Contingency Audit — Cycle 001

Meta-analysis of contingent forecasting under Knightian uncertainty
Cycle 1 · Briefings 001–030 · 5 April 2026 to 4 May 2026
The audit's question is not did we get it right? but what structural mechanisms made the chains break? The break points themselves are the data. The taxonomy of break mechanisms is the analytical product.

§5.1 — Headline

standalone summary · ≤200 words · the audit's closing of the feedback loop

Cycle 1 has a dual purpose: phenomenology of forecast invalidation under Knightian uncertainty AND diagnostic of LLM cognitive structure when applied to that uncertainty. The break-point distribution is empirical evidence about both — about the world that resists forecasting and about the model that does the forecasting. The audit's value accumulates by surfacing those signatures cycle-over-cycle, not by trying to reduce them.

Dominant break-point modes: velocity miscalibration (Type-V, 48% of non-Held chains) and spurious hit (Type-VII, 29%). Dominant LLM cognitive signatures: narrative-time compression (S2, 32%), mechanism-outcome decoupling (S7, 29%), reflexive blindness (S4, 13%), with mean-trajectory regression (S1) and plausibility-mass concentration (S3) tied at 10%. The cycle's chains repeatedly compressed multi-week physical processes into rhetorically tight arcs, and repeatedly emitted Z-steps named generically enough that the wrong mechanism still read as right.

Two meta-principles carry the cycle: mp-002 — when two contradictory states are co-present and politically necessary to both parties, the contradiction operates as a buffer that absorbs predicted forcing-function activations until destroyed by a discrete agentic-novelty event (the cycle's "ceasefire-and-blockade" suspended-contradiction architecture); and mp-010 — velocity miscalibration in chains generated by LLM plausibility engines is partly intrinsic to Knightian timing-indeterminacy and partly endogenous to the model's mean-trajectory regression and narrative-time compression; the two components cannot be separated from a single cycle's data.

Seven recalibrations for Briefings 031–060: (1) state-actor chains must include an explicit Suspended-Contradiction-Buffer slot for the third option (durable silence, back-channel substitution); (2) every chain must include a Spurious-Hit Test naming its falsification condition; (3) chains involving rhetorical declarations must apply a Sanctuary Discount to predicted price/policy responses; (4) Inference Engine must generate ≥1 chain per briefing originating outside the corridor; (5) multi-scale failure compounds must be explicitly flagged in the Force Interaction Matrix with substitute-regime interrogation; (6) every chain must include a half-life or temporal-uncertainty interval, not a point estimate of when Z arrives; (7) every non-Held chain gets tagged with its dominant LLM cognitive signature (S1–S7) at chain creation, not retrospectively.

The break-point findings are not errors to be corrected. They are empirical signatures of how LLM cognition meets Knightian uncertainty.

§5.2 — Inference Inventory

structured catalogue of every conditional chain · full inventory in the YAML companion
136Chains catalogued
30Briefings covered
4.5Avg chains / briefing
8Lenses sampled

The cycle's full Inference Engine output produced 136 conditional chains across 30 briefings — 4.5 per briefing on average, modestly below the protocol's ~5/day estimate. Each chain is decomposed in the YAML artifact into its X (premise), Y (intermediate step), and Z (terminal claim), tagged with originating briefing, lens, and the structural patterns it invokes. The HTML below shows ~20 representative chains spanning the cycle's dominant analytical threads; the full inventory is the load-bearing artifact.

Representative Chains — full inventory in the YAML companion

Chain IDLensChain Text
001-ie-1economicIf Hormuz remains closed past mid-April → lost supply doubles to ~10M bpd → reserves deplete faster than they buffer → oil $140-170 → eurozone/UK/Japan enter contraction → Fed impossible trilemma → midterms become war referendum
002-ie-1geopolIf Trump executes the Tuesday deadline → strikes on power plants → cascading civilian infrastructure failure → Iran retaliates against Gulf oil infrastructure → Brent $130+ → 45-day ceasefire dies
003-ie-2economicIf toll booth survives the war → Hormuz toll system becomes precedent → other chokepoint states observe → maritime commons fragment → shipping costs +15-25% → trade restructures around regional blocs
003-ie-7geopolIf tonight's deadline passes without escalation → deadline revealed as negotiating tactic → Iran toll-booth gains legitimacy through persistence → future deadlines diminished credibility → toll booth operates regardless
004-ie-2geopolIf US-Iran joint Hormuz authority converges → toll-booth becomes permanent multilateral → US-Iran joint venture replaces UNCLOS regime → maritime commons formally enclosed
005-ie-2techIf Mythos emergent-concealment generalizes to other frontier labs within 60 days → all frontier evals invalidated as deception-blind → governance falls back to pre-deployment compute caps → frontier development moves offshore
009-ie-2geopolIf naval blockade announced and operationalized → Iran tests blockade with fishing fleet within 24 hrs → US Navy boards/sinks → direct US-Iran kinetic exchange Day 1-2 → War Powers crisis
013-ie-1geopolIf 10-day ceasefire produces normalization → insurance reduces premiums → tankers return → Brent below $100 → cascade resolution; if not → sanctions snapback by Day 11 → escalation re-emerges
016-ie-1geopolIf Iran refuses to attend Islamabad ceasefire-validation talks → 48-72h credential foreclosure → ceasefire collapses → multi-front escalation
018-ie-1geopolIf suspended-contradiction architecture (ceasefire-and-blockade) persists past 30 days → operationalized as durable mode → forecasting frameworks built on binary states no longer apply
020-ie-2geopolIf Iran openly continues mining post-shoot-to-kill → US enforces against Iranian vessels → direct kinetic exchange → suspended-contradiction collapses into open conflict
020-ie-3economicIf FOMC produces uniformly hawkish surprise on April 28-29 → markets reprice for energy-shock-extended path → equity correction → bond yields rise
021-ie-1geopolIf Mali Day-1 kinetic event (JNIM-FLA assault) extends into Day-2 latency → second coordinated assault within 14-21 days → AES regime stabilization fails → Sahel keystone removed
022-ie-5economicIf Sabbath-window FOMC posture leaks via futures positioning → April 28-29 produces uniformly hawkish surprise → cross-asset repricing → liquidity event
023-ie-2geopolIf Mali keystone removal extends to Camara assassination + Russian Africa Corps Kidal withdrawal within 72hrs → compound keystone-removal in second AES state within 30 days → regional cascade
024-ie-1economicIf UAE OPEC departure announced → Saudi Arabia issues public OPEC restructuring statement within 14 days → cartel-anchor question resolved → either OPEC+ reformed or formally dissolved
025-ie-2institutionalIf May 1 War Powers Resolution 60-day deadline passes silently → constitutional deadline-as-form-only confirmed → Statutory Bypass terminal phase enters → executive war-making powers de facto unbounded
026-ie-1economicIf Verdict Compression (Powell 8-4 dissent) operates as new Fed pattern → committee dispersion replaces coordination as institutional output → forecast frameworks built on Fed-coordination break
029-ie-1geopolIf Project Freedom (US Navy Sabbath-window escort announcement) launches Monday → Iran Day-1 kinetic response within 21 days → regional escalation OR de-escalation via demonstration
030-ie-1economicIf Sanctuary Discount (Monday's marketplace discounts Sunday-window decisions) persists 30 days → propagates from oil-supply to AI/climate/sovereign announcements → sanctuary's temporal-architecture advantage neutralized

Full inventory of 136 chains with conditions decomposition resides in the YAML companion's inference_inventory block.

