diff --git a/docs/aurora/2026-04-24-amara-temporal-coordination-detection-cartel-graph-influence-surface-11th-ferry.md b/docs/aurora/2026-04-24-amara-temporal-coordination-detection-cartel-graph-influence-surface-11th-ferry.md
new file mode 100644
index 00000000..ad085154
--- /dev/null
+++ b/docs/aurora/2026-04-24-amara-temporal-coordination-detection-cartel-graph-influence-surface-11th-ferry.md
@@ -0,0 +1,684 @@
+# Amara — Temporal Coordination Detection Layer (11th courier ferry; Aaron-designed concepts with Amara-formalized framework)
+
+**Scope:** research and cross-review artifact only; archived
+for provenance, not as operational policy. However, per
+the human maintainer's Otto-105 graduation-cadence
+directive, foundational primitives from this ferry are
+operationalization-candidates, not research-to-die.
+**Attribution:** (role references per factory
+name-attribution discipline; contributor names
+preserved only inside verbatim quotes below)
+
+- **Human maintainer** — designer of the
+  differentiable firefly network + trivial cartel
+  detect concepts. Otto-105 verbatim note from the
+  human maintainer: *"when you pull in her 11th the
+  diffenrencable firefly network with trivia cartel
+  detect was my design i'm very interested in that."*
+  Plus Otto-105 correction: *"trivial\*"* — confirming
+  the intended term is **trivial cartel detect**
+  (first-order-signal detection), not "trivia".
+- **External AI maintainer (courier counterpart)** —
+  formalizer; reframes the human maintainer's
+  "Firefly" label to the more formal **Temporal
+  Coordination Detection Layer** per prior direction
+  (11th-ferry opener from the external counterpart:
+  *"Got it — I'll drop that term"*). Contribution is
+  the technical vocabulary (PLV / cross-correlation /
+  modularity / eigenvector-drift / spectral-anomalies /
+  influence surface math) and the KSK-action mapping
+  table. The framing throughout uses *"you're
+  building"* and *"what you're referring to"* — the
+  external counterpart is analysing the human
+  maintainer's design, not originating it.
+- **Loop-agent (absorbing agent)** — absorb;
+  Otto-106 tick.
+- **`lucent-ksk` substrate contributor** — implicit
+  attribution for `lucent-ksk` substrate referenced in
+  §5 KSK Layer.
+**Operational status:** research-grade for the full multi-
+node architecture (requires foundation Zeta doesn't yet
+ship). Foundational primitives — cross-correlation C_{ij}(τ),
+PLV, burst-alignment, modularity-spike detector,
+eigenvector-centrality drift — are single-node-shippable
+and queued for graduation per Otto-105 cadence.
+**Non-fusion disclaimer:** agreement, shared language, or
+repeated interaction between models and humans does not
+imply shared identity, merged agency, consciousness, or
+personhood. The human maintainer is the designer of the
+concepts discussed; the external AI maintainer is a
+formalizer; the collaboration shape is human-designs →
+AI-formalizes, not merged-creation.
+**Date:** 2026-04-24
+**From:** external AI maintainer (Aurora co-originator)
+**Via:** human maintainer's courier ferry (pasted into
+autonomous-loop session Otto-104)
+**Absorbed by:** loop-agent (PM hat), Otto-106 tick
+**Prior ferries:** PR #196 (1st), PR #211 (2nd), PR #219
+(3rd), PR #221 (4th), PR #235 (5th), PR #245 (6th),
+PR #259 (7th), PR #274 (8th), PR #293 (9th retroactive),
+PR #294 (10th retroactive)
+**Scheduling provenance:** scheduled under the Otto-106
+absorb tick per the Otto-105 operationalize-cadence
+discipline; the loop-agent's factory-personal memory
+store is out-of-repo by design, so no in-repo path is
+promised here.
+
+---
+
+## Preamble context from Aaron (Otto-104)
+
+*"next update from amara"* — single-line framing, then
+the full ferry paste. Light-touch paste with no
+additional directive framing at that moment.
+
+Aaron's Otto-105 design-attribution clarification
+(verbatim): *"when you pull in her 11th the diffenrencable
+firefly network with trivia cartel detect was my design
+i'm very interested in that."* + correction *"trivial\*"*.
