[[home|← Ultimentality Wiki]] # The Outcomes of Attack ## Definition When a valid attack lands on a content-bearing claim, the framework recognizes exactly four things that can happen to the claim — and one thing that is **not** among them. The four standardized outcomes are: 1. **Deletion.** The claim dies. It is removed from the corpus and its dependents are rerouted or fall with it. 2. **Revision.** The claim changes. Its wording, scope, dependency, or formalization is rewritten, and the earlier form is retained as the reason the new one exists. 3. **Demotion or retyping.** The claim survives at a different level or status. Its mark moves (for instance FT → CV), its abstraction level shifts, or its tier changes, without the underlying relation being destroyed. 4. **Invariant recognition.** The attack — run as a genuine removal test — shows that deleting the relation destroys the function it regulates. The claim survives *as load-bearing*, and what is recorded is the function that would be lost, not a verdict of proof. The excluded fifth is **confirmation**. A valid attack that does not land is logged as a *failed attack* — a record type, never an outcome, and never evidence that the claim it failed against is therefore true. Counting a survived attack as support is the drift toward a **forced** tier — the [[textual-nephilim|Textual Nephilim]] — and is prohibited across the frame. Every resolution — whichever of the four it reaches — is recorded as a fixed object, not a narrative: ```text resolution: target: attack_type: evidence: outcome: deletion | revision | demotion/retyping | invariant-recognition changed_content: changed_marks: changed_links: remaining_kill_condition: ``` A resolution missing these fields is not a resolution; it is an anecdote. The record is the operative object, and it is precisely what [[crystallization|Crystallization]] and the [[scar-record|scar record]] later read. | Outcome | The claim | What moves | What the record fixes | |---|---|---|---| | Deletion | dies | it leaves the corpus; dependents rerouted or dropped | evidence, the removed content, downstream link repairs | | Revision | changes | wording, scope, dependency, or formalization | old form, new form, why the old failed, the surviving kill condition | | Demotion / retyping | survives at a new level or status | its E/A mark, tier, or abstraction level | old mark, new mark, the counter-instance that forced the move | | Invariant recognition | survives as load-bearing | nothing in the claim; the *understanding* of it | the attempted removal, the function that would be lost, the still-live kill condition | ## Type and formal status **E (epistemic):** Derived, **CV**. This is a classification — a carving of the outcome-space — and by the invariant that *every count is a contestable carving*, the count of four is contestable by a counter-instance (a genuine fifth outcome) or by a better carving that merges or re-cuts the set. It is not FT: the partition does not follow from the definition of *attack* alone; it is a proposed exhaustive-and-disjoint cover of what a landed attack does, and a cover can be wrong. **A (alethic):** the classification aspires to *map* the actual behavior of a claim under a valid attack. Its accuracy claim is that these four exhaust the ways a claim is moved and that "confirmation" names no motion at all. Exhibiting a real movement outside the four would make the carving **inaccurate**, not merely unstylish — the axes do not predict each other, so *contestable* here never means *probably wrong*. This page is a **treatise-side extension, held contestable**. It fixes no authority and licenses no exempt or forced status for anything, including itself. ## What it regulates This page supplies the outcome vocabulary that [[postfalsifiability|Postfalsifiability]] presupposes: a postfalsifiable corpus is one whose invariants are the residue of attacks that were *killed, revised, retyped, or recognized as invariants* — and that residue is only legible if each of those verbs is defined and recorded. Without a fixed outcome set, "the claim survived falsification" degrades into an unaudited boast. It regulates one characteristic excess directly: **confirmation creep** — the drift by which a survived attack is silently reclassified as support and, from there, the claim is nudged toward exemption. The [[falsification-standard|Falsification Standard]] forbids this in principle; this page operationalizes the forbidding by giving survival no cell labeled *confirmed*. It regulates the **failed-attack / invariant-recognition boundary**, the very seam its own kill condition targets. Invariant recognition is a positive result — a removal test that produced a *specified functional loss*. A failed attack is an absence — an attack that never reached the seam. Collapsing the second into the first is how "you couldn't defeat it" becomes "it is necessary," which is the [[textual-nephilim|Textual Nephilim]]. Finally it regulates **attack bookkeeping**: the eight recorded fields force every resolution to name its target, type, evidence, and remaining kill condition, so that no attack can dissolve into a story about who prevailed. ## What regulates it **[[causal-falsifiability|Causal Falsifiability]]** is the primary regulator. An outcome counts only when its motion is a traceable state-change: a *deletion* that removes nothing, a *revision* that alters no wording, mark, dependency, or behavior, is not an outcome but a display. Causal falsifiability supplies the requirement that each of the four names a real change; this classification supplies the names. Neither is sufficient alone. The **[[same-level-attack-rule|Same-Level Attack Rule]]** and **[[attack-type-matching|Attack-Type Matching]]** sit *upstream*. They decide whether an attack is valid at all — whether it contests the claim at the type and abstraction level where the claim is made. An ill-typed attack has no outcome on this page; it is rejected before classification, not slotted into a default cell. The **[[two-mark-system|Two-Mark System]]** regulates the record's `changed_marks` field: demotion and retyping are motions *within* the E/A ledger, and this page cannot invent a mark the ledger does not carry. And the **CV mark on its own four-count** regulates the page from inside. The classification is contestable by the same moves it classifies; a better carving would revise it, and a demonstrated fifth outcome would enlarge it. ## Valid attack surface A valid