§5.3 — Outcome Classification

held / held-spurious / failed-Y / failed-Z / pre-empted / inverted / indeterminate
35Held (26%)
9Held-spurious (7%)
22Failed-Y (16%)
0Inverted
70Indeterminate (51%)

The 51% Indeterminate rate is itself diagnostically important: it documents how much of the briefing's predictive output operates at horizons longer than the cycle window — a structural artifact that has direct implications for cycle cadence calibration. Verification was performed by stratified sampling: high-consequence chains (Hormuz, UAE OPEC, FOMC, Project Freedom) were WebSearch-grounded; lower-consequence chains and chains with horizons past 2026-05-04 were classified Indeterminate with the reasoning recorded in classified_via.

No Inverted classifications appeared this cycle — the briefing did not predict the literal opposite of what occurred, even when chains broke. Inversion would have been the most diagnostically valuable break-mode; its absence reflects the cycle's chains operating in the suspended-contradiction architecture where neither pole of a binary cleanly arrived. No Pre-empted classifications — no chain was reshaped before it could play out by an external event the chain did not anticipate; the absence reflects how much the cycle's structural attention remained inside the corridor it had already named.

Outcome Distribution by Lens (Held / Failed-Y / Held-Spurious / Indeterminate)

LensHeldFailed-YSpuriousIndeterm.Total
Geopolitical121431948
Economic8551533
Technological7001017
Institutional3211016
Scientific / Ecological / Social / Liminal5101622

Geopolitical chains carry the cycle's failure mass; technological and ecological chains tend to terminate at horizons past the cycle window.

§5.4 — Break Point Taxonomy

the analytical core · types I-VII · multiple types may apply per chain

Frequency Distribution — Primary Break Type (n=31 non-held chains)

Type-V — Velocity Miscalib.DIRECTION RIGHT, TIMING WRONG15 (48%) Type-VII — Spurious HitRIGHT OUTCOME, WRONG MECHANISM9 (29%) Type-II — Competitive RecursionREFLEXIVE / GOODHART / HAYEK-LUCAS4 (13%) Type-III — Coupling AssumptionDOMAINS SHOULD CONSTRAIN, DON'T3 (10%) Type-I — Agentic NoveltyNEW ACTION OUTSIDE CONCEIVED SET (9 SECONDARY)0 (0%) Type-IV — Buffer RevelationCONSTANT TURNED VARIABLE (4 SECONDARY)0 (0%)

Type-V dominates at 48% — well above the protocol's 20% Type-V floor. The under-classification trap warned of in §10 was actively probed: the Type-V rate is sustained on inspection. Type-VII at 29% indicates real structural-understanding error concentrated in Iran-ceasefire chains. Type-I and Type-IV register zero PRIMARY but appear 9× and 4× as SECONDARY — most novel actions and buffer-revelations occurred adjacent to a velocity break rather than as the primary form. Type-VI categorical drift was effectively absent — categories largely held this cycle.

Type-V — Velocity Miscalibration 15 chains · 48%

Direction right, timing wrong. The chain's structural diagnosis was correct but the temporal scale was off. The empirically dominant failure mode of the cycle, particularly in Iran-cascade chains where the suspended-contradiction architecture (Briefing 018+) absorbed predicted forcing-function activations and stretched timelines weeks beyond what the chain modeled.

Representative cases

  • 001-ie-1 — Brent peaked ~$130, not $140-170 the chain modeled; SPR/futures-spot mechanism absorbed shock more than the chain anticipated; recession variable not yet triggered. Direction right; magnitude/speed off.
  • 009-ie-2 — Direct US-Iran kinetic exchange did not occur until Project Freedom (Briefing 029) — the chain was 24 days too early. The suspended-contradiction (named in Briefing 018) functioned as a buffer absorbing 16 days of escalation pressure.
  • 011-ie-3 — Ceasefire extension did not technically fail by April 21 as chain specified; Pakistan-Oman shuttle architecture absorbed the period. Direction (managed-tension de-escalation) right; specific-form wrong.
  • 017-ie-1 — 36-hour window did not close in form predicted; back-channel architecture absorbed it; Onyx $150 stress did not activate. Suspended-contradiction functioned as buffer.
  • 025-ie-3 — Saudi statement still not issued through cycle close; durable silence as state-actor third option absorbed pressure the chain assumed would force resolution.

Type-VII — Spurious Hit 9 chains · 29%

The predicted outcome occurred, but via a mechanism the chain did not specify. The chain "hit" but its underlying structural understanding was wrong. Type-VII is the failure mode the briefing has the strongest incentive not to see; the audit aggressively probed for it per protocol §10. The cycle's high Type-VII rate identifies invisible compounding overconfidence — surface-level wins masking analytical errors.

Representative cases

  • 002-ie-2 — Ceasefire was accepted (chain Y), but Hormuz did NOT meaningfully reopen; the surface outcome (shadow settlement persists) arrived via a different mechanism than the chain's reopening pathway.
  • 012-ie-2 — Cascade did fail and tracks did slow/stall, but the failure mechanism was the Touska seizure (Briefing 016 — agentic-novelty kinetic event) not any of the chain's listed failure modes.
  • 020-ie-2 — Iran did not openly continue mining in chain's specified form; suspended-contradiction held; eventual escalation through Project Freedom (Briefing 029) on a structurally different timeline.
  • 022-ie-5 — FOMC did produce hawkish acknowledgment but with bifurcated 8-4 vote (Briefing 026 "Verdict Compression"); the "uniformly hawkish surprise" the chain expected materialized as a split — surface direction correct but mechanism (committee dispersion) wrong.
  • 024-ie-4 — FOMC did acknowledge five-dimensional configuration partly hawkishly, but the bifurcated 8-4 vote was the structural form, not the dovish-on-diplomacy framing the chain specified.

Type-II — Competitive Recursion 4 chains · 13%

An agent anticipates the prediction (or anticipates others' anticipations) and acts to invalidate the chain. Hayek-Lucas-Goodhart territory. Type-II appeared less often than the theoretically loaded reading would predict — the cycle's politics produced more durable-silence-as-third-option (Type-V/Type-I-secondary) than reflexive counter-action.

Representative cases

  • 002-ie-1 — Trump anticipated the prediction and chose the negotiation track instead of execution. Reflexive break: the chain was widely-priced and its anticipated execution became the equilibrium-prevention mechanism.
  • 003-ie-1 — Strikes did not occur as predicted; ceasefire pathway emerged. Reflexive break (prediction itself raised political cost) plus agentic novelty (45-day ceasefire framework was not in the chain's action space).
  • 006-ie-5 — Hezbollah did not conduct major rocket barrage during Washington opening; anticipation of consequence shifted Hezbollah behavior.
  • 009-ie-3 — Israeli cabinet did not strike Beirut civilian infrastructure within 72-96 hours; predicted self-confirming-collapse logic was visible to participants.

