Three of Spades: Reviewing the Grading Metric

Overview

The Ace of Spades has the founding members write a grading metric and send it upward once. That is a founding act, and a founding act is not a check. A metric written once and never revisited does not stay still; it goes on measuring what its authors could measure at the beginning while the craft moves underneath it, and its meaning drifts in practice where its text cannot. Left alone long enough it becomes the property of whoever interprets it, without anyone having amended a word.

This technic turns that single act into a standing office: the metric deliberated on a rhythm, its effects read and not just its wording, its revisions made prospectively, and the network’s review of it answered in the open. It does not replace what the Ace does. It is what keeps the Ace’s work honest after the founding is over. This technic is graded by its own standard like any other.

How it works

The office exists because of one asymmetry. Whoever can revise the standard can revise it toward themselves, and they can do it in complete good faith, because the change that favours you is also the change you find most obviously correct. No amount of integrity in the reviewers answers this, so it is answered structurally instead. Changes are prospective. A revision takes effect from a stated date, binds no grade already in flight, and never reaches backwards. A reviewer with a grade pending stands out of the sitting. Reviewers are drawn by lot from those who did not author what is under review, wherever the pool allows it. None of this assumes bad faith; it removes the situation in which good faith would have to be trusted.

The office reviews the standard and never applies it. That separation is made structural later in the suit, but the seed of it belongs here, because the shortest route from a body that decides what a grade means to a body that decides who gets one is convenience. A body holding both corners has the whole power, and it will not feel like power while it is being assembled.

What is reviewed is the metric’s effects as much as its text. A standard can read perfectly and work badly: it can be advancing the fluent over the competent, stalling the people whose work is hardest to display, or drifting toward measuring alignment despite having been built so that it could not express alignment at all. So the review reads the grades actually given, the causes actually published, and who has moved and who has not — measuring conditions rather than virtue, turned on the cell’s own instrument.

The second thing the review catches is amendment by reinterpretation. Most standards change without anyone changing them: a reading hardens into practice, then into precedent, and the text stands unaltered while what it means has moved. So anything the metric is being read to mean but does not say is either written into the text or dropped. Practice does not amend a standard.

Network review is then answered rather than merely received. The cell publishes what it changed, what it declined to change, and why. Refusing the review is not open to the cell, because that would cancel the self-application clause; refusing a particular recommendation, with the reason published, always is. The network reads how a cell grades. It does not grant the cell permission to exist, and a cell that starts treating its approval as a licence has installed the centre its constitution forbade.

Excluded by name: no retroactive revision; no amendment by practice; no reviewer sitting on a standard they will be graded under before it changes hands; no licence-to-operate; and no criterion that can express alignment, allegiance, opinion, or way of life, which the floors beneath the Ace already forbid and which every revision must be checked against again.

Operation

1. Convene the review on a stated rhythm rather than on complaint. Fix when it sits before anyone knows what it will consider.

2. Seat the reviewers by lot from those who did not author the metric under review, and stand down anyone whose own grade is pending.

3. Read the effects before the text: the grades given, the causes published, who has advanced and who has stalled, and whether the measure has begun to express alignment despite being built not to.

4. Read the metric against the floors it may not fall beneath — a grade as a verified record of demonstrated competence and conduct, lowered only for demonstrated failure of competence or honesty, with the cause published and the appeal open.

5. Name the interpretations that have hardened in practice. Whatever the metric is being read to mean but does not say is either written into the text or dropped.

6. Decide the changes in the manner of Deciding Together, and make them prospective: effective from a stated date, binding no grade already in flight, never reaching backwards.

7. Publish the revised metric with what changed and why, alongside what was proposed and declined.

8. Submit the revision upward for network review under the self-application clause, and answer what comes back in the open — changed, declined, and the reason for each.

9. File the sitting as a record whether or not anything changed. A review that changed nothing is a finding; a review that left no record is indistinguishable from one never held.

Cautions & failure modes

The revision that lands on its author. Someone will propose a change while their own grade is pending, and will mean it honestly. The prospective rule and the stand-down exist so that nobody has to judge whether they meant it honestly.

The office that starts grading. The body deciding what a grade means is one convenient step from deciding who gets one, and the step is usually taken to save a meeting. It puts two corners of one power in the same hands and will not feel like power while it happens.

Amendment by practice. The most common way a standard changes is that nobody changes it. Meaning drifts sitting by sitting, the text stands, and the cell ends up governed by a metric it never wrote and cannot point to.

The metric that measures everything. This is the failure that arrives from running the technic well. Every good review finds something the standard failed to capture, so every good review adds a criterion, and a metric that grows at each sitting ends up measuring a person rather than a competence — arrived at by a long series of reasonable additions. Ask at every review what can come out, not only what should go in.

Network review read as licence, or as optional. Treated as permission to exist it becomes the centre the constitution forbade; treated as optional it cancels the self-application clause outright. The review is owed and answered, and it grants nothing.