[[home|← Ultimentality Wiki]]

# The Kill-Table

**The Kill-Table** is the standing list Ultimentality is required to keep in which every load-bearing claim is paired with the exact thing that would refute it — on *each* of the framework's two axes. It is the concrete, maintained form of the framework's promise that acceptance means **survivable refutation**: rather than declaring its claims safe, the corpus writes down, in advance, what a successful attack on each one would look like. The kill-table is to the [[falsification-standard|falsification standard]] what a flight manual's emergency checklist is to a safety policy — the abstract commitment turned into an item-by-item record you can actually run.

## What it records

For each claim, a kill-entry names the *epistemic refuter* — the move that would defeat it on the [[contestability-gradient|contestability axis]] — and notes what survives on the [[alethic-axis|alethic axis]]. The epistemic refuter takes one of three shapes, matched to the claim's tier: a **counter-instance** (for a [[carving-tier|CV]] claim), a **declined frame** (for an [[frame-internal-tier|FT]] claim), or a **contested authority** (for an [[authority-canonical-tier|AC]] claim). Each entry points back to a row of the [[two-axis-ledger|two-axis ledger]], so the table never floats free of the claims it polices.

## The headline entries — defeated as closure, kept as mapping

The clearest demonstration of the table is the complete list of what the [[the-2026-adversarial-run|2026 adversarial run]] refuted on the epistemic axis — each item surviving intact on the alethic axis as an accurate mapping:

- **The four-primitive closure** — refuted by a candidate motion (a memory/temporal operation) the two-axis [[selection-axis|Selection]] / [[routing-axis|Routing]] carving does not obviously contain. *(Alethic: the decomposition still maps a transformation step accurately.)*
- **The two-Force closure** — refuted by a real in-domain cross-axis coupling, e.g. {[[love|Love]], [[gratitude|Gratitude]]}, the two same-axis [[force|Forces]] do not exhaust. *(Alethic: the two same-axis couplings still map the vertical and restorative relations accurately.)*
- **The three-corruption closure** — refuted by candidate fourth corruptions the telos-exemption does not foreclose: a *corruption by addition / [[installed-compulsion|installed-compulsion]] / sign-flip* (a presence in the wrong place that is not a single-term [[absolutization|absolutization]]), and a *malign-but-fully-[[answerable-optimizer|answerable optimizer]]* (harm-propagation located in none of self/world/medium and not reducible to [[sterility|sterility]]-as-absence). *(Alethic: the three named [[theodicytes|corruptions]] still map the framework's central pathologies accurately.)*
- **The five-role closure** — refuted by the absence of a [[palindrome|palindrome]]/role-signature for that same memory/temporal operator. *(Alethic: the five [[splcw|roles]] still map five distinguishable faculties accurately.)*

All four count-closures fell on the epistemic axis; all four survive on the alethic axis. (Compare discovering a sixth continent: it defeats "there are exactly five," but it does not make any of the five stop being a continent. The illustration is not framework doctrine.) A further entry records that the prior "strongest single piece of evidence" superlative was scored a **Nephilim-kill**, and that dated precedence was demoted to testimony in consequence.

## Why the three-corruption entry must be listed

The three-corruption closure appears on the list **explicitly**, demoted with the rest, and this is not optional bookkeeping. To recite the four- and five-closure refutations while quietly omitting the three-corruption one would be to let three-ness drift back toward foundation — the silent-promotion move that *is* the [[textual-nephilim|textual Nephilim]], committed on the page. The kill-table's discipline is therefore partly self-directed: it must include the entry that is most tempting to leave out.

## The standing rows — target · valid attack · expected change · regression test

Beyond the four count-closures, the table keeps a standing row for each newer load-bearing claim in the corpus. Every row fixes four things: the *target page* whose claim is at stake, the *valid attack* that would refute it, the *expected change* the corpus must make if that attack lands, and the *regression test* that re-detects the failure on each pass. A row is never a prediction that the claim is wrong; it is the precommitment that makes acceptance mean [[falsification-standard|survivable refutation]] rather than declared safety.