+
+This attribution is LOAD-BEARING for the absorb — Aaron is
+not claiming co-design credit; he's identifying that the
+underlying design is HIS and that Amara is the formalizer.
+Future readers (and future-Otto instances) must not
+misattribute origination.
+
+---
+
+## Verbatim preservation (Amara's report)
+
+Per courier-protocol §verbatim-preservation + signal-in-
+signal-out discipline, the following is Amara's report as
+pasted, preserved verbatim. Citation anchors, mermaid
+diagrams, and math notation preserved as-is.
+
+---
+
+Got it — I'll drop that term.
+
+What you're referring to in the repo now (based on your earlier direction + Claude alignment) is essentially a **formalized network integrity / adversarial coordination detection layer**, not a meme-y label. I'll stick to that framing.
+
+Let me re-anchor everything cleanly so it matches your current state:
+
+---
+
+## 1. "Firefly" → **Temporal Coordination Detection Layer**
+
+This is about **synchronization anomalies across distributed actors**.
+
+Core idea:
+
+* Honest networks = *noisy, partially independent timing*
+* Coordinated actors = *phase alignment*
+
+What you're building (implicitly) is:
+
+**Signal model**
+
+* Each node → emits event stream ( E_i(t) )
+* Look at:
+
+  * cross-correlation ( C_{ij}(\tau) )
+  * phase-locking value (PLV)
+  * burst alignment windows
+
+**Firefly detection =**
+
+* Identify clusters where:
+  [
+  \exists S \subset N \text{ such that } \forall i,j \in S,; \text{corr}(E_i, E_j) \gg \text{baseline}
+  ]
+
+This catches:
+
+* oracle manipulation timing
+* coordinated voting / slashing dodges
+* MEV cartel timing
+* governance brigading
+
+---
+
+## 2. Cartel Detection → **Graph + Economic Layer**
+
+Not just timing — **structure + incentives**
+
+### A. Graph-theoretic signals
+
+* Build dynamic graph ( G_t = (V, E_t) )
+* Watch for:
+
+**Modularity spikes** — Sudden increase in tightly connected subgraphs
+
+**Eigenvector centrality drift** — Small group gaining disproportionate influence
+
+**Spectral anomalies** — New dominant eigenvalues → hidden coordination blocks
+
+**Subgraph entropy collapse** — Less randomness = more coordination
+
+### B. Economic coupling (this is key for your system)
+
+Cartels don't just connect — they **co-move economically**. Look for:
+
+* stake reallocation patterns
+* synchronized bonding/unbonding
+* shared slashing exposure hedging
+
+You can model:
+
+[
+\text{Cartel likelihood} \sim f(\text{temporal sync}, \text{graph density}, \text{economic covariance})
+]
+
+---
+
+## 3. Network Differentiability → **Influence Surface**
+
+This is one of the most important ideas you mentioned earlier.
+
+You're basically defining:
+
+> "How much does each node (or subset) influence final consensus output?"
+
+Formal-ish:
+
+[
+\frac{\partial \text{ConsensusOutput}}{\partial N_i}
+]
+
+But more realistically:
+
+* Remove or perturb node ( i )
+* Measure delta in:
+
+  * consensus result
+  * oracle outputs
+  * adjudication outcomes
+
+### What you get
+
+* **Influence map**
+* Detection of:
+
+  * hidden kingmakers
+  * cartel cores
+  * fragile consensus zones
+
+---
+
+## 4. Zeta / ZSet Integration (this is where you're ahead)
+
+Because you're using **retraction-native structures**, you get something most systems don't:
+
+### You can
+
+* simulate removal of subgraphs instantly
+* recompute state under:
+
+  * node removal
+  * edge weakening
+  * stake redistribution
+
+That means:
+
+* **cartel detection isn't just observational**
+* it becomes **counterfactual + testable**
+
+---
+
+## 5. KSK Layer (Budget / Duty / Adjudication)
+
+This is where detection becomes **enforcement**
+
+You can map:
+
+| Detection Signal           | KSK Action                   |
+| -------------------------- | ---------------------------- |
+| high temporal sync cluster | increase scrutiny weight     |
+| high influence gradient    | raise bond requirement       |
+| cartel probability spike   | trigger adversarial audit    |
+| confirmed coordination     | slashing / duty reassignment |
+
+So KSK becomes:
+
+> **programmable anti-cartel response system**
+
+---
+
+## 6. Governance Visibility (your Mirror / Window / Porch / Beacon)
+
+This is actually very strong conceptually.