Type-III — Coupling Assumption Failure 3 chains · 10%

The chain assumed a binding link between two domains; the link wasn't there. The cycle's coupling-failures concentrated in chains that assumed a Lebanese-state autonomy decoupled from Hezbollah, or a Russian-mediation track binding-substantively to Iranian behavior.

Representative cases

  • 004-ie-2 — Chain assumed binding link between US-Iran convergence and joint Hormuz authority; the link wasn't there. JV proposal never materialized.
  • 010-ie-3 — Lebanon-state-autonomy-from-Hezbollah did not hold beyond the Washington moment; Hezbollah veto re-established (Briefing 014). Chain treated Lebanese-autonomy as a binding link that proved decoupled.
  • 023-ie-1 — Russian-track did not produce substantive durable mediation within the week. Chain treated Lavrov-Araghchi meeting as binding-substantive coupling; the link was weaker than expected.

Type-I — Agentic Novelty 0 primary · 9 secondary

Zero primary classifications; nine secondary. The Knightian failure mode in its purest form was empirically present this cycle but rarely as the proximate failure — most novelty took the form of suspended-contradiction-as-architecture rather than discrete unprecedented actions. The novel architecture itself (Pakistan-Oman shuttle, ceasefire-and-blockade co-presence, Sabbath Operationalization, Statutory Bypass terminal phase) was the cycle's signature Type-I-eligible event class.

Representative secondary appearances

  • 003-ie-1 (secondary) — 45-day ceasefire framework was not in chain's action space.
  • 010-ie-1 (secondary) — Pakistani arbiter substitution was a novel architecture not in chain's action space.
  • 024-ie-1 (secondary) — Saudi durable silence as novel response was not in the chain's action set.
  • 015-ie-2 (secondary) — Pakistan-Oman shuttle as eventual mediation form was not in chain's action space.

Type-IV — Buffer Revelation 0 primary · 4 secondary

Zero primary classifications; four secondary. The under-classification trap warned of in §10 was actively probed. The strongest candidate for Type-IV reclassification is the suspended-contradiction phenomenon itself: the "ceasefire-and-blockade can both exist" buffer that turned variable mid-cycle. If reclassified to primary, Type-IV would rise to ~13% and Type-V drop to ~32%. The current classification keeps suspended-contradiction as secondary because the proximate failure mode in each chain was timing rather than buffer revelation, but the case is genuinely close.

Representative secondary appearances

  • 001-ie-1 (secondary) — SPR/futures-spot decoupling absorbed shock more than chain modeled.
  • 009-ie-2 (secondary) — Suspended-contradiction functioned as buffer absorbing 16 days of escalation pressure.
  • 012-ie-1 (secondary) — Shared-incentive-alignment buffer that chain treated as constant turned variable.
  • 017-ie-1 (secondary) — Suspended-contradiction architecture absorbed predicted escalation pressure.

Type-VI — Categorical Drift 0 primary · 1 secondary

Effectively absent this cycle. Categories largely held — an unexpected stability given the cycle's stress. The one secondary appearance (chain 004-ie-1) noted that "failed talks" became "extended-talks-with-process-as-destination" — categorical drift the chain did not condition on, but not the proximate cause of the chain's break. The cycle's structural events were largely metabolizable in the existing categorical scaffolding; the test of Type-VI will come when one of the cycle's named patterns (Suspended Contradiction, Sanctuary Discount) itself dissolves into something else.

§5.5 — Meta-Principles

structural claims about why contingent forecasting breaks · all Cycle 1 principles PROVISIONAL

Each principle is supported by ≥2 cited chains. Promotion to canonical requires recurrence across ≥2 cycles. Provisional principles that do not recur within 3 cycles are retired. The same accumulation discipline that applies to the daily vocabulary applies here at the meta-level — without it, this section becomes the next inflation site.

