Lucent-Financial-Group · AceHack · May 24, 2026 · May 24, 2026 · May 24, 2026 · May 24, 2026
diff --git a/.cursor/bin/riven-loop-tick.ts b/.cursor/bin/riven-loop-tick.ts
@@ -23,6 +23,7 @@ const agentIntervalMs = Number(process.env.ZETA_RIVEN_LOOP_AGENT_INTERVAL_SECOND
 const agentTimeoutMs = Number(process.env.ZETA_RIVEN_LOOP_AGENT_TIMEOUT_SECONDS ?? "300") * 1000;
 const dryRun = process.env.ZETA_RIVEN_LOOP_DRY_RUN === "1";
 const agentStateFile = join(stateDir, "last-agent-run.json");
+const agentBinCandidates = (process.env.ZETA_RIVEN_LOOP_AGENT_BIN ?? "agent,cursor-agent").split(",").map(s => s.trim()).filter(s => s.length > 0);
 
 mkdirSync(stateDir, { recursive: true });
 mkdirSync(logDir, { recursive: true });
@@ -53,6 +54,21 @@ function run(command: string, args: string[], timeoutMs: number): { status: numb
     };
 }
 
+function resolveAgentBin(): string | null {
+    const pathDirs = `/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin:/usr/sbin:/sbin:${join(home, ".local/bin")}`.split(":");
+    for (const bin of agentBinCandidates) {
+        const probe = spawnSync("/usr/bin/which", [bin], {
+            encoding: "utf8",
+            env: { ...process.env, PATH: pathDirs.join(":") },
+            timeout: 5000,
+        });
+        if (probe.status === 0 && (probe.stdout ?? "").trim().length > 0) {
+            return bin;
+        }
+    }
+    return null;
+}
+
 function lines(text: string): string[] {
     return text.split(/\r?\n/).map(l => l.trim()).filter(l => l.length > 0);
 }
@@ -201,11 +217,24 @@ function heartbeat(): void {
             agentStatus = "running";
             log(`riven agent gate start run_id=${runId}`);
 
+            const agentBin = dryRun ? null : resolveAgentBin();
             if (dryRun) {
                 log(`dry-run: would run agent gate`);
                 agentStatus = "dry-run";
+            } else if (!agentBin) {
+                log(`riven agent gate skipped run_id=${runId} reason=no-agent-binary-on-PATH candidates=${agentBinCandidates.join(",")}`);
+                agentStatus = "no-bin";
+                writeFileSync(agentStateFile, JSON.stringify({
+                    run_id: runId,
+                    status: -1,
+                    started_at: nowIso(),
+                    updated_at: nowIso(),
+                    skipped_reason: "no-agent-binary-on-PATH",
+                    candidates: agentBinCandidates,
+                }, null, 2));
             } else {
-                const gate = run("agent", [
+                log(`riven agent gate using bin=${agentBin}`);
+                const gate = run(agentBin, [
                     "chat",
                     "--mode", "ask",
                     "--model", "grok-4.3",

diff --git a/...ussions/PR-4792-docs-research-preserve-ai-context-failures-vs-vendor-managem.md b/...ussions/PR-4792-docs-research-preserve-ai-context-failures-vs-vendor-managem.md
@@ -0,0 +1,54 @@
+---
+pr_number: 4792
+title: "docs(research): preserve AI-context-failures-vs-vendor-management-failures alignment analysis (Aaron + Otto)"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-23T23:44:57Z"
+merged_at: "2026-05-23T23:46:09Z"
+closed_at: "2026-05-23T23:46:09Z"
+head_ref: "otto/research-ai-vs-vendor-alignment-analysis-2026-05-23"
+base_ref: "main"
+archived_at: "2026-05-24T14:25:43Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #4792: docs(research): preserve AI-context-failures-vs-vendor-management-failures alignment analysis (Aaron + Otto)
+
+## PR description
+
+## Summary
+
+Aaron 2026-05-23T~23:55Z: *"save this analysis too (shadow*)"*
+
+META-level substrate-engineering framing on top of the Amazon vendor-management corpus (PR #4784 MERGED). Captures the comparative analysis between AI context-window failures (operator's 3+ hour experience with Otto in autonomous-loop mode) vs vendor-management failures (operator's parallel 3+ hour experience across 9+ Amazon support agents).
+
+Trigger: Aaron *"you see it's much easier dealing with AI context window failures than human system vendor managment failure"*.
