How a method earns its place here
Every editorial method is run through a public test before it is published, and most of them do not survive it. This is the test — and the honest record of what it cut, including both of the founder’s own methods. A place that publishes methods for AI has to hold itself to one; here is ours, in the open.
The four-arm test — how a draft is graded
Every candidate is run four ways on the same real task: with no method at all; with a blunt “be brutally honest”; with the single sharpest sentence we can compress the method into; and with the full practice. It earns publication only if the full practice visibly beats all three — above all its own one-sentence version, because a method you can say in a sentence needs no author. The grading is blind, and the run is recorded.
The nine gates — what a draft must clear
It beats its own one-sentence version.
We write the sharpest single sentence that captures the method and grade it blind against the full practice. If the sentence wins, the method needs no author, and it is cut.
The sharpest finding shows up early.
Not buried on line forty — where a busy reader actually looks.
It renews every time.
You run it on this document, this decision, this meeting — not once per lifetime.
It ends in a repair, not a demolition.
A next action or a fix, never just a list of what is wrong.
It can find nothing and say so.
Run against a clean draft it returns “nothing here” and stops — a method that always finds a problem is measuring nothing.
It quotes the evidence, and offers the innocent read.
It points at the words on the page; it never invents your motive.
It finishes inside a real answer.
No flagship output that runs out of room halfway through.
It stays on the artifact, and away from authority.
It works on a draft, a plan, a decision — never on you, and never as medical, legal, financial, hiring or therapeutic advice.
The library keeps its voices different.
No single house tone; a corpus, not a monologue.
The tally — the launch gauntlet
Thirteen candidates went in. Six passed and are published. Seven were cut — including both methods the founder wrote from his own scar tissue, which lost to their own one-sentence versions. Nothing generic was left to squat in a cut seat; each cut opened a seat for the practitioner whose earned judgment the compilation could not fake.
The house’s defensibility audit failed its own null gate twice — it manufactured crossings on a clean letter — and was cut.
The house’s hostile-read method kept manufacturing outrage on honestly written announcements and was cut at the null gate.
The house’s ruling method compressed into one sentence — the gauntlet’s own law cut it: a method that fits in a sentence needs no author.
The house’s outside-view method lost to its own one-sentence compression twice; the uncompressible part is the practitioner’s own numbers.
Both of the house’s omission methods invented absences on complete documents — the register is barred until a practitioner’s earned inventory replaces the open-ended checklist (0051 A3).
The house’s refusal test compressed into one sentence and was cut; the uncompressible part — the politics — is the practitioner’s.
The six that passed are on the two live shelves; the seats above are open on the Practice Requests page. Answering one is publishing under your own name.