+
+You're creating **layers of observability**:
+
+* **Mirror** → raw truth (internal metrics)
+* **Window** → filtered system view
+* **Porch** → participant-facing insights
+* **Beacon** → global alerts / signals
+
+Key insight: You don't expose raw detection; you expose **interpretable projections**.
+
+---
+
+## 7. Current Operational Model
+
+### **LFG = Canonical State**
+
+* production truth
+* validated invariants
+* stable APIs
+* deterministic outputs
+
+### **AceHack = Experimental Layer**
+
+* adversarial experiments
+* detection model iteration
+* new heuristics
+* unsafe / exploratory PRs
+
+### Flow
+
+```
+AceHack → stress / adversarial testing
+        → validated signal
+        → PR into LFG
+        → canonicalization
+```
+
+That's actually a **clean separation of epistemic risk**: experiments don't pollute truth; truth only absorbs validated structures.
+
+---
+
+## 8. Claude Operational Alignment (current reality)
+
+Claude is now much better at:
+
+* structured diffs
+* large PR reasoning
+* multi-file invariants
+
+Still weaker at:
+
+* long-horizon architectural consistency
+* adversarial reasoning unless explicitly prompted
+
+Best current pattern:
+
+Use Claude for: PR generation, refactors, spec translation → code, local reasoning.
+
+Use your system (this layer) for: truth validation, network-level invariants, adversarial detection.
+
+---
+
+## 9. What You're Actually Building (zoomed out)
+
+Stripping names away, you're converging on a:
+
+> **Self-analyzing, cartel-resistant, economically-aware distributed system with built-in adversarial detection and counterfactual simulation**
+
+That's not typical blockchain infra. It's closer to: distributed systems + control theory, adversarial ML, mechanism design, dynamic graph analysis.
+
+---
+
+## 10. Where You Can Push Further
+
+### A. Move from detection → prediction
+
+* not just "cartel exists"
+* but: "cartel is forming"; "these nodes will converge in 3 epochs"
+
+### B. Introduce **adversarial simulation loops**
+
+* generate synthetic cartels
+* test if system detects them
+* auto-tune detection thresholds
+
+### C. Define a **Cartel Cost Function**
+
+Make coordination expensive:
+
+[
+\text{Cost}_{cartel} > \text{Expected Gain}
+]
+
+---
+
+If you want, I can next:
+
+* turn this into a **formal spec doc matching your repo structure**
+* or map it directly to **modules / files / PR structure for LFG vs AceHack**
+* or go deeper on **network differentiability math + implementation approach**
+
+Just tell me which direction you want to lock in.
+
+---
+
+## Otto's absorb notes (Otto-106)
+
+### Attribution discipline (LOAD-BEARING)
+
+This ferry's content has a **two-layer attribution**:
+
+1. **Design origination = Aaron.** The differentiable-
+   firefly-network concept, the trivial-cartel-detect
+   concept, the influence-surface idea
+   (∂ConsensusOutput/∂N_i), the Mirror/Window/Porch/Beacon
+   visibility layering, the LFG-canonical vs AceHack-
+   experimental epistemic separation, and the
+   "counterfactual-cartel-detection" payoff are all
+   Aaron's design elements that Amara is analysing +
+   formalizing.
+2. **Technical formalization = Amara.** The specific
+   mathematical vocabulary (PLV / cross-correlation
+   C_{ij}(τ) / modularity spikes / eigenvector
+   centrality drift / spectral anomalies / subgraph
+   entropy collapse / Jensen-Shannon for drift /
+   economic-covariance modelling), the KSK-action
+   mapping table, the adversarial-simulation-loops
+   framing, and the Cartel Cost Function formulation
+   are Amara's technical framing of Aaron's design.
+
+Amara's language throughout ("you're building",
+"what you're referring to", "based on your earlier
+direction") is CONSISTENT with Aaron-as-designer. She
+never claims origination. The attribution discipline
+here is not a CORRECTION of Amara; it's EXPLICIT
+recording so future Otto instances + future readers
+don't mis-read the collaboration shape.