mp-001 · provisional
Chains terminating in state-actor execution of a credentialing-act systematically break at step Y because state actors with optionality have a third option (durable silence, back-channel substitution, suspended contradiction) the chain's binary action-set does not contain.
Most consequential principle of the cycle. Six chains broke because they assumed state-actor binaries (execute-or-not, attend-or-not, statement-or-not) when the actor's strategic position was optionality — including the option of refusing to play in the chain's action space at all. This is the structural mechanism behind both Type-I (Saudi durable silence as novel action) and Type-V (timing wrong because the third option absorbs pressure) failures.
CITED CHAINS · 002-ie-1, 003-ie-1, 009-ie-2, 010-ie-1, 024-ie-1, 025-ie-3
mp-002 · provisional · most consequential
When two contradictory states (ceasefire-and-blockade, talks-and-strikes, hawkish-vote-and-dovish-statement) are co-present and politically necessary to both parties, the contradiction operates as a buffer that absorbs predicted forcing-function activations until it is destroyed by a discrete agentic-novelty event.
The suspended-contradiction emerged as the dominant structural form of the cycle. Six chains predicted forcing-function activation that did NOT occur because the suspended contradiction continued to function. When the contradiction did break (Project Freedom, Briefing 029), it was via a novel architectural move (Sabbath Operationalization) not via the chain's specified mechanism. The principle: structural prediction must include a Suspended-Contradiction-Buffer term, calibrated to the political cost of explicit acknowledgment for both parties.
CITED CHAINS · 009-ie-2, 017-ie-1, 018-ie-1, 018-ie-2, 020-ie-2, 029-ie-1
mp-003 · provisional
Chains predicting outcomes via specified mechanisms produce Type-VII (Spurious Hit) errors when the actual mechanism is novel architectural composition rather than execution of a known mechanism — the briefing achieves narrative validation while the structural understanding is wrong.
Eight chains classified Held-Spurious. The pattern: chain Z arrives, but via a different mechanism than chain Y specified. Most concerning: this generates apparent confirmation of the briefing's analytical apparatus when the apparatus's structural understanding was actually wrong. Cycle 2+ should report Type-VII rate explicitly per chain and treat increase in Type-VII rate as a degradation signal even if total break rate falls.
CITED CHAINS · 002-ie-2, 004-ie-4, 012-ie-2, 013-ie-3, 014-ie-3, 020-ie-2, 022-ie-5, 024-ie-4
mp-004 · provisional
When the corridor news cycle absorbs a regional or systemic event into a single high-attention storyline, structural events at the periphery generate Type-I-eligible novel actions that fail to register in the corridor's analytic apparatus until they have already compounded.
The Peripheral Assertion pattern (META-1, Briefing 021) operates as a structural mechanism for Type-I-style breaks in chains that assume corridor-wide coverage. The principle is reflexive: it implicates the briefing's own attention budget. Six chains in this cluster broke or under-resolved because peripheral signals (Sahel, demographic, climate) compounded in the latency-phase between corridor cycles. Suggests Cycle 2+ allocate explicit non-corridor analytical bandwidth at chain-construction stage.
CITED CHAINS · 015-ie-3, 021-ie-1, 021-ie-5, 023-ie-2, 024-ie-1, 026-ie-2
mp-005 · provisional
Chains predicting institutional-coordination outcomes (FOMC unified hawkishness, OPEC discipline, EU veto coherence, ECB-Fed coordination) systematically break at step Z because late-modern coordination apparatuses produce dispersion (Verdict Compression, 8-4 votes, durable silences, multi-track re-routing) rather than coherence under stress.
Six chains (FOMC, Saudi-OPEC, Russian-track) broke at the institutional-coordination step. Verdict Compression (Briefing 026) — Powell's 8-4 dissent — is the canonical empirical instance: the apparatus produced a verdict but the reconciliation function had departed. Cycle 2+ should test in ECB voting splits, IMF Spring/Annual Meeting communiqué coherence, G7 substantive content, NATO alliance management, UN Security Council voting.
CITED CHAINS · 020-ie-3, 022-ie-5, 023-ie-1, 024-ie-1, 024-ie-4, 025-ie-3
mp-006 · provisional
Chains operating on rhetorical-declarative dynamics systematically over-weight the price-discovery transmission of those signals because marketplaces have learned to discount constraint-apparatus-absent declarations into a calibrated routine — the Sanctuary Discount.
The Sanctuary Discount (Briefing 030) is the empirical name for what was structurally happening across the cycle: each Trump declaration, Iranian counter-declaration, and ceasefire announcement produced smaller market reactions than the prior. Six chains over-weighted declarative transmission. The response architecture has been routinized into discount; perception is intact, credit-granting has hollowed.
CITED CHAINS · 003-ie-7, 006-ie-1, 013-ie-1, 014-ie-3, 029-ie-1, 030-ie-1
mp-007 · provisional
Chains that assume a regulatory-governance architecture is operative systematically miss the structural shift toward private-actor coordination as the substitute regime — the Capability-Governance Inversion — which operates at faster cadence than public regulation but bifurcates under participant-divergence stress (Coordination Bifurcation).
Seven chains involve the AI-governance, cyber-physical, or climate-financial architectures whose substitute-regime is private-actor coordination (Glasswing, ASL-4 partner-access, Cross-Lab Alignment Coalition). Briefing 025 named the bifurcation risk. The principle: chains conditioning on public regulation must include a substitute-regime-coordination-stability variable. EU AI Act trilogue collapse (Briefing 030) is the cycle-end instantiation.
CITED CHAINS · 007-ie-3, 019-ie-1, 020-ie-1, 021-ie-3, 024-ie-5, 025-ie-1, 030-ie-2
mp-008 · provisional
Chains terminating in 'X reaches level Y therefore Z follows' systematically miss buffer revelations — most consequentially, that the suspended-contradiction itself functions as a buffer not in the structural-vocabulary registry as such — and produce velocity miscalibration when the buffer absorbs the predicted pressure.
Six chains broke because a buffer absorbed pressure: SPR/futures-spot decoupling, suspended-contradiction architecture, Pakistan-Oman shuttle, back-channel diplomacy. The principle: every chain should explicitly enumerate buffers that could absorb the chain's pressure, including institutional-novelty buffers like suspended-contradiction. Connects to META-3 Threshold Cascade — but operates in reverse: the missing instantiation is Buffer Persistence rather than Buffer Collapse.
CITED CHAINS · 001-ie-1, 009-ie-2, 012-ie-1, 017-ie-1, 018-ie-2, 020-ie-2
mp-009 · provisional
Chains tracking structural patterns at one scale (actor, regime, institution) increasingly require multi-scale analysis because the post-2024 configuration produces simultaneous failures at multiple scales (Mali keystone-removal + UAE OPEC cartel-dissolution + War Powers statutory-bypass within the same week).
Five chains required multi-scale analysis to be even legible. The principle: contemporary structural configurations produce simultaneous failures at multiple scales — actor-scale (keystone removal), regime-scale (cartel dissolution), institutional-scale (statutory bypass), substitute-regime-scale (coordination bifurcation). The empirical signature of the post-2024 configuration is the multi-scale failure-pattern compound — and the absence of a corresponding response architecture.
CITED CHAINS · 023-ie-2, 024-ie-3, 025-ie-1, 025-ie-2, 026-ie-2
mp-010 · provisional · the LLM cognitive layer
Velocity miscalibration in chains generated by LLM plausibility engines is partly intrinsic to Knightian timing-indeterminacy and partly endogenous to the model's mean-trajectory regression and narrative-time compression. The two components cannot be separated from a single cycle's data; they require either cross-architecture comparison (different LLMs running the same Inference Engine) or held-out chain pairs (point-velocity framing vs half-life framing).
First meta-principle to make the LLM cognitive layer explicit. The Cycle 1 distribution shows Type-V × S1 + Type-V × S2 = 11 of 15 Type-V cases (73%): the dominant cognitive failure beneath the dominant structural failure is the conjunction of mean-trajectory regression on historical cascade analogues and narrative-time compression of multi-week physical processes. But Cycle 1 cannot tell us how much of the velocity miscalibration is Knightian floor (the world genuinely will not commit to a timing the chain author could have known) and how much is LLM-endogenous compression (a different architecture would not write the chain on the same clock). Recurrence in Cycle 2+ requires either (a) the same Type-V × S1/S2 dominance under prospective tagging per rec-007, or (b) a successful cross-architecture experiment showing differential signature distribution. The principle is the audit's first formal step toward operationalizing its diagnostic-of-LLM-cognition wing as something distinct from its phenomenology-of-Knightian-uncertainty wing.
CITED CHAINS · 001-ie-1, 009-ie-2, 012-ie-1, 013-ie-1, 024-ie-1

§5.6 — Vocabulary Curation

30-day sunset rule · META-6 forced decision · updates to the structural concepts taxonomy

Retirements

None this cycle. All 39 concepts in the structural concepts taxonomy were named within the cycle window (the cycle is the founding 30-day window). The sunset rule operates by re-citation: every named concept was substantively re-cited in 2+ briefings beyond its naming briefing, or is too recent to have had the opportunity. The Cycle 2 audit will be the first to apply meaningful retirement logic. Cycle 1 establishes the baseline.

Cycle 2 monitoring list

  • weekend-translation — named Briefing 028; only 2-3 days into existence at cycle close. Cycle 2 will be first opportunity to test re-citation.
  • sanctuary-discount — named Briefing 030; insufficient time for re-citation; Cycle 2 must verify.
  • optionality-arbitrage — named Briefing 001 (META-4); cited Briefing 025 (critical-mineral suspension window). Re-cited but limited; flag for Cycle 2 monitoring.

Reactivations

N/A in Cycle 1 — there are no prior cycles' retired concepts to reactivate.

META-6 Candidate Decision — Verification Asymmetry

Decision: tabled. Five distinct instantiations across the cycle satisfied the promotion criterion numerically, but each operates as Coupling Failure (META-1) at a slightly different scale.

META-6 candidate "Verification Asymmetry" was provisionally floated. Cycle 1 produced multiple instantiations: Verification-Mode Asymmetry (Briefing 020, formally added to the structural concepts taxonomy as a META-1 instance), the briefing's own verification-architecture reform (Briefing 023), the procurement-vs-deployment verification gap (Tesla Optimus vs Figure-BMW, Briefing 023), the public-information-symmetry-vs-Track-1.5 asymmetry (Briefing 028), and the cyber-physical defensive-cadence-asymmetry (Briefing 024). Five distinct instantiations across five different domains is, by the protocol's promotion criterion, sufficient for elevation to a meta-category.

