Nothing here is a summary written after the fact. These are the problem, mechanism,
and move of each article, in the reading order, joined by its own connective lines.
Any one of them opens into the full argument.
Start with the split that decides everything downstream, between the half of a rule a machine can settle from the artifact and the half that needs somebody to read the situation.
Automated checks are written as though every rule were fully decidable from the artifact, and a check that fires on cases it cannot read trains the people who read it to dismiss it.
Mechanism
A rule whose breach is decidable from the artifact alone can be mechanized and failed on, and the half requiring judgment can only be reported, so a mechanized ruling enforces the first half and addresses the second to a person.
Move
Split every rule you are about to automate before you automate it, fail the build on the decidable half, and demote the rest to a note that names a human reader.
Before a check can rule well, it has to read correctly, and the commonest misreading is mistaking an explanation for the thing being explained.
Static checks are run over raw source, and a file that explains why it avoids a forbidden pattern contains the forbidden pattern, so the most carefully documented work is the work most likely to be flagged.
Mechanism
A pattern matcher has no access to whether a string is doing something or describing something, so a check reading raw text charges the same penalty for committing a fault and for documenting its avoidance.
Move
Strip the comments before matching on any check whose pattern could plausibly appear in an explanation, and treat a rule that taxes documentation as a defect in the rule.
Move from what a machine may rule on to what it may ratify, where the cleanest control is a format in which the self-approving sentence cannot be written.
When the same actor may both propose a relationship and record it as accepted, the record stops being a review and becomes an echo, and the control that forbids this arrangement elsewhere is a rule a person can decide to break.
Mechanism
Fixing the status field to a single permitted value and the promotion flag to a single permitted value removes the self-ratifying document from the format, so there is no ratified claim to reject and no override to find.
Move
Give agents a proposal format that cannot express its own acceptance, and keep promotion as a separate human edit in a different file, reviewed like any other.
Ratification aside, machines and people both file findings, and the honest ones grade their own recommendation lower than the observations that led to it.
A report that grades its findings and its recommendation on one scale invites the reader to spend the credibility of the observations on the repair, and most engineering write-ups are built exactly that way.
Mechanism
Every observation is a statement about a state the system was caught in, while a repair is a claim about every state it has not yet been in, so evidence accumulates on the diagnosis and does not carry across to the fix.
Move
Grade the recommendation separately from the findings, in the same table, and let a low grade on your own proposal stand where the reader can see it.
End where deciding goes wrong without anyone being wrong, in a system that let two parts settle the same fact and had nowhere to notice they disagreed.
The same state is often classified in more than one place in a running system, and each classifier is correct given the evidence it accepts, so the disagreement never surfaces as a fault.
Mechanism
Two independent decisions about one fact produce two answers that never meet, so the system never holds a contradiction, only two truths, each locally consistent and consumed by a different half of the layout.
Move
Name one owner for each fact a system acts on, make every consumer read that owner, and treat a second place that decides the same thing as a defect on sight rather than a redundancy.
You have walked What a machine may decide end to end: the split inside a rule, the misreading underneath it, the line at ratification, the honesty of an unknown, and the cost of two authorities for one fact. None of these turned on how capable the machine was. Every one turned on whether somebody had decided, in advance and in writing, which questions it was allowed to close.