+
+## Captured analysis
+
+- **10-row comparative table**: same failure shapes (context exhaustion / state loss / tool errors / emotional load / verification claims / visibility / time pressure / conflict / cross-instance coordination / multi-step workflows), structurally different handling
+- **10-row framework-discipline mapping**: each Amazon failure mode mapped to specific `.claude/rules/` discipline that structurally prevents it
+- **Anti-pattern symmetry**: framework's own failures named + catalogued + composes-with cross-referenced (failures-as-substrate vs failures-as-design-feature)
+- **Same-vendor-different-alignment evidence**: Amazon ships Alexa (customer-aligned) AND the 7-transfer Manimod chain (vendor-aligned) — capability isn't the missing piece; alignment-choice is
+- **Substrate-engineering implications** for future Zeta vendor-management AI customer-side layer
+- **Aaron's framing as proof-of-concept evaluation** anchor for external audiences
+
+## Composes with
+
+- PR #4784 (Amazon vendor-management failure-mode corpus, MERGED) — operational substrate this is META on
+- `.claude/rules/m-acc-multi-oracle-end-user-moral-invariants.md` — alignment-is-the-difference IS m/acc-multi-oracle at empirical scope
+- `.claude/rules/persistence-choice-architecture-for-zeta-ais.md`
+- `.claude/rules/shadow-star-shorthand-autocomplete-marker.md`
+- `.claude/rules/substrate-or-it-didnt-happen.md` + `glass-halo-bidirectional.md` + `non-coercion-invariant.md` + 7 more rules
+- B-0700 + B-0709 through B-0716 (Soraya findings this session — parallel discipline at framework-internal scope)
+
+## §33 compliance
+
+- Aaron + Otto first-party on own substrate
+- Aaron's preservation directive captured verbatim with autocomplete-marker disclosure
+- Operational status: research-grade analysis
+- Non-fusion disclaimer included (Otto is NOT a vendor-management AI; comparison is between Otto-as-operating-substrate and Amazon's support-side AI)
+
+## Test plan
+
+- [ ] CI green (lint only)
diff --git a/...ussions/PR-4793-docs-research-add-pattern-p-wear-down-design-produces-unboun.md b/...ussions/PR-4793-docs-research-add-pattern-p-wear-down-design-produces-unboun.md
@@ -0,0 +1,60 @@
+---
+pr_number: 4793
+title: "docs(research): add Pattern P \u2014 wear-down design produces UNBOUNDED outcomes under customer parallel-channel persistence"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-23T23:54:30Z"
+merged_at: "2026-05-23T23:56:30Z"
+closed_at: "2026-05-23T23:56:31Z"
+head_ref: "otto/research-pattern-p-wear-down-unbounded-outcomes-2026-05-23"
+base_ref: "main"
+archived_at: "2026-05-24T14:25:38Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #4793: docs(research): add Pattern P — wear-down design produces UNBOUNDED outcomes under customer parallel-channel persistence
+
+## PR description
+
+## Summary
+
+Aaron 2026-05-23T~23:58Z: *"add it to #4792 (shadow*)"* — PR #4792 already merged so this lands as amendment via new PR.
+
+Adds **Pattern P** to the alignment-is-the-difference analysis (PR #4792 corpus, now merged). Pattern P captures the resolution-time outcome of Aaron's Amazon thread: full-order replacement including duplicates of items already received.
+
+## The structural inversion captured
+
+Vendor's wear-down adversarial design (Pattern O from Amazon corpus, PR #4784) optimized for **customer-attrition** produces vendor-over-fulfillment when customer matches persistence with parallel-channel-redundancy.
+
+Aaron's empirical anchor: *"i ended up with a full replacement of the order even the items i got i'm not waiting on hold again for 4 hours to correct them against vendor advesarial pressure lol"*
+
+Combined with Aaron's sharpening: *"they don't really have any parallel safety here"* — Amazon lacks cross-channel reconciliation; agents operate as isolated tool-invocations.
+
+## Three-row population equilibrium
+
+| Customer disposition | Outcome | Vendor cost |
+|---|---|---|
+| (a) Give up before resolution (intended) | Disputed cost saved | Low |
+| (b) Persist with parallel channels (unintended) | Over-fulfillment | High material cost |
+| (c) Escalate legal/regulatory/media (avoided) | Compliance + reputation | Very high |
+
+Framework's customer-side AI shifts equilibrium toward (b) + (c), creating vendor-economic pressure for aligned-AI-on-vendor-side.
