Three kinds of work

Three kinds of work sit behind these properties, and the reason to keep them apart is not tidiness. They carry different obligations and they fail in different ways. Research is obliged to publish the question it could not answer. A forum on responsible practice is obliged to write rules that bind the party paying for the room. A product is obliged to work on a Tuesday, in the period after this one, in a room of fifty, on a device shared between three students. That is a narrower and far less forgiving test than either of the others.

What goes wrong when the three are merged is specific. Research that must support a sale has its finding decided before its question is asked; the sentence we do not know becomes commercially expensive and stops being written. A forum that must protect its convener’s revenue writes guidance the convener already satisfies; its silences then map onto the convener’s weak points, and the guidance works as a moat rather than a floor. A product built to demonstrate a research thesis ships the argument instead of the thing a teacher asked for, and the classroom pays for the argument.

The separation is therefore a set of prohibitions rather than a description of character. Research here does not clear a product for sale. The Council certifies nothing PrepGraph makes. Deployment confers no authority: that a system is running in schools is not evidence that it is right. The formulation quoted above states what each kind of work owes. It is not a report that each is discharging the debt, and in the Council’s case there is nobody yet to discharge it: its founding instruments were published as drafts on 29 August 2026, no voting members are in office, and no ethics review panel has been appointed.

What each property is, and whether money changes hands. Nothing below describes what any product does. That belongs on pages where it can be read as a product claim and argued with on those terms.

The honest count is five out of five. The platform, the Academy and the partner portal are commercial surfaces and read as such. This site is not commercial, but it is written, funded, staffed and edited by PrepGraph and argues for positions PrepGraph holds. It claims no independence and should be discounted exactly that much. The Council is convened and paid for by PrepGraph as well. A forum funded by a company selling into the same schools is structurally exposed, and any page telling you otherwise would be the first symptom of the problem.

What is offered instead of innocence is legibility. The funding is stated rather than buried, and the terms are written down and published: who may sit and who may not, what the forum may and may not issue, how a member records dissent, and what happens when the convener wants a particular outcome. Those terms are founding drafts, issued on 29 August 2026 before the Council was constituted, open for comment, and subject to reaffirmation, amendment or withdrawal by the voting members once seven of them are in office. Written rules can be read by a stranger, checked against behaviour, and found to have been broken. Goodwill can only be asserted. These particular rules bind nobody today, because there is not yet anybody to bind.

Nothing here has been tested by a case that cost PrepGraph something it minded losing, and no such case can arise before the Council is constituted and a first study is under way. Neither has happened. The boundary is a design, and the only use a reader has for a design is to hold it against the record when a record exists.

The design is falsifiable. If the Council issues anything that functions in practice as an endorsement, if PrepGraph’s own products sit permanently outside what it examines, if the published terms are amended without record, or if dissent never once appears on the record after members take office, the boundary has failed. The honest description then is that it failed, not that it was always aspirational.

This page exists so that a reader arriving from the Council can place what they are reading. The register here is argument and open problem. Where a research page would ordinarily reach for a figure, this one reaches for a question and says when the question is still open.

No result is reported here, and none has been reported elsewhere either; the record to date is empty, and saying so is worth more than a figure would be. The policy does not change, because the remedy is not a number on a page like this one. It is work published with its method attached, where someone can attack it properly.