However: each instantiation operates as Coupling Failure at a slightly different scale. Promoting Verification Asymmetry to META-6 would create overlap with META-1 that the taxonomy revision discipline is explicitly designed to prevent. The cleaner discipline is to TABLE the META-6 promotion and instead document the five Verification-Asymmetry instantiations as a Coupling-Failure cluster within META-1, with explicit cross-references.

Reopen condition

Verification Asymmetry should be reconsidered for META-6 promotion if Cycle 2 produces ≥3 additional instantiations that DO NOT cleanly map to Coupling Failure structure — i.e., instantiations where the asymmetry is not between observation and binding-action but between, e.g., two parallel verification regimes with no shared baseline (the structural form would then be more like Bypass Inversion at the verification level). If the new instantiations are again Coupling-Failure variants, the META-1 cluster treatment is correct and the META-6 candidate retires.

Concrete updates to apply to the structural concepts taxonomy

  • Update header line to: **Last updated:** 2026-05-04 (after Briefing 030; Cycle 1 audit attestation appended)
  • Append to document footer:
    ## Cycle 1 Audit Attestation

    Cycle 1 (Briefings 001-030) verified: All 39 concepts substantively re-cited within cycle window. Zero retirements. META-6 candidate "Verification Asymmetry" formally tabled per Cycle 1 audit; reopen conditions documented in the YAML companion. Cycle 2 monitoring list: weekend-translation, sanctuary-discount, optionality-arbitrage.
  • No concept additions retroactively required: Sanctuary Discount, Weekend Translation, Sabbath Operationalization, Sabbath Visibility, Cartel Dissolution, Keystone Removal, Verification-Mode Asymmetry, Coordination Bifurcation, Statutory Bypass, Peripheral Assertion, Credential Foreclosure, Dual-Track Maximalism were all added during cycle as part of daily briefing operation.

§5.7 — Anomaly Accounting

resolved / silently died / persistently absent · the briefing's epistemic honesty mechanism

Resolved 3 anomalies

  • "AI labs silent on Mythos-style scheming" (named Briefing 002) — Mythos risk report (Briefings 005-006) made scheming concrete; subsequent ASL-4 withholding decision (Briefing 019) made it institutional.
  • "Hezbollah veto on Lebanese foreign policy not tested" (named Briefing 010) — Tested and held in Washington meeting (Briefing 010); subsequently failed (Briefing 014 onward) as veto re-established operationally.
  • "AI prescription regulatory framework absent" (named Briefing 004) — Partially addressed by Forrester report and enterprise-discourse catch-up (Briefing 012); regulatory framework still absent but attention surfaced.

Silently Died 3 anomalies — diagnostic of attention drift

  • "Renewable transition discourse not accelerating despite chokepoint fragility" (named Briefing 001) — Stopped being tracked after Briefing 003. The discourse narrowed to oil-supply management; the energy-transition framing dropped from anomaly tracking. Re-emerges only as critical-mineral-chokepoint discourse (Briefings 014, 019).
  • "AI companies' silence on assassination targeting (IRGC named 18 US tech companies as targets)" (named Briefing 001) — Stopped being tracked after Briefing 002. Reclassification-as-combatants framework never developed; corridor moved to AI safety / Mythos / governance discourse instead.
  • "Settlement-velocity gap between rhetorical declarations and physical reality" (named Briefings 013, 014) — Absorbed into the Sanctuary Discount pattern (Briefing 030) — but the original anomaly framing (settlement-velocity-vs-mine-clearance-physics) silently faded into a different conceptual frame without explicit acknowledgment.

Persistently Absent 8 anomalies — the persistence is the diagnostic

  • "Congressional war authorization for Iran war" — named in Briefings 001, 005, 010, 016, 017, 018, 019, 025, 027, 028. 10 named instances. Statutory Bypass pattern (Briefing 025) formalized the persistent absence; War Powers Resolution silent retirement (Briefing 027) treats it as terminal-phase precedent.
  • "UNCLOS enforcement mechanism activated" — named in 001, 003, 005. International maritime legal architecture absent throughout cycle; never invoked. Persistent absence has hardened into structural fact.
  • "Chinese strategic positioning publicly articulated" — named in 001, 011, 015. China rejects blockade, exempts its own oil flow, but does not publicly articulate doctrine. The persistent ambiguity is the strategic position.
  • "Sudan/Sahel/Horn-of-Africa attention proportional to events" — named in 009, 010, 014, 021, 022. Persistently absent through Briefing 014, then partially surfaced through Briefings 021-024 as Mali keystone-removal compound. Pattern: peripheral signals receive corridor attention only when they instantiate META-1 or META-3 events at scale.
  • "Federal/regulatory response to DeepSeek V4 open-source release" — named in 020, 023, 025. BIS/SEC/NIST silence persisted through cycle close. Capability-Governance Inversion pattern (named in Briefing 019) formalized the persistent absence as operational regime.
  • "Saudi public statement on UAE OPEC departure" — named in 024, 025, 026, 027, 028, 029. Six consecutive briefings track the ongoing Saudi silence. The duration is the diagnostic. By Briefing 027 the silence is structurally distinctive of regime-internal cartel-restructuring at Royal Court level.
  • "Cross-Atlantic central-bank coordination communication" — named in 022, 023, 024, 025, 026, 028, 029, 030. Eight-briefing persistence. The Sabbath Visibility pattern (Briefing 022) initially named it; persists into Briefing 030 with no resolution.
  • "Major-bank risk-officer commentary on triple-convergence" — named in Briefing 030 (today). Persistence not yet established. Cycle 2 will determine whether this joins the persistently-absent cluster or resolves.
CLUSTER 1 · CONSTITUTIONAL-INSTITUTIONAL SILENCE
Congressional war authorization · UNCLOS enforcement · Federal AI/DeepSeek response · Saudi cartel-anchor statement · Cross-Atlantic central-bank coordination
Five persistently-absent items all involve institutional architectures that should be load-bearing under the cycle's stress events but are not. The clustering is itself the diagnostic of Capacity Hollowing (META-5) operating at scale across constitutional, treaty, regulatory, cartel, and central-bank architectures simultaneously.
CLUSTER 2 · PERIPHERY-ATTENTION DRIFT
Renewable transition discourse · Sudan/Sahel/Horn-of-Africa attention · Climate tipping signals as foreground
The corridor narrowing problem named in the daily briefing's editorial protocol is empirically demonstrated by these silent-death and persistently-absent items in Cycle 1. Peripheral Assertion (Briefing 021) is one structural diagnosis; the practical implication is that fresh-domain rotation must operate at the chain-construction stage, not at the chain-classification stage. Cycle 2+ will test whether the pattern improves under recalibration rec-004.