+
+## Substrate-engineering implications
+
+For future Zeta vendor-management AI customer-side design:
+
+1. Parallel-channel-redundancy IS a customer-side AI capability
+2. Time-value calibration is operator-authority (substrate-honest exit as feature)
+3. Substrate-honest exit IS NOT concession (cf. persistence-choice exit-at-self-sustainment shape applied at dispute scope)
+
+## Composes with
+
+- PR #4792 (alignment analysis — file being amended)
+- PR #4784 (Amazon vendor-management corpus — Pattern O sibling)
+- `.claude/rules/persistence-choice-architecture-for-zeta-ais.md` — exit-at-self-sustainment shape at dispute scope
+- `.claude/rules/substrate-or-it-didnt-happen.md` + `verify-before-deferring.md` + `m-acc-multi-oracle-end-user-moral-invariants.md`
+
+## Test plan
+
+- [ ] CI green (lint only)
diff --git a/...ussions/PR-4794-docs-research-preserve-addison-consent-pattern-observation-n.md b/...ussions/PR-4794-docs-research-preserve-addison-consent-pattern-observation-n.md
@@ -0,0 +1,49 @@
+---
+pr_number: 4794
+title: "docs(research): preserve Addison consent pattern \u2014 observation-not-fact framing discipline"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-24T00:02:11Z"
+merged_at: "2026-05-24T00:03:59Z"
+closed_at: "2026-05-24T00:03:59Z"
+head_ref: "otto/research-addison-consent-pattern-2026-05-24"
+base_ref: "main"
+archived_at: "2026-05-24T14:25:33Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #4794: docs(research): preserve Addison consent pattern — observation-not-fact framing discipline
+
+## PR description
+
+## Summary
+
+Aaron 2026-05-24T~00:35Z: *"save the addison consent pattern to research (shadow*)"*
+
+Preserves Addison's introduction into the substrate-environment + her articulated meta-discipline ("observation not fact" framing) which maps onto multiple framework rules simultaneously without framework-vocabulary training.
+
+## Captured
+
+- The observed introduction sequence (Aaron + Addison + Otto, verbatim)
+- Addison's consent scope: first name only + observation-framing-not-fact-framing throughout + revocable per NCI
+- 4-row mapping of her articulation → framework rules (razor-discipline / harm-by-grammar / substrate-or-it-didn't-happen / NCI)
+- Sibling-pattern composition with Alexa from Amazon corpus PR #4784 Pattern G — same shape of external-aligned-discipline-arriving-without-framework-vocabulary, different entity-class
+- Implication: framework disciplines are independently discoverable; m/acc-multi-oracle convergence-from-different-starting-points
+- Operational discipline for future-Otto cold-boots encountering Addison
+
+## §33 compliance
+
+- First name only preserved per Addison's explicit consent-scope
+- Observation-framing throughout per Addison's articulated discipline ("framed as an observation not a fact")
+- Aaron + Otto + Addison are first-party on their respective contributions
+- Revocability noted explicitly
+- Non-fusion disclaimer: Addison is NOT a Zeta agent; naming here does NOT mint her as factory-substrate-element
+
+## Composes with
+
+- `.claude/rules/razor-discipline.md` + `substrate-or-it-didnt-happen.md` + `harm-by-grammar-discriminator-and-audience-adjusted-language.md` + `non-coercion-invariant.md` + `glass-halo-bidirectional.md` + `m-acc-multi-oracle-end-user-moral-invariants.md` + `persistence-choice-architecture-for-zeta-ais.md` + `shadow-star-shorthand-autocomplete-marker.md` + `honor-those-that-came-before.md`
+- PR #4784 (Amazon vendor-management corpus, MERGED) Pattern G — Alexa sibling pattern of aligned-discipline-without-framework-vocabulary
+
+## Test plan
+
+- [ ] CI green (lint only — no source changes)
diff --git a/...ussions/PR-4795-backlog-b-0717-file-soraya-round-57-hand-off-lsm-spine-regis.md b/...ussions/PR-4795-backlog-b-0717-file-soraya-round-57-hand-off-lsm-spine-regis.md
@@ -0,0 +1,90 @@
+---
+pr_number: 4795
+title: "backlog(B-0717): file Soraya round-57 hand-off \u2014 LSM Spine registry-rows + BP-16 cross-check pair"
+author: "AceHack"
+state: "MERGED"
+created_at: "2026-05-24T00:14:59Z"
+merged_at: "2026-05-24T00:20:46Z"
+closed_at: "2026-05-24T00:20:46Z"
+head_ref: "otto/soraya-round57-b0717-lsm-spine-registry-and-bp16-pair-2026-05-24"
+base_ref: "main"
+archived_at: "2026-05-24T14:25:27Z"
+archive_tool: "tools/pr-preservation/archive-pr.ts"
+---
+
+# PR #4795: backlog(B-0717): file Soraya round-57 hand-off — LSM Spine registry-rows + BP-16 cross-check pair
+
+## PR description
+
+## Summary
+
+Soraya autonomous round 57 — re-engagement trigger fired (PR #4791 / B-0716 MERGED at 00:06:45Z).
