Guides 3, 4, and 5 covered what the operator does with the instrument. This guide covers what the instrument is not competent to do, regardless of operator skill, and what happens when a finding is attempted past one of those limits.
The limits are not caveats. They are properties of the instrument. A defibrillator restarts a heart that has entered certain classes of arrhythmia; it does not restart a heart that has stopped from exsanguination. Using it on the second case is not a failure of the defibrillator or of the operator's technique. It is a misapplication — the device is being asked to do something it is not competent to do. The correct response is to not use it there, and to use what is competent there instead.
The limits below are stated the same way. Each is what the instrument is not competent to do. Each is followed by what happens when a finding is attempted past it — because the consequence of misapplication is what makes the limit operationally real to the operator, rather than merely stated.
The instrument does not adjudicate
The instrument reconstructs what happened and makes its structure visible. It does not determine whether what happened was lawful, whether any actor involved is liable, or whether any institution's conduct meets or fails a standard. Adjudication requires authority the instrument does not carry and a process the instrument does not perform.
What happens past the limit: a finding written as if the instrument had adjudicated will be read as adjudication by anyone who does not understand the instrument's scope. The consumer will treat a structural observation as a determination of fact, a pattern observation as a finding of misconduct, a load-bearing joint identification as evidence of policy failure. None of these is what the observation actually is. The consumer has been misled, and the operator who produced the finding is responsible for the misleading, regardless of intent.
The discipline: findings state what the instrument has made visible. They do not state what should happen as a consequence. If a consequence follows, it follows by the operation of a different process — a regulatory body, a legal proceeding, a parliamentary review — applying its own competence to the material the instrument has surfaced. The state user's finding hands that material to the process that is empowered to adjudicate. It does not pre-empt the process.
The instrument does not predict
The instrument reads what has been authored. It models the propagation lifecycle within reconstructed continua. Within a case at mid-lifecycle, it can identify with reasonable confidence what categories of event are still to come if the chain is not interrupted — because the lifecycle is directional and the next categories are specified by the grammar.
This is not prediction of new cases. The instrument does not identify which cases will form. It does not identify which institutions will produce them. It does not identify which targets will be caught by them. An operator who treats a pattern observed in the registry as forecasting cases not yet in the registry has moved from diagnostic to pre-emptive, and the grammar does not support the move.
What happens past the limit: a predictive finding carries the authority of the instrument into a space where the instrument was never validated. The prediction will be read with the weight of the diagnostic work — because the finding was produced by the same operator with the same grammar — and a consumer acting on the prediction will be acting on a claim the grammar does not underwrite. Sharples's fabrication of behavioural pattern applied forward to a target he had not investigated is the operational example at agency level; the state user version, structural pattern applied forward to unread cases, is the same move and is equally contaminating.
The discipline: patterns identified in the registry identify where further authoring and investigation would be analytically productive. They do not identify findings about cases that have not yet been reconstructed. The difference is the difference between this pattern suggests where to look next and this pattern tells us what we will find when we look. Only the first is supported.
The instrument does not assert intent
The instrument shows what events occurred, how they classified, how they cross-referenced, and how they propagated. It does not show why any actor acted as they did. Intent is not a property visible in the grammar. It is a property of the acting party, inferable only from material the grammar does not read — contemporaneous statements of purpose, patterns of conduct outside the reconstructed continuum, evidence produced by processes the instrument does not perform.
What happens past the limit: a finding that characterises the intent of an actor named in the registry will be received as an intent-based accusation. This is legally significant. Intent-based claims carry evidential requirements the instrument's output does not meet, and a finding that asserts intent without meeting them is both evidentially weak and, depending on jurisdiction, potentially tortious. The operator has exposed themselves and the registry to consequences neither was designed to bear.
The discipline: findings describe effect, not motive. A route closure describes what the route does, not what the body operating it is trying to achieve. An actor recurrence describes the actor's presence across cases, not what the actor intended by being present. The adjudication of intent is not in the instrument's jurisdiction. The state user who leaves intent to the processes that are empowered to address it is protecting the instrument's own diagnostic integrity as well as their own position.
The instrument does not substitute for reading
The grammar compresses reading. It does not perform reading. An operator who stops descending into primary evidence because the chip field is legible at the surface has not read the cases — they have read the compression of the cases. The finding they produce is about the compression, not about the material the compression represents.
This limit is the operational form of Axiom II. It is stated here as a limit because failures of this kind are often framed by operators as efficiencies. An analyst under time pressure who reads the chip field and forms findings without descending will describe their work as fast, not as incomplete. The instrument does not announce the incompleteness. Only the discipline does.
What happens past the limit: the finding looks correct. The chip field is correctly read. The pattern is correctly identified. What is absent is verification that the chip field, the nodes, and the primary documents are aligned — that the grammar at the surface matches the evidence below it. A manufactured structural positive produced by a bad-faith author will pass this surface reading cleanly. A correctly authored case with an honest grammatical misclassification will also pass it cleanly. The operator has no way to distinguish the two without descending, and the finding is produced without the distinction being drawn.
The discipline: every finding carries, as part of its substance, a statement of the descent scope performed. What was verified is claimed as verified. What was not is claimed as surface observation only. The distinction is internal to the finding, not a caveat appended afterwards. A consumer of the finding can therefore calibrate their reliance on it against what was actually checked.