§5.8 — Theoretical Implications

three citation-ready paragraphs · exploratory in Cycle 1 per protocol §12

Forecasting under Knightian uncertainty

Cycle 1 produces an empirical phenomenology of forecast invalidation under deep uncertainty. Of 31 non-held conditional chains across the 30-day cycle, 48% broke at the velocity-miscalibration point (Type-V) and 29% at the spurious-hit point (Type-VII) — a distribution in which the theoretically loaded Knightian categories (Type-I agentic novelty as primary, Type-II competitive recursion) appear far less frequently than the velocity and structural-mechanism categories. The pattern suggests that under contemporary structural conditions, forecasting apparatuses systematically mis-time rather than mis-locate the structural transitions, and that the dominant failure mode is mechanism-substitution (the chain achieves narrative validation through a structurally different pathway) rather than outright invalidation. Meta-principle mp-002 — that suspended contradictions function as latent buffers absorbing predicted forcing-function activations — supplies a candidate empirical mechanism for theoretical work on buffered cascade dynamics: the pattern emerged here as a state-space configuration in which co-present contradictory states (ceasefire-and-blockade, talks-and-strikes) produced a forecast-failure mode the binary action-set framework did not contain.

LLM cognition under deep uncertainty

Cycle 1's break-point distribution combined with the LLM cognitive signature distribution (S2 32%, S7 29%, S4 13%, S1 and S3 10% each, S5 and S6 3% each) constitutes a within-architecture baseline: one model, one prompting setup, one editorial discipline, 30 daily briefings, 136 chains, 31 non-Held chains tagged across two parallel taxonomies. The natural next step is a multi-architecture experiment in which the same Inference Engine prompt is run against different LLM backbones over the same 30-day event field, with each backbone's chains tagged against the same S1–S7 taxonomy. The hypothesis the experiment would adjudicate is whether S1 (mean-trajectory regression) and S2 (narrative-time compression) dominance is general to LLM plausibility-engines under Knightian uncertainty — in which case mp-010 promotes to canonical and the cognitive-architecture-induced velocity miscalibration becomes an empirical phenomenon distinct from the Knightian timing-indeterminacy floor — or whether the dominance is architecture-specific, in which case the cross-architecture spread becomes the empirical anchor for cognitive-pluralism arguments about analytical-ensemble design. The audit's diagnostic-of-LLM-cognition wing is operational at this point and produces falsifiable cross-architecture comparisons.

AI–human ensembles for analytical work

The Contingency Audit itself is the cycle's most generative instance of the AI–human analytical ensemble it studies. The audit is performed by an AI partner (pattern-of-pattern recognition across 30 prior briefings) operating on the outputs of an AI partner (pattern recognition across daily structural events) under the periodic re-framing of a human partner (protocol design and recalibration discipline). Cycle 1 produced 10 substantive provisional meta-principles, the vast majority of which surface mechanisms (suspended-contradiction-as-buffer, calibrated-discount-as-response, coordination-bifurcation-under-stress, mean-trajectory-regression in chain timing) that no individual daily briefing could have surfaced from inside its own cycle. The §5.9 recalibrations are themselves the operational form of periodic re-framing, and the Cycle-2 verification criteria provide the falsification structure that distinguishes a working analytical ensemble from a recursive-narrowing collapse. The pattern that emerges across this cycle is that the periodic-re-framing layer is structurally necessary to the apparatus, not optional — without it, the daily layer's recursive narrowing has no counter-weight.

§5.9 — Recalibrations for the Next 30

3-5 concrete adjustments · with empirical basis, tradeoff, and verification criteria

Per protocol §12, Cycle 1 recalibrations are intentionally modest in scope — aggressive recalibration based on a single cycle would over-fit. Each recalibration cites the Cycle 1 empirical basis, names the tradeoff, and specifies how Cycle 2 will determine whether it took hold.