attack on this page must do one of the following, at the level of the classification itself: - Exhibit a **genuine outcome** of a *landed valid attack* that is irreducible to deletion, revision, demotion/retyping, or invariant recognition — a fifth motion the four cannot represent. (A candidate must be a motion of the *claim*, not a fact about the attacker or the audience.) - Show that **two of the four collapse** — for instance, that demotion is always a species of revision, or that invariant recognition is only deletion-of-the-attack under another name — so the carving is redundant. - Show that the classification **cannot in practice separate a failed attack from invariant recognition**, i.e. that any survival can be filed under either cell at will. What is *not* a valid attack: objecting that the count "four" is arbitrary (all counts are CV; arbitrariness of a *number* is not a counter-instance), or offering a failed attack as though its failure were an argument against the page. A failed attack against this page is logged, per the page's own rule, as a failed attack. ## What happens if isolated Isolated from **causal falsifiability**, the four labels become bookkeeping theater: resolutions are written, outcomes are stamped, and nothing in the governed corpus moves. "Deletion" and "revision" without a traced state-change are ornamental — the failure the [[causal-error-mark|causal error mark]] names, transposed to the ledger. Isolated from the **failed-attack guard**, the classification self-seals. With no cell for *failed* and a permissive reading of *invariant recognition*, every survived attack finds a home under "the removal test showed the relation is necessary," and from there the claim is quietly promoted toward the exempt tier. Classification without the guard is a confirmation engine wearing a falsification costume — exactly what the [[self-sealing-test|Self-Sealing Test]] exists to catch. Isolated from the **upstream validity rules**, outcomes get assigned to attacks that never earned an outcome: a process offered against an operator, a substrate difference offered against a function, is mis-typed but still "resolved," and the ledger fills with resolutions of non-attacks. ## What larger property emerges from the coupling Coupled with causal falsifiability and read into the postfalsifiability ledger, this classification yields **an auditable falsification record**: a corpus whose surviving invariants are the compressed residue of *logged, causally effective defeats*, each survival recorded honestly and none mislabeled as proof. The coupling is specific. The classification alone gives outcome *names* but cannot guarantee they correspond to anything. Causal falsifiability alone demands a real state-change but does not say which states count as which motion. Wired together — each outcome name bound to a required traced change — they produce a ledger that [[crystallization|Crystallization]] can consume: high structural compression achieved *because* unstable arrangements were eliminated and the eliminations are on the record, distinguishing a crystal (still answerable) from a fossil (a dead form merely preserved). [[semantic-natural-selection|Semantic natural selection]] operates over exactly this record; without honest outcome-accounting it would mistake mere persistence for survival-by-selection. ## What would actually kill the claim The claim dies if **the system cannot distinguish a failed attack from invariant recognition** — if there is no operable criterion by which "the attack didn't connect" is separated from "a removal test demonstrated a specific functional loss," so that any survival can be booked as either. Then the classification collapses into the confirmation engine it was built to prevent. It also dies if a critic **exhibits a real fifth outcome** of a landed valid attack, irreducible to the four, or shows that two of the four are never distinct in practice. Either result revises or re-carves the page and is logged, on this very page, as a *revision* or *retyping* — with a new kill condition attached. And it is undercut if the eight record fields can be **filled without any causally traceable change** — if a complete resolution can be written for a motion that never occurred. That routes the failure to [[causal-falsifiability|Causal Falsifiability]] and to the [[kill-table|Kill-Table]], where the corresponding row already stands: *postfalsifiability treats all failures as support.* ## Prohibited misreadings - **Invariant recognition read as confirmation, proof, or a forced tier.** Recognizing that removing a relation destroys a regulated function does not lift the claim out of contestability. The claim stays Derived, keeps its E/A marks, and keeps a live kill condition. Reading load-bearingness as forcedness is the [[textual-nephilim|Textual Nephilim]] and the local face of [[absolutization|absolutization]]. - **A failed attack read as strengthening the claim.** Failure to defeat is not evidence for. A failed attack is logged as a failed attack and contributes nothing to the truth of what it failed against. This is the framework-wide rule that a failed attack is never counted as confirmation. - **The four outcomes read as a fixed, exhaustive, exempt set.** The count is CV, like every count. It invites a better carving; it does not close. - **Demotion or retyping read as deletion — or as promotion.** Survival at another level or status is neither death nor victory. It is a recorded move within the [[two-mark-system|two-mark ledger]], with the counter-instance that forced it on the record. - **Classification read as guaranteeing an outcome.** Validity is decided upstream by the [[same-level-attack-rule|Same-Level Attack Rule]] and [[attack-type-matching|Attack-Type Matching]]. An invalid attack yields *no* outcome, not a default one. - **The resolution record read as optional narrative.** The eight fields are the operative object of the resolution, not a summary of it. Without them there is no logged outcome, only a claim about who won — which the framework does not accept. ## See also [[postfalsifiability|Postfalsifiability]] · [[causal-falsifiability|Causal Falsifiability]] · [[same-level-attack-rule|The Same-Level Attack Rule]] · [[attack-type-matching|Attack-Type Matching]] · [[self-sealing-test|The Self-Sealing Test]] · [[crystallization|Crystallization]] · [[falsification-standard|The Falsification Standard]] · [[kill-table|The Kill-Table]] · [[scar-record|The Scar Record]]