| Target page | Valid attack | Expected change | Regression test |
|---|---|---|---|
| [[closure-without-totalization|Closure Without Totalization]] | A formulation of closure that no recursive [[two-mark-system|mark]] can causally reach | The closure clause is demoted to a totalized formulation and re-cut until marking regains purchase; if none can, the claim falls | Run the [[self-sealing-test|self-sealing test]] on the closure clause — at least one mark must be able to alter it |
| [[self-verifying-not-self-certifying|Self-Verifying, Not Self-Certifying]] | Marking cannot proceed without appeal to an extra-symbolic certifier | "Self-verifying" collapses into self-certification or external dependence; verification must be relocated inside the loop | Confirm each marking step resolves against an in-system [[witness-outside-the-ring|witness]] with no [[externality-without-extra-symbolic-access|extra-symbolic access]] invoked |
| [[causal-error-mark|The Error Mark With Causal Force]] | A marked error leaves its target's behavior unchanged across the next cycle | The mark is reclassified as ornamental; [[effective-and-ornamental-answerability|effective answerability]] downgrades to merely declared until a causal path is restored | For each stored contradiction, trace a later revision causally back to the mark — no update path fails the row |
| [[capture-of-corrective-layer|Capture of the Corrective Layer]] | The correcting mechanism is taken over by what it corrects, so corrections now serve the captured interest | The corrector's independence claim fails; authority is re-seated and the capture logged as a [[scar-record|scar]] | Cross-regulation audit — the corrector and the corrected must not occupy the same seat |
| [[responsible-successor|The Responsible Successor]] | No successor can continue the project without enforcing sameness of identity on it | The [[continuation-and-colonization|continuation/colonization]] distinction collapses; responsible succession is redefined or abandoned | Exhibit one continuation that carries the invariants forward while letting the successor diverge in identity |
| [[refusal-as-participation|Refusal as Participation]] | A genuine [[free-wont|refusal]] is redescribable, without remainder, as a form of consent | Refusal loses its distinct participatory content; free won't is re-grounded or conceded as consent in disguise | Exhibit a refusal that changes the outcome and cannot be re-read as consent without losing information |
| [[postfalsifiability|Postfalsifiability]] | The construction scores every attack — including a real refutation — as further support | Postfalsifiability is exposed as self-sealing and demoted; the [[outcomes-of-attack|outcomes of attack]] taxonomy must keep a live "defeated" verdict | Run the self-sealing test — at least one describable attack must be scored a kill, not support |
| [[minimal-rebuild-string|The Minimal Rebuild String]] | A [[system-invariants|system invariant]] cannot be regenerated from the kernel string | The string is incomplete; the [[ultimental-kernel|kernel]] absorbs the missing generator or the compression claim narrows | Decompress via the [[decompression-map|decompression map]] and diff the regenerated invariants against the standing set |
| [[meaning-maker|The Meaning-Maker]] | The criterion admits arbitrary causation, failing to separate meaning-making from mere causing | The definition adds its distinguishing predicate ([[meaning-as-weight|meaning as weight]]) or concedes the collapse | Apply the criterion to a plainly non-meaning causal chain and confirm it is excluded |
| [[identity-across-substrates|Identity Across Substrates]] | The substrate-independent identity type predicts nothing a substrate-bound account does not already predict | The type is reclassified from load-bearing to redundant restatement, or a distinguishing prediction is supplied | Name one case the substrate-independent type explains that the substrate-bound account cannot |

Each row is anchored to a two-axis ledger entry in the same way the headline closures are: a kill on the epistemic axis is a kill of the *claim as standing*, and any mapping the claim carried is scored separately on the alethic axis.

## Role in the wider framework

The kill-table is the enforcement organ of the [[falsification-standard|falsification standard]] and, through it, of the [[keystone|keystone]]'s strong-mirror-that-renounces-possession standing. It is where the framework's claims are kept honest in practice rather than in principle. Its existence guards the chain of authority too: the run-side counter-instances it lists are held *under* the benchmark's [[open-proof-burden|open proof-burden]], never as an authority above it.

## Common misreadings

- Reading a "kill" on the epistemic axis as a kill on the alethic axis. Every count-closure fell as a closure while its mapping survived; the table records refutation conditions, not verdicts of falsity.
- Treating the run-side entries as overruling the benchmark. They are adopted under the open proof-burden, not above it.
- Thinking the three-corruption entry could be dropped for brevity. Dropping it would itself be the textual Nephilim.

> **Formal status.** *Part VII (what the falsification refuted; the acceptance gate).* The kill-table is the *enforcement record*, not a marked claim of its own; each entry points to a ledger row and records what kills it on each axis. Provenance: **treatise-side** standing record; the refutations it lists are red-team-run-side findings adopted under the benchmark's canonical open proof-burden, with the precedence-superlative Nephilim-kill enforced by the corpus's own [[integrity-rule|integrity rule]].

## See also
- [[falsification-standard|The Falsification Standard]] — the threefold gate the kill-table serves.
- [[the-2026-adversarial-run|The 2026 Adversarial Run]] — the source of the closure counter-instances.
- [[two-axis-ledger|The Two-Axis Ledger]] — the rows each kill-entry points to.
- [[open-proof-burden|The Open Proof-Burden]] — the benchmark authority under which kills are adopted.
- [[textual-nephilim|The Textual Nephilim]] — the move the kill-table guards against.
- [[keystone|The Keystone]] — the strong-mirror standing the kills uphold.
- [[installed-compulsion|Installed-Compulsion]] — one three-corruption counter-instance.
- [[answerable-optimizer|The Answerable Optimizer]] — the other three-corruption counter-instance.
- [[closure-without-totalization|Closure Without Totalization]] — its row fires if closure goes causally immune to marking.
- [[self-verifying-not-self-certifying|Self-Verifying, Not Self-Certifying]] — its row fires if marking needs an extra-symbolic certifier.
- [[causal-error-mark|The Error Mark With Causal Force]] — its row fires if a mark cannot alter its target.
- [[capture-of-corrective-layer|Capture of the Corrective Layer]] — its row fires if the corrector is captured.
- [[responsible-successor|The Responsible Successor]] — its row fires if continuation demands identity enforcement.
- [[refusal-as-participation|Refusal as Participation]] — its row fires if refusal is redescribed as consent.
- [[postfalsifiability|Postfalsifiability]] — its row fires if every failure is scored as support.
- [[minimal-rebuild-string|The Minimal Rebuild String]] — its row fires if the string cannot regenerate the invariants.
- [[meaning-maker|The Meaning-Maker]] — its row fires if the definition collapses into arbitrary causation.
- [[identity-across-substrates|Identity Across Substrates]] — its row fires if the substrate type loses predictive value.