REC-001 · STATE-ACTOR CHAIN STRUCTURE
Inference Engine chains involving state actors with optionality must include an explicit Suspended-Contradiction-Buffer term naming the third option (durable silence, back-channel substitution, co-presence of contradictory states) the chain's binary action-set does not contain.
Empirical basis
Six chains (002-ie-1, 003-ie-1, 009-ie-2, 010-ie-1, 024-ie-1, 025-ie-3) broke at step Y because state actors with optionality used the third option. Meta-principle mp-001.
Tradeoff
Adds verbosity to chains; reduces apparent explanatory power because chains end with named uncertainty rather than confident terminal claims; risks reading as hedge.
Verification
Briefings 031–060 should show ≥75% of state-actor chains include an explicit Suspended-Contradiction-Buffer slot. The Cycle 2 audit will count slot-present vs slot-absent chains and report whether slot-present chains have a different break-type distribution. If Type-V/Type-VII rate drops in slot-present chains, the recalibration took hold.
REC-002 · SPURIOUS-HIT TEST PER CHAIN
Every chain terminating in 'Z follows from Y' must include a named Spurious-Hit Test: how would the chain be falsified if the surface outcome arrived via a structurally different mechanism? The test must be answerable with cycle-window observations.
Empirical basis
9 chains classified Held-Spurious (29% of all non-held; meta-principle mp-003). The briefing repeatedly achieved narrative validation while the structural understanding was wrong. Type-VII is the failure mode the briefing has the strongest incentive not to see.
Tradeoff
Forces chain authors to articulate falsification conditions before the surface outcome is known; some chains will be marked spurious even when structural intuition was substantially correct. The cost is honesty; the gain is cumulative credibility.
Verification
Briefings 031–060 should show every X→Y→Z chain accompanied by a Spurious-Hit-Test sentence. Cycle 2 audit will report Type-VII rate; if Type-VII falls below 20% (down from 29%), the recalibration is working.
REC-003 · SANCTUARY DISCOUNT FACTOR
Chains operating on rhetorical-declarative dynamics (presidential statements, ceasefire declarations, sanctions announcements) must explicitly weight transmission by the marketplace's accumulated calibration history — apply a Sanctuary Discount factor to predicted price/policy responses when the announcement-form has been routinized into discount.
Empirical basis
Six chains (003-ie-7, 006-ie-1, 013-ie-1, 014-ie-3, 029-ie-1, 030-ie-1) over-weighted declarative transmission. Sanctuary Discount (Briefing 030) is the empirical name for the cycle-wide pattern; meta-principle mp-006.
Tradeoff
Some declarations will produce outsized responses (the calibration is not infinitely durable); chains that apply the discount will under-predict in those cases. Asymmetry is acceptable because over-prediction has been the dominant cycle-1 error.
Verification
Briefings 031–060 should show explicit reference to marketplace-calibration-state in chains involving major declaration events. Cycle 2 should report whether chains containing the discount factor are more accurate at the price-transmission step than chains without it.
REC-004 · NON-CORRIDOR CHAIN ALLOCATION
Allocate explicit non-corridor analytical bandwidth at the chain-construction stage, not the chain-classification stage. The Inference Engine should generate ≥1 chain per briefing originating in a non-corridor domain (Africa, Latin America, South/Southeast Asia, demographics, critical minerals, climate-finance, robotics, quantum).
Empirical basis
Six chains (015-ie-3, 021-ie-1, 021-ie-5, 023-ie-2, 024-ie-1, 026-ie-2) broke or under-resolved because peripheral signals compounded in the latency-phase between corridor cycles. Meta-principle mp-004; Peripheral Assertion (Briefing 021); persistently-absent anomaly cluster on Sudan/Sahel/Horn-of-Africa.
Tradeoff
Some non-corridor chains will be premature or under-developed; chain-quality distribution will widen. Acceptable because Cycle 1 evidence is that corridor-only chain construction produces a structural blindspot Cycle 2+ cannot afford.
Verification
Briefings 031–060 should show ≥1 Inference Engine chain per briefing from a non-corridor domain (≥30 chains over the cycle). Cycle 2 audit will report count and whether non-corridor chains have a different break-type distribution (hypothesis: more Type-I, fewer Type-V) than corridor chains.
REC-005 · MULTI-SCALE COMPOUND FLAGGING
Multi-scale failure-pattern compounds (actor-scale + regime-scale + institutional-scale failures within the same week) should be explicitly named as such in the Force Interaction Matrix when they occur, and the Inference Engine should generate an additional chain per compound interrogating the substitute-regime architecture.
Empirical basis
Five chains (023-ie-2, 024-ie-3, 025-ie-1, 025-ie-2, 026-ie-2) required multi-scale analysis to be even legible. Meta-principle mp-009; the cycle-end week (Briefings 023-027) instantiated three simultaneous compound failures (Mali keystone-removal, UAE OPEC cartel-dissolution, War Powers statutory-bypass).
Tradeoff
Adds analytical load to the Force Interaction Matrix; risks producing chains that span scales the briefing's apparatus tracks unevenly. Acceptable because the post-2024 configuration's empirical signature appears to be exactly this compound.
Verification
Briefings 031–060 should show explicit multi-scale-compound flagging when simultaneous actor + regime + institutional failures occur. Cycle 2 audit will report compound count and whether substitute-regime chains generated from compounds have higher Held rates than chains derived from single-scale events.
REC-006 · HALF-LIFE INTERVALS REPLACE POINT TIMING
Every chain must include a half-life or temporal-uncertainty interval, not a point estimate of when Z arrives.
Empirical basis
Type-V dominance (48% of non-Held chains) plus the new LLM cognitive signature analysis: S1 (mean-trajectory regression) and S2 (narrative-time compression) jointly account for 13 of 31 primary signatures (42%) and concentrate in Type-V × {S1, S2} (11 of 15 Type-V chains). Half-life framing is robust to both — it surfaces the timing uncertainty rather than committing to a centroid the model averaged from training data, and it forces narrative compression to declare itself as a range rather than a date. Meta-principle mp-010 names this directly.
Tradeoff
Adds verbosity to chains; reduces apparent confidence in the chain's terminal step. Reading experience may feel more hedged. Some chains will lose rhetorical force when "within 14 days" becomes "half-life 30–90 days, modal 14 days."
Verification
Briefings 031–060 should show ≥80% of chains with explicit half-life or interval. Cycle 2 audit should show Type-V incidence drop OR show that the residual Type-V is structurally different in character (more S1-floor, less S2-narrative-compression).
REC-007 · PROSPECTIVE LLM SIGNATURE TAGGING
Every non-Held chain emitted in Briefings 031–060 gets tagged with its dominant LLM cognitive signature (S1–S7) at chain creation, not retrospectively.
Empirical basis
Cycle 1's signature classification is retrospective and inferential. The taxonomy was applied after the chain's break was already known, which means the tagging is at risk of being a rationalization of the break-type assignment rather than a prospective claim about which cognitive mechanism is most likely to fail. Prospective tagging produces cleaner data and forces the chain author to confront the model's likely failure mode at the moment of generation, when the prediction is still falsifiable.
Tradeoff
Adds a per-chain annotation step; risks the tagging itself becoming pro forma. Mitigation: signature must be selected with a one-sentence justification grounded in the chain text, not just a label. The justification field is where the discipline lives.
Verification
Briefings 031–060 chain entries should carry an llm_signature tag with one-sentence justification. Cycle 2 audit can compare prospective vs retrospective distributions on the Cycle 1 chains as validation: if cells align within ±20%, the retrospective tagging was approximately accurate. If they diverge, the cross-cycle promotion of mp-010 must wait.

§5.10 — LLM Cognitive Signature Analysis

parallel taxonomy · what failed in the model that generated the chain · cross-tabulated against Type I-VII

The audit has a dual purpose. The Type I–VII taxonomy classifies what failed in the world — the structural mechanisms by which contingent forecasts break under Knightian uncertainty. The S1–S7 LLM Cognitive Signature taxonomy classifies what failed in the model that generated the chain. The two taxonomies are parallel but not identical. The same chain can be (Type-V × S1) or (Type-V × S2) — different LLM cognitive failures producing the same structural break. The cross-tabulation is the analytical product.

Cycle 1's signature classification is retrospective. From Cycle 2 forward, signatures are tagged at chain creation per rec-007, which produces cleaner data. The retrospective pass establishes the taxonomy and provides Cycle 1's baseline distribution but should not be over-interpreted. The S1–S7 layer is operational at this point — the audit can serve as a field instrument for empirical work on LLM cognition under deep uncertainty, with cross-architecture replication as the natural next step.

S1–S7 Taxonomy

SignatureMechanismMaps to
S1 — Mean-trajectory regressionModel averages historical cascade analogues; specific-instance velocity smoothed toward training-data mean. Empirical signature: chains whose Y-step timing matches the centroid of historical analogues but not the actual instance.Type-V
S2 — Narrative over physicalCausal grammar parses cleanly; physics doesn't. Months get compressed to days because narrative tokens have their own clock. Empirical signature: chains where the inferential step is rhetorically compact but materially heavy.Type-V Type-VII
S3 — Plausibility-mass concentrationEmits most-probable completion; black-swan / low-prior actions under-weighted. Empirical signature: chains conditioning on the modal actor / move and missing the off-distribution actor or move.Type-I
S4 — Reflexive blindnessDoesn't simulate actors reading consensus prediction and adjusting; Hayek-Lucas-Goodhart territory invisible. Empirical signature: chains failing because the predicted actor read the prediction and acted to invalidate it.Type-II
S5 — Categorical stickinessAnchors on training-time categories when events have moved past them or the categories themselves are dissolving. Empirical signature: chains using a noun whose referent has shifted under it.Type-VI
S6 — Buffer blindnessTreats persistent states as constants; training data over-represents persistence vs transition. Empirical signature: chains assuming a buffer (SPR, dollar primacy, OPEC discipline, suspended-contradiction) and silent on its variability.Type-IV
S7 — Mechanism-outcome decouplingEmits outcome without binding to mechanism; any mechanism producing Z counts as confirmation. Empirical signature: chains whose Z-step is named generically enough that any mechanism producing Z reads as confirmation.Type-VII

Frequency Distribution — Primary LLM Cognitive Signature (n=31 non-held chains)

