The publishing discipline

Trust is the only currency this kind of work runs on, and trust is built by procedure, not by sincerity. Here is the publishing discipline that everything on our shelf follows. It’s boring on purpose. Boring is what checkable looks like.

Every claim carries its receipt. A number without a source is an opinion wearing a costume. Before anything publishes, every load-bearing claim traces to a source a stranger can check — and an independent pass re-fetches the sources and checks each claim against them, done by someone who didn’t write the draft. The writer never verifies their own work; wanting a claim to be true is exactly the condition under which you shouldn’t be the one checking it.

Targets are labeled targets. Ten thousand conversations is a goal, not a count. The count is whatever the counter says. The fastest way to lose a fact-checker’s respect — and deserve to — is to let an aspiration wear a statistic’s clothes. Scenario, target, estimate, verified count: four different things, labeled as four different things, every time.

What isn’t verified yet says so on its face. Anything not yet trustworthy is flagged as exactly that, in public, rather than quietly published and hoped about. An honest “not verified yet” outperforms a confident error every single week.

Corrections are a front door, not a shame ledger. A standing public corrections page; anyone can file; fixes propagate to every page carrying the same fact, and the correction stays visible. Skeptics who find real errors are thanked and asked to go harder. A movement that can’t survive its fact-checkers doesn’t deserve to. Publish your failures in the same typeface as your wins — August audit season, every year, on purpose.

Machines are disclosed. AI does substantial work across this project — research assembly, drafting, cross-checking. That’s stated plainly on a standing methods page, along with what stays human: judgment, rulings, the decision to publish anything at all. Undisclosed machine work would be a receipt withheld, and the discipline doesn’t allow those.

The author’s name is the one thing withheld — and that’s the same discipline, not an exception. Everything here is designed to be checkable without trusting anyone, and that includes the founder. A named founder becomes the story, the target, and the single point of failure all at once; an anonymous steward whose work carries its receipts can be judged only on the work. The founding steward writes under a pen name, claims no authority, holds no office, and is designing their own exit — every role names its successor and the condition that hands it over. If the project only worked while one particular person ran it, it would have already failed the only test that matters. Identity restraint protects one more thing: everyone else. No lists of who attended what, no personal data beyond what a signature form needs, nothing held that would hurt someone if it leaked. Institutions get criticized with evidence; people don’t get scapegoated, including the people who run the institutions.

Copy the whole discipline or don’t copy it at all — receipts without corrections is a brochure; corrections without receipts is a suggestion box. Together they’re an immune system.