The instrument does not fabricate the absent
Where the authored record is silent, the instrument is silent. It does not fill gaps. It does not reconstruct events for which no evidence has been authored. It does not assert, by inference from surrounding material, what must have happened at a point where no material exists.
This is Axiom I stated as a limit on the instrument rather than a discipline on the operator. The instrument honours silence because silence is what the instrument's data actually looks like at points where the record is thin. An operator who fills silence by inference is producing findings the instrument did not produce — the operator's inference has been laundered through the instrument's authority.
What happens past the limit: a finding that names a silence-filling inference as if the grammar had surfaced it attributes the operator's reasoning to the instrument. If the inference is later challenged, the operator has two defences, both bad. They can say the instrument produced the finding, which is false and exposes the operator's misrepresentation. They can say the inference was their own reasoning from surrounding material, which is accurate but raises the question of why the finding presented the inference as an instrument output rather than as an operator interpretation.
The discipline: where the record is silent, the finding records the silence. Where inference from surrounding material is appropriate, the inference is stated as the operator's, with the surrounding material cited. The inference may still be reported — the state user is not forbidden from reasoning — but the reasoning is attributed to the operator, not to the instrument. The reader can then evaluate the inference on its own merits, and the instrument's diagnostic authority is not borrowed for it.
The instrument does not author on the operator's behalf
The grammar is applied to a case by the case's author — typically the target user reconstructing their own material, in the companion corpus's register. The state user reading the registry encounters cases that have been authored. The state user does not author. Observations the state user produces live at the registry surface; they do not enter individual cases as amendments, additions, or reclassifications.
An operator who edits a case in the registry to align it with a pattern they have observed has corrupted the case's authorial provenance. The case now contains material the original author did not produce, authorised by the state user without the authority to do so. The registry can no longer distinguish what the author asserted from what the state user added. The integrity of the entire registry is at risk, because a single unauthorised edit sets the precedent that cases may be modified in the service of patterns.
What happens past the limit: the registry stops being a directory of independently authored cases and becomes a directory of cases that have been edited to support the analyst's view. At that point the instrument's core architectural property — the sovereignty of each case's author — has been destroyed. Every finding produced afterwards rests on material that may or may not reflect what any author actually authored.
The discipline: the state user's work lives in the findings they produce, not in modifications to the cases those findings reference. A pattern observed across cases points back to the cases as they are. If the cases are wrong, the state user reports their observation of the pattern and the cases' apparent conflict with it; the state user does not edit the cases to resolve the conflict. Case correction is the authoring author's responsibility, operating under the target-user corpus's discipline. The state user is a custodian of the registry's integrity, not its editor. Guide 8 — Corpus Governance — addresses custodianship in detail.
The instrument does not launder its own misuse
Each of the limits above describes a specific failure mode. A more general property underlies them: the instrument does not protect itself against misuse by operators in bad faith. Its authority can be borrowed. Its output can be misrepresented. Its silences can be backfilled. Its patterns can be applied beyond its validated scope. When any of these happens, the instrument's legible output carries the misuse forward without visible alteration.
A finding produced in bad faith looks the same as a finding produced in good faith. The chip field does not signal the operator's motive. The domain-chain view does not indicate whether descent was genuinely performed. The registry does not show whether patterns were honestly observed or strategically selected. The only protection against misuse is the operator's own discipline, supplemented by the review discipline of whoever reads the operator's findings afterwards.
This property is the instrument's most significant exposure at state level. An adversary with access to the authoring or analytical layers, acting strategically, can produce grammatically flawless output whose underlying evidential basis is selectively curated or fabricated. The output will read as cleanly as legitimate work. Only descent to the primary evidence — the discipline of Axiom II — and verification of authoring provenance — the discipline of Axiom III — can catch the manipulation.
The defence is therefore not technical. It is procedural and cultural: every finding that will influence governance decisions is reviewed by operators applying the full axiom set, primary-evidence descent is performed at the specific points that materially affect the finding, and authoring provenance is verified for every case that contributes to a pattern observation. No operator, however senior or trusted, is exempt from these reviews, because trust is the specific vector by which strategic manipulation most effectively enters the registry.
Limits as instrument properties
The seven limits above are not failures of the instrument. They are the specific shapes of what the instrument is not competent to do. An instrument without limits is not an instrument — it is a claim of unbounded capability, which no real instrument can support.
The state user's responsibility is to operate within the limits, to state them when producing findings, and to hand material that exceeds them to the processes that are competent to address it. Adjudication to courts, legal authorities, or regulators. Prediction to investigators who can gather new evidence. Intent-based claims to processes that can test them under appropriate evidential standards. Silences to authors who may subsequently address them, or to the record as permanent gaps. Case editing to the cases' own authors. Bad-faith use detection to the review layer and to the corpus governance discipline stated in guide 8.
An operator who knows the instrument's limits can use it correctly within them. An operator who does not know the limits will produce confident findings past them. The first operator makes the instrument useful. The second makes it dangerous. The difference is not skill. It is this guide.
Guide 7 — Register and the Downstream Reader — addresses what the operator owes to the audiences their findings reach. Knowing the limits is necessary but not sufficient. The limits must also be visible to the reader, in a form the reader can use. The translation is the subject of the next guide.