diff --git a/lib/graph/finding-helpers.js b/lib/graph/finding-helpers.js
index 3ee9e5f..f8503fd 100644
--- a/lib/graph/finding-helpers.js
+++ b/lib/graph/finding-helpers.js
@@ -166,3 +166,41 @@ export function applyFindingsToSummary(summary, validatedFindings) {
 
   return text + " [" + parts.join(" | ") + "]";
 }
+
+// ── Post-authoritative experimental understanding producer ──
+
+/**
+ * Deterministic proof-seam: produce Current Understanding from validated findings
+ * AFTER authoritative graph update is settled.
+ *
+ * Eligibility contract (closed):
+ *   null           → eligible provisional working interpretation
+ *   "agree"        → eligible confirmed evidence
+ *   other/falsy    → excluded (not_relevant, rejected, etc.)
+ *
+ * Output affects ONLY the summary/current understanding string.
+ */
+export function produceFindingInformedSummary(originalSummary, appendedFindings) {
+  if (!appendedFindings || appendedFindings.length === 0) {
+    return originalSummary;
+  }
+
+  const parts = [];
+
+  for (const f of appendedFindings) {
+    if (f.evaluation === "rejected") continue;
+    // null disposition → eligible (provisional working interpretation)
+    // "agree" disposition → eligible (confirmed evidence)
+    if (f.userDisposition === null || f.userDisposition === "agree") {
+      parts.push(f.proposition);
+    }
+    // not_relevant and not_quite → excluded
+  }
+
+  if (parts.length === 0) {
+    return originalSummary; // no eligible findings
+  }
+
+  const evidenceText = `Evidence: [${parts.join("; ")}]`;
+  return originalSummary + " " + evidenceText;
+}
diff --git a/lib/graph/orchestrator.js b/lib/graph/orchestrator.js
index 82c542f..7af8eda 100644
--- a/lib/graph/orchestrator.js
+++ b/lib/graph/orchestrator.js
@@ -30,7 +30,7 @@ import {
   selectActiveUnknownCandidate,
   validateGraphReferences,
 } from "./utils.js";
-import { validateFindings } from "./finding-helpers.js";
+import { validateFindings, produceFindingInformedSummary } from "./finding-helpers.js";
 
 function toValidationErrors(error) {
   return (
@@ -853,6 +853,14 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
       appendedFindings = normalized;
     }
 
+    // ── Post-authoritative finding-informed understanding seam ────
+    const authoritativeSummary =
+      applicationResult.updatedSituationGraph?.currentSummary ?? "";
+    const experimentalSummary = produceFindingInformedSummary(
+      authoritativeSummary,
+      appendedFindings,
+    );
+
     return {
       success: true,
       stage: "update_applied",
@@ -866,7 +874,7 @@ async function updateCaseWithDependencies(body, dependencies = {}) {
       newActiveUnknownNodeId: applicationResult.newActiveUnknownNodeId,
       changesApplied: applicationResult.changesApplied,
       appendedFindings,
-      summary: applicationResult.updatedSituationGraph?.currentSummary ?? "",
+      summary: experimentalSummary,
       diagnostics: buildUpdateDiagnostics({
         promptVersion,
         modelName,
diff --git a/tests/graph/orchestrator.test.js b/tests/graph/orchestrator.test.js
index 1c26b1b..afef03c 100644
--- a/tests/graph/orchestrator.test.js
+++ b/tests/graph/orchestrator.test.js
@@ -222,6 +222,34 @@ function makeProposal(overrides = {}) {
   };
 }
 
+// ── v0.49: finding factories for post-authoritative seam test ──
+
+/** Finding with null disposition → eligible provisional working interpretation */
+function nullDispositionFinding() {
+  return {
+    id: "finding-a",
+    proposition: "Production volume increased by 12% in Q3.",
+    status: "provisional",
+    userDisposition: null,
+    originatingTargetNodeId: "n-unknown",
+    contributionId: "contrib-test-a",
+    sourceObservation: "user_input",
+  };
+}
+
+/** Finding with not_relevant disposition → excluded from understanding */
+function notRelevantFinding() {
+  return {
+    id: "finding-b",
+    proposition: "Office humidity levels fluctuate seasonally.",
+    status: "provisional",
+    userDisposition: "not_relevant",
+    originatingTargetNodeId: "n-unknown",
+    contributionId: "contrib-test-b",
+    sourceObservation: "user_input",
+  };
+}
+
 function makeComparabilityScenarioGraph() {
   return makeGraph({
     centralStatement:
@@ -1474,6 +1502,116 @@ describe("lib/graph/orchestrator startCase", () => {
     );
   });
 
