, no rhetorical flourish. A modest conclusion stays modest.
The guardrails
- Showing that an argument fails does not prove that its conclusion is false. Refuting a bad proof of a thesis leaves the underlying question open. These texts do not conclude on the existence of God, nor do they accuse anyone of fraud.
- No single target. When an example serves as a thread, the same criticism is shown to hold for neighbouring positions. The aim is not to target someone, but to examine a piece of reasoning.
In practice
Every text is produced by a reproducible chain: an argument is broken down into atomic claims, each one receives a dedicated search and a strength-of-evidence rating, then the writing ties each sentence to its source. A final quality check verifies that no claim is left unsupported and flags the points to watch.