Part 4 · 4.2 Analytical Rationale
Why this is the general shape, not a press-release peculiarity
How do we define and fill systemic gaps — and why the worked example in 4.5 is an illustration of a pattern, not the pattern itself.
§The nine moves, renamed in standard method
| move (4.1) | standard analytical name |
|---|---|
| 1. state the object and purpose | scoping |
| 2. the machine proposes a method | method selection, with a stated rationale |
| 3. break into questions | decomposition into sub-questions |
| 4. delegate each to a reader | tasking a controlled sub-task, with a brief |
| 5. ask; get a tagged answer | source-bound extraction, epistemic-status tagging |
| 6. attest or archive | quality review and sign-off |
| 7. the finding returns | release into the study's own record |
| 8. synthesise once every branch is in | synthesis under a barrier |
| 9. arrange into a document | write-up / composition |
This is, essentially, the shape any competent piece of evidence-based work already follows, whether the "reader" in move 4 is a person, a research assistant, or — as in this book's own worked example — a language model given a narrow brief. Nothing about the mechanism is press-release-specific; the four vectors of 4.5 happen to be which facts, where, what relations, in what form only because that is what suits studying a company's announcement. A different object gets different questions from a different method on the shelf, and every other move stays identical.
§Why "any programme on the shelf" is not an empty claim
PROGRAMME_MANUAL.md (cited fully in Part 3) names three further study designs built on this same arc, only one of which — the press-release study — has actually been walked live at the time of writing: reading a physical phenomenon by decomposing it into aspects worked under a shared budget; checking a claim against sources with a declared evidence family and a fetch step behind a human gate; and drafting an article by lane, with citations carried by the represent mechanism of Part 2 and versions living as fresh lanes rather than edits. Each of these swaps out what move 3 asks and what move 4's brief says; none of them touches moves 1, 2, 6, 7, 8 or 9 at all — those are the machine's own fixed contract, covered next in 4.3.
§What the epistemic-status tag actually buys, and its honest limits
Tagging every finding declared, happened, or unclear forces a distinction most casual reading skips past silently — a company saying it will do something and a company having done something are not the same fact, however similar the two sentences read. This book's own worked example, examined closely in 4.3, shows the tag is not infallible: a reported number can be marked happened as though independently confirmed, when the material only shows the company itself announcing it — a distinction a careful reader has to catch, because the extraction step has no way to know, on its own, whether "the company said X" and "X occurred" are actually the same claim.