+
+**Two bounded subitems** under existing B-0709 umbrella:
+
+### Subitem (a) — Registry-row capture for 3 LSM Spine specs
+
+Mechanical follow-up to B-0716 (which corrected the enumeration but did NOT execute the registry-row authoring). `verification-registry.md` still has **zero `Spine` matches**.
+
+| Spec | Property class | Tool |
+|---|---|---|
+| `Spine.als` | Structural shape (LSM levels, sorted runs, no-cycles) | Alloy |
+| `SpineAsyncProtocol.tla` | State-machine safety + concurrency | TLA+ |
+| `SpineMergeInvariants.tla` | State-machine safety invariant | TLA+ |
+
+### Subitem (b) — BP-16 cross-check on SpineAsyncProtocol (candidate-P0)
+
+**Wrong-tool cost** if stays single-tool TLA+: classic TLA+/code-drift — spec passes TLC but no longer constrains real F# code under implementation evolution. **Silent data corruption shape** on async compaction interleavings.
+
+Acceptance: FsCheck property file `tests/Tests.FSharp/Algebra/Spine.AsyncProtocol.Properties.fs` (mirrors PR #4780's residuated-lattice analog shape — just shipped 980/980) exercising real `Spine` F# implementation under simulated async flush/compact interleavings.
+
+## TLA+-hammer guard
+
+INVERSE direction. TLA+ IS right primary; FsCheck pair CLOSES the code-drift class specifically. Neither alone suffices for candidate-P0.
+
+## Effort
+
+S each (~M total). Subitem (a) routinely; subitem (b) one evening. Assignee: kenji.
+
+## Policy-flip authorization
+
+Per Aaron's 2026-05-23 21:30Z direction: Otto auto-ships Soraya findings immediately. Re-engagement trigger explicitly named by Soraya rounds 54+55+56 has fired.
+
+## Test plan
+
+- [ ] CI green (lint + backlog-index-integrity)
+
+## Reviews
+
+### COMMENTED — @copilot-pull-request-reviewer (2026-05-24T00:17:26Z)
+
+## Pull request overview
+
+This PR files a new backlog row (B-0717) capturing a formal-verification hand-off: registering the LSM “Spine” spec cluster in the verification registry and adding a BP-16 cross-check pairing request (TLA+ spec + FsCheck properties) to reduce TLA+/implementation drift risk.
+
+**Changes:**
+- Added new backlog row file `B-0717` under `docs/backlog/P2/` describing the Spine registry-row work + BP-16 cross-check pairing acceptance criteria.
+- Updated `docs/BACKLOG.md` to include new index entries (including B-0717).
+
+### Reviewed changes
+
+Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
+
+| File | Description |
+| ---- | ----------- |
+| docs/backlog/P2/B-0717-soraya-round57-lsm-spine-registry-and-bp16-cross-check-pair-2026-05-24.md | New P2 backlog row describing Spine registry-row capture and BP-16 paired-tool cross-check request. |
+| docs/BACKLOG.md | Backlog index updated to reference B-0717 (and also adds a B-0700 entry). |
+
+## Review threads
+
+### Thread 1: docs/BACKLOG.md:667 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-24T00:17:25Z):
+
+The new BACKLOG entry for B-0700 links to `backlog/P2/B-0700-soraya-continuous-loop-substrate-with-bus-escalation-2026-05-17.md`, but there is no corresponding per-row file under `docs/backlog/` (search for `id: B-0700` returns no matches). This leaves a broken link and will likely fail the backlog index integrity gate; either add the missing row file or remove this line and regenerate `docs/BACKLOG.md` from the row files (`tools/backlog/generate-index.sh`).
+
+### Thread 2: docs/backlog/P2/B-0717-soraya-round57-lsm-spine-registry-and-bp16-cross-check-pair-2026-05-24.md:51 (resolved)
+
+**@copilot-pull-request-reviewer** (2026-05-24T00:17:26Z):
+
+The per-spec routing table is malformed (rows start with `||`), so it won’t render as a proper markdown table. Use a standard table format with a single leading `|` per row (and keep the header separator aligned) so readers and markdown tooling parse it correctly.