The rules we hold ourselves to

The Legibility Index (v1.0) scores the argument and the rhetorical move on 3 axes: Unfalsifiability, Motte-and-bailey, False legibility. It never scores a person's character or motives. A number attached to a claim is an opinion grounded in the disclosed evidence beside it, not a verdict on a human being.

Every episode states the strongest good-faith version of the subject's argument before any critique, in its own block the subject could forward. An episode page cannot ship without that steelman populated. If we cannot make the argument well, we have not earned the right to take it apart.

Factual claims become receipts. A computed receipt re-derives its figure live in your browser and in CI; a sourced receipt carries a citation that fails the build if the source drifts. We can only compute what a subject sourced, so most receipts are cited rather than recomputed, and we say so. "Unsupported" is our default verdict, never "lying."

Clips are short excerpts quoted under commentary, with a source block and an outbound link to the rights-holder's own player. Our feed and radio carry only our own commentary audio. This is an editorial guideline, not a claim of legal immunity: the real protection is fair, transformative commentary and careful subject selection.

The Legibility Desk is a firewalled sibling of Between Systems, the interview show. We never put an interview guest, or anyone plausibly in our outreach graph, through the rubric, and we never invite a critique subject onto the interview show as if nothing happened. Selection is the real firewall: we critique already-public output from large platforms, not peers.

Any named subject gets the last word, verbatim and unedited, at the top of the page, plus a standing invitation to defend the argument in conversation. Critique in public, and mean the invitation.

We skip it when the subject is a private individual; when the claim is about someone's personal life rather than a public argument; when the target has less reach than we do and would be harmed disproportionately; when we cannot source the correction; when the only available framing is mockery; when the likely outcome is a pile-on; or when the question is genuinely contested among experts.

The show opens by running its own rubric on its own essays and resolving predictions. Fallacy cross-cuts (one systems fallacy, several short clips) and historical or already-resolved cases come next. Critiquing a named, currently-active figure is a graduation we earn only after the guards are boring, and only with a designated takedown process and outside review of this charter in place first.