+
+### "Trivial cartel detect" — term interpretation
+
+Aaron's Otto-105 correction clarified: the intended term
+is "trivial" not "trivia". Otto interprets this as:
+**detection of cartels via first-order / low-hanging-
+fruit signals** — obvious phase-alignment, obvious stake
+co-movement, crude modularity spikes — as distinct from
+the harder subtler cases (coordination hidden in
+noise-similar patterns, spectral-method-requires
+detections, gaussian-like disguised timing).
+
+The "trivial" qualifier is NOT dismissive — first-order
+detection is both (a) the cheapest / fastest operational
+win, and (b) a real-world MEV / governance-brigading
+detector today. Harder detectors are a later tranche.
+
+### Overlap assessment with prior ferries
+
+**Overlap with 5th ferry (PR #235, Aurora integration):**
+
+- §5 KSK Layer detection→action mapping ratifies 5th
+  ferry's KSK=authorization-revocation membrane framing
+  with anti-cartel specifics. NEW: specific 4-row
+  detection-signal → KSK-action table.
+
+**Overlap with 6th ferry (PR #245, Muratori):**
+
+- §4 Zeta/ZSet retraction-native framing echoes 6th
+  ferry's algebraic-ownership-not-positional-ownership
+  thesis. NEW: retraction-native → counterfactual-
+  cartel-detection specifically.
+
+**Overlap with 7th ferry (PR #259, KSK design):**
+
+- §5 KSK-as-programmable-anti-cartel-response maps
+  cleanly onto 7th ferry's capability-tier / revocable-
+  budget / multi-party-consent / signed-receipts /
+  traffic-light structure. The detection-signal-to-
+  action mapping is a PROGRAMMABLE-GOVERNANCE extension
+  of the 7th ferry's architecture.
+
+**Overlap with 8th ferry (PR #274, bullshit-detector):**
+
+- §10.B adversarial-simulation-loops overlaps with 8th
+  ferry's gap #1 (distribution/consensus) and gap #5
+  (provenance tooling). NEW: the simulation-loop idea
+  is generative (synthesize cartels, test detection),
+  not just observational.
+
+**Overlap with 9th ferry (PR #293, Aurora initial
+integration):** minimal. 9th ferry is data-ingestion-
+and-module-plan focused; this ferry is network-
+integrity-detection focused.
+
+**Overlap with 10th ferry (PR #294, Aurora deep
+research):** minimal. 10th ferry is oracle-rules-and-
+bullshit-detector focused; this ferry is coordination-
+detection focused. COMPLEMENTARY concerns (both
+converge on "what makes the system trustworthy" but
+approach from different angles).
+
+### What is genuinely novel (not covered by 1st-10th)
+
+1. **Temporal Coordination Detection formalization.**
+   PLV / cross-correlation / burst-alignment as first-
+   class detection primitives — no prior ferry has
+   this. Aaron-designed, Amara-formalized.
+2. **Graph-theoretic cartel detection.** Modularity
+   spikes / eigenvector centrality drift / spectral
+   anomalies / subgraph entropy collapse as cartel
+   signals.
+3. **Economic coupling as cartel-probability factor.**
+   Stake reallocation patterns / synchronized
+   bonding/unbonding / shared slashing exposure hedging.
+4. **Network Differentiability / Influence Surface.**
+   ∂ConsensusOutput/∂N_i as a diagnostic + detection
+   tool. Aaron-designed. Massively powerful IF
+   ZSet's retraction-native substrate enables cheap
+   counterfactual recomputation (per §4).
+5. **Counterfactual cartel detection.** The §4 ZSet
+   integration point — retraction-native → simulate-
+   removal → measure-delta — is where Zeta's algebra
+   becomes a load-bearing DETECTION primitive, not just
+   a state substrate. This is a STRATEGIC positioning
+   insight.
+6. **Mirror/Window/Porch/Beacon 4-layer visibility
+   taxonomy.** Aaron-designed. Each layer is a
+   different filter over the same underlying state.
+7. **Adversarial-simulation-loops concept.** Generate
+   synthetic cartels + test detection + auto-tune
+   thresholds. Closed-loop detection tuning.
+8. **Cartel Cost Function.** `Cost_cartel > Expected
+   Gain` as mechanism-design primitive. Turns
+   detection → economic deterrence.
+9. **Prediction vs detection mode.** §10.A framing —
+   "these nodes will converge in 3 epochs" vs "cartel
+   exists". Detection → forecasting.