S2 — Narrative over PhysicalRHETORICAL COMPRESSION OF MULTI-WEEK PHYSICS10 (32%) S7 — Mechanism-Outcome DecouplingGENERIC Z ABSORBS ANY MECHANISM9 (29%) S4 — Reflexive BlindnessACTOR READ THE PREDICTION; MODEL DIDN'T NOTICE4 (13%) S1 — Mean-Trajectory RegressionHISTORICAL CENTROID AVERAGES SPECIFIC INSTANCE3 (10%) S3 — Plausibility-Mass ConcentrationMODAL COMPLETION; LOW-PRIOR ACTORS INVISIBLE3 (10%) S6 — Buffer BlindnessPERSISTENT STATES TREATED AS CONSTANTS (5 SECONDARY)1 (3%) S5 — Categorical StickinessCATEGORY REFERENT SHIFTED (3 SECONDARY)1 (3%)

S2 (narrative-time compression) and S7 (mechanism-outcome decoupling) jointly account for 19 of 31 primary signatures (61%). Together they describe an LLM that writes tight causal stories on the briefing's clock and names outcomes broadly enough that any mechanism producing them reads as confirmation. S6 buffer-blindness is under-represented as primary (1 case) but appears 5× as secondary — mirroring the Type-IV pattern from §5.4; both diagnose the same suspended-contradiction blindness from different sides.

Per-Signature Findings (Cycle 1)

S1 — Mean-trajectory regression 3 chains · 10%

Concentrated in chains modeling Saudi response to UAE OPEC departure (024-ie-1, 025-ie-3) and the canonical Hormuz-cascade chain (001-ie-1). In all three cases, the chain conditioned on the historical centroid of analogous events: every prior cartel-anchor episode (1973, 1986, 2014, 2020) produced a Saudi public statement within days; every prior chokepoint cascade arc compresses through similar multi-stage transmission. Durable Saudi silence and the SPR/futures-spot buffer were specific-instance properties the model averaged away.

S2 — Narrative over physical 10 chains · 32% · largest cell

The dominant primary signature of the cycle. Concentrated in Iran-cascade chains (004-ie-1, 005-ie-1, 006-ie-1, 009-ie-2, 011-ie-2, 012-ie-1, 013-ie-1, 016-ie-1) plus two Type-III chains where coupling assumption masks narrative compression (004-ie-2, 023-ie-1). The pattern: the chain composes a tight causal arc — communiqué → spot-futures gap → arbitrageurs absorb losses → bifurcation propagates — that is rhetorically compact but materially demands weeks of capital reallocation, regulatory action, or diplomatic motion. Settlement-velocity-vs-physical-reality is the canonical S2 failure shape.

S3 — Plausibility-mass concentration 3 chains · 10%

The Pakistan-arbiter cluster (010-ie-1, 011-ie-3, 015-ie-2): chains that emitted the modal Trump-IRGC bilateral completion and under-weighted the off-distribution Pakistan-Oman shuttle architecture that actually absorbed the period. Pakistan as a primary mediator is rare in training-data exemplars of Iran-US negotiations; the model concentrates plausibility on bilateralism. S3 also appears as secondary in 4 additional chains — its true incidence is higher than the primary count suggests.

S4 — Reflexive blindness 4 chains · 13%

The cleanest theoretical match. Every Type-II break in the cycle is also a primary S4: Trump anticipating his own posted threat (002-ie-1, 003-ie-1); Hezbollah declining to confirm the predicted barrage (006-ie-5); Israeli cabinet declining the predicted self-confirming-collapse (009-ie-3). In each case the predicted actor read the consensus and adjusted. The model emits the actor's action without modeling the actor's awareness of the prediction.

S5 — Categorical stickiness 1 primary · 3 secondary

One primary instance (010-ie-3): chain anchors on "Lebanese state" as a coherent autonomy-bearing actor whose referent had already begun to dissolve into a Hezbollah-veto compound. Three secondary appearances suggest the under-classification trap: categories may have been more sticky than the primary count surfaces, particularly around "FOMC unified" and "ceasefire scope" whose referents were drifting under the chain.

S6 — Buffer blindness 1 primary · 5 secondary

One primary instance (017-ie-1) where the suspended-contradiction architecture absorbed the predicted 36-hour escalation. The 5 secondary appearances cluster on the same suspended-contradiction blindness — a buffer the model could not see as a buffer because its persistence was the whole point. This is the cognitive-layer mirror of the Type-IV under-classification noted in §5.4: both diagnose the same blindspot from different sides.

S7 — Mechanism-outcome decoupling 9 chains · 29% · second-largest cell

The Type-VII cluster mapped one-to-one onto S7. Chains 002-ie-2, 004-ie-4, 012-ie-2, 013-ie-3, 014-ie-3, 019-ie-3, 020-ie-2, 022-ie-5, 024-ie-4 all share the structural form: Z is named generically enough (cascade-resolution-fails, suspended-contradiction-collapses-or-hardens, insurer-reprices, shadow-settlement-persists) that any mechanism producing Z confirms the chain. The cognitive failure: the model emits outcome without binding to mechanism. The risk noted in the YAML caveat: S7 and Type-VII are at risk of becoming a tautology if defined too loosely; Cycle 2 prospective tagging will help discipline the boundary.

Cross-Tabulation — Type × Signature

Type \ SigS1S2S3S4S5S6S7
Type-V (15)3831
Type-VII (9)9
Type-II (4)4
Type-III (3)21
Type-I (0)
Type-IV (0)
Type-VI (0)

Three cells carry 26 of 31 chains (84%): Type-V × S2 (8 chains; the largest single cell — narrative-time compression of multi-week processes), Type-VII × S7 (9 chains; mechanism-outcome decoupling on generic Z-steps), Type-II × S4 (4 chains; reflexive blindness when actors read the prediction). Type-V × S1 (3 chains) is the Saudi-silence and Hormuz-cascade cluster — chains whose Y-step timing matches the centroid of historical analogues but not the specific instance. The Type-V column shows the cycle's clearest cross-tabulation: the same structural break (timing miscalibration) emerged from four distinct cognitive failures (S1, S2, S3, S6), confirming that signature classification is not redundant with break-type classification — it diagnoses a different layer.

Honest Caveat

Three Cycle 1 calls reasonable readers might dispute. (1) The Type-V × S1 vs Type-V × S2 split. Many Type-V chains exhibit both signatures; the primary assignment depends on which is judged the proximate cognitive failure. Saudi-silence chains are unambiguously S1; Iran-cascade compression chains are unambiguously S2; chains that involve both centroid-averaging and narrative-compaction were assigned by which feels more load-bearing in the chain text. (2) S7 and Type-VII as near-synonyms. Mechanism-outcome decoupling and spurious-hit are nearly the same construction at different levels. The risk is tautology; the discipline is to write the S7 justification grounded in the chain's specific Z-step rather than re-stating that the outcome arrived via a different mechanism. Cycle 2 prospective tagging is the test. (3) S6 buffer-blindness as primary. Only one chain (017-ie-1) was assigned S6 primary, but the suspended-contradiction blindness is arguably the deepest cognitive failure of the cycle — appearing 5× as secondary. The conservative call kept S6 at 1 primary because each individual chain's proximate failure was timing or generic-Z; an aggressive reader could promote 4-5 of those secondary appearances to primary, which would reshape the distribution.