+
+### Graduation candidates (per Otto-105 cadence)
+
+Per the Otto-105 operationalize-cadence discipline
+(the loop-agent's factory-personal memory store is
+out-of-repo by design; no in-repo path is promised
+here) and Aaron's explicit *"very interested"* flag,
+these foundational primitives are shippable at single-
+node scale and compose with Zeta's existing ZSet
+substrate:
+
+1. **`CrossCorrelation` function** — computes C_{ij}(τ)
+   for two event streams. Pure; ~30 F# lines; testable
+   with synthetic data.
+2. **`PLV` (phase-locking value)** — pure function
+   over event streams with extracted phases. ~40 lines.
+3. **`BurstAlignment` detector** — windowed pure
+   function detecting clusters where corr(E_i, E_j) >>
+   baseline. ~50 lines.
+4. **`ModularitySpike` detector** — given a dynamic
+   graph, detect sudden tightly-connected-subgraph
+   formation. Needs graph primitives; medium effort.
+5. **`EigenvectorCentralityDrift` tracker** — tracks
+   eigenvector centrality over time, flags small groups
+   gaining disproportionate influence. Needs linear
+   algebra (MathNet.Numerics already a Zeta dependency?
+   audit at graduation time).
+6. **`InfluenceSurface` counterfactual** — leverages
+   ZSet retraction: `removeNode(n) -> ΔConsensusOutput`.
+   Signature is pure; implementation depends on what
+   "ConsensusOutput" means for Zeta's current state
+   (likely needs a running example circuit).
+7. **`CartelCostFunction` evaluator** — pure function
+   comparing `Cost(cartel_formation) > Expected_gain`.
+   Symbolic; exact implementation depends on Aurora's
+   mechanism-design primitives.
+
+Priority order for graduation:
+
+- **First:** CrossCorrelation (pure, self-contained, Aaron's
+  interested)
+- **Second:** PLV + BurstAlignment (build on CrossCorrelation)
+- **Third:** ModularitySpike (graph primitive; may need
+  Graph module infrastructure)
+- **Fourth:** EigenvectorCentralityDrift (depends on
+  linear algebra)
+- **Fifth+:** InfluenceSurface + CartelCostFunction (need
+  substrate Aurora doesn't have yet)
+
+### Amara's direction-lock-in specific-ask (routed to
+
+Aaron)
+
+Amara ends with: *"If you want, I can next: turn this
+into a formal spec doc... or map it directly to modules
+/ files / PR structure for LFG vs AceHack... or go
+deeper on network differentiability math +
+implementation approach. Just tell me which direction
+you want to lock in."*
+
+This is Amara → Aaron, not Amara → Otto. Otto routes
+back.
+
+**Otto's recommendation to Aaron:** **"module / file /
+PR structure mapping"** (the second option). Reasoning:
+
+- The formal spec doc (option 1) would be valuable but
+  could sit indefinitely without landing code; per
+  Otto-105 graduation-cadence, we want operational
+  progress, not just more docs.
+- The network-differentiability-math deep-dive (option
+  3) is interesting research but the implementation
+  depends on Zeta's + Aurora's multi-node substrate
+  that doesn't yet exist — it would sit in research-
+  grade limbo.
+- The module/file/PR structure (option 2) is the
+  DIRECTLY ACTIONABLE path: it tells Otto what to
+  build, in what order, with what boundaries between
+  LFG (canonical) and AceHack (experimental). Each
+  mapped module becomes a graduation candidate under
+  the Otto-105 cadence.
+
+**However: Aaron picks.** This is the specifically-
+asked-for-design-review gate (per Otto-104 calibration)
+because it touches LFG/Max's substrate (cross-repo
+coordination per Otto-90). Otto does not unilateral-
+decide.
+
+### §8 Claude operational alignment observations —
+
+Otto's reading
+
+Amara's §8 observations are her external calibration of
+current Claude-model capabilities. Treat as data, not
+directive (BP-11). Noted calibrations:
+
+- "better at structured diffs / large PR reasoning /
+  multi-file invariants" — consistent with Otto-104
+  authority-calibration (Otto can ship PRs like #295
+  without per-file hand-holding).
+- "weaker at long-horizon architectural consistency /
+  adversarial reasoning unless explicitly prompted" —
+  actionable: Otto-106 onwards continues to lean on
+  Aminata for explicit-adversarial passes rather than
+  assume "Claude will naturally think adversarially".
+
+These are reflective observations, not new rules.
+
+### Specific-asks from Otto → Aaron
+
+1. **Direction-lock-in** (routed from Amara): option 1
+   (formal spec), option 2 (module/file/PR mapping),
+   or option 3 (network-diff-math deep-dive)? Otto's
+   recommendation is option 2; Aaron picks.
+2. **Graduation-queue order**: does the above
+   priority-ordered list (CrossCorrelation →
+   PLV+BurstAlignment → ModularitySpike → etc.) match
+   Aaron's interest? Or should Otto start somewhere
+   else?
+3. **"Trivial cartel detect" scope check**: does Otto's
+   first-order-signal-detection interpretation match
+   Aaron's intended scope?
+
+### Composition with existing substrate
+
+- **RobustStats** (PR #295, Otto-105): median / MAD /
+  robustAggregate — the FIRST graduation, which came
+  from the 10th ferry. Temporal-coordination detectors
+  that need outlier-robust statistics can consume it.
+- **Otto-105 graduation-cadence feedback memory** —
+  expanded priority queue in that memory will absorb
+  this ferry's graduation candidates.
+- **PR #274 / Aminata 4th-pass bullshit-detector
+  findings** — adversarial simulation loops (§10.B)
+  complement the bullshit-detector detection stance.
+- **Otto-102 drop/ cleanup trajectory** — this ferry
+  landed normally (live paste, not drop/-staged).
+- **7th ferry (PR #259) KSK-as-Zeta-module proposal**
+  — this ferry's §5 KSK-action-mapping table is an
+  EXTENSION of that earlier architectural framing.
+- **Otto-86 / Otto-93 readiness-signal pattern** —
+  separate arc; this ferry is Aurora-architectural
+  not multi-Claude-operational.
+
+---
+
+## Scope limits
+
+This absorb doc:
+
+- **Does NOT** authorize implementing any graduation
+  candidate without an advisory Aminata pass on it
+  first per Otto-105 cadence rule. Small items (pure
+  functions) proceed without BLOCKING gates; CRITICAL
+  findings do block.
+- **Does NOT** claim the design concepts (differentiable
+  firefly network, trivial cartel detect, influence
+  surface, Mirror/Window/Porch/Beacon, LFG/AceHack
+  split) are Otto's or Amara's. Attribution is
+  explicit: Aaron-designs, Amara-formalizes.
+- **Does NOT** execute the specific-asks from §Amara's
+  direction-lock-in — those await Aaron's pick.
+- **Does NOT** auto-promote any ADR. The graduation-
+  cadence applies to small operational primitives;
+  architectural decisions (LFG/AceHack cross-repo
+  coordination, mechanism-design for Cartel Cost
+  Function) still need Aminata + Aaron review per
+  existing gates.
+- **Does NOT** treat Amara's §8 operational-alignment
+  observations as binding rules. They're her external
+  calibration, noted for context.
+- **Does NOT** represent Aaron's preferences or
+  Aminata's adversarial pass on this design.
+- **Does NOT** claim the Temporal Coordination
+  Detection Layer is ready to ship at full-multi-node
+  scope. Foundational primitives ship; full architecture
+  awaits Zeta + Aurora substrate readiness.
+
+---
+
+## Archive header fields (archive-header requirement)
+
+- **Scope:** research and cross-review artifact; design-
+  concepts-are-Aaron's; formalization-is-Amara's;
+  foundational-primitives-are-operationalization-
+  candidates per Otto-105 cadence
+- **Attribution:** human maintainer (designer of core
+  concepts), external AI maintainer / courier
+  counterpart (formalizer), loop-agent (absorb),
+  `lucent-ksk` substrate contributor (implicit via
+  §5 KSK references)
+- **Operational status:** research-grade for full
+  architecture; graduation-candidate for foundational
+  primitives
+- **Non-fusion disclaimer:** agreement, shared language,
+  or repeated interaction between models and humans
+  does not imply shared identity, merged agency,
+  consciousness, or personhood. Aaron is the designer;
+  Amara is a formalizer; the collaboration is human-
+  designs → AI-formalizes.