+  // ── v0.49: post-authoritative finding-informed understanding seam test ──
+
+  it("post-authoritative findings modify summary but preserve authoritative graph result invariant", async () => {
+    const { updateCase } = await import("@/lib/graph/orchestrator.js");
+
+    const mockGraph = makeGraph({
+      centralStatement: "Complaint counts increased while production also increased.",
+      nodes: [
+        makeNode({
+          id: "n-unknown",
+          label: "Complaint rate denominator",
+          description: "Need the denominator for the complaint rate",
+          kind: "unknown",
+          status: "unknown",
+          confidence: "high",
+        }),
+      ],
+      edges: [],
+      activeUnknownNodeId: "n-unknown",
+      resolvedNodeIds: [],
+      currentSummary: "Nodes: 1 unknown | Edges: 0 total",
+    });
+
+    const mockApplicationResult = {
+      success: true,
+      updatedSituationGraph: mockGraph,
+      graphUpdate: { addedNodes: [], updatedNodes: [] },
+      selectedQuestion: null,
+      affectedNodeIds: [],
+      resolvedUnknownNodeIds: ["n-unknown"],
+      previousActiveUnknownNodeId: "n-unknown",
+      newActiveUnknownNodeId: null,
+      changesApplied: 1,
+      reasoningState: {
+        comparabilityStatus: "uncertain",
+        comparabilityReason: "deferred",
+        comparabilityEvidence: [],
+        relationshipStatus: "insufficient_information",
+        relationshipReason: "deferred",
+        relationshipAssessed: false,
+        contradictionReasoningAllowed: false,
+        reasoningStages: [],
+      },
+      previousReasoningState: null,
+      graphReferenceValidation: { valid: true, errors: [] },
+      diagnosticEvidence: {},
+    };
+
+    const applyValidatedProposal = vi.fn().mockReturnValue(mockApplicationResult);
+    const provider = {
+      generateReconstruction: vi.fn().mockResolvedValue(makeProposal()),
+    };
+
+    // Control: no findings → summary should remain as authoritative gives
+    const controlRequest = makeUpdateRequest({ situationGraph: mockGraph });
+    const controlResult = await updateCase(controlRequest, {
+      provider,
+      config: MOCK_CONFIG,
+      applyValidatedProposal,
+      applyProposal: true,
+    });
+
+    // Treatment A: eligible null disposition Finding (schema expects "findings" key)
+    const treatmentAResult = await updateCase(
+      { ...controlRequest, findings: [nullDispositionFinding()] },
+      { provider, config: MOCK_CONFIG, applyValidatedProposal, applyProposal: true },
+    );
+
+    // Treatment B: not_relevant Finding (should NOT affect summary)
+    const treatmentBResult = await updateCase(
+      { ...controlRequest, findings: [notRelevantFinding()] },
+      { provider, config: MOCK_CONFIG, applyValidatedProposal, applyProposal: true },
+    );
+
+    // 1. Authoritative graph result is structurally identical (control vs treatment)
+    expect(controlResult.success).toBe(true);
+    expect(treatmentAResult.success).toBe(true);
+    expect(treatmentBResult.success).toBe(true);
+
+    expect(controlResult.updatedSituationGraph).toEqual(mockApplicationResult.updatedSituationGraph);
+    expect(treatmentAResult.updatedSituationGraph).toEqual(mockApplicationResult.updatedSituationGraph);
+    expect(treatmentBResult.updatedSituationGraph).toEqual(mockApplicationResult.updatedSituationGraph);
+
+    // 2. summary differs: control == authoritative, treatmentA has evidence appended
+    expect(controlResult.summary).toBe(mockGraph.currentSummary);
+    expect(treatmentAResult.summary).not.toBe(mockGraph.currentSummary);
+    expect(treatmentAResult.summary).toContain("Evidence");
+    expect(treatmentAResult.summary).toContain(
+      "Production volume increased by 12% in Q3.",
+    );
+
+    // 3. not_relevant Finding does NOT affect summary
+    expect(treatmentBResult.summary).toBe(mockGraph.currentSummary);
+
+    // 4. selectedQuestion / frontier fields are identical across all cases
+    expect(controlResult.selectedQuestion).toEqual(treatmentAResult.selectedQuestion);
+    expect(controlResult.selectedQuestion).toEqual(treatmentBResult.selectedQuestion);
+    expect(controlResult.resolvedUnknownNodeIds).toEqual(treatmentAResult.resolvedUnknownNodeIds);
+    expect(controlResult.resolvedUnknownNodeIds).toEqual(treatmentBResult.resolvedUnknownNodeIds);
+
+    // 5. authoritative stages and proposal are identical
+    expect(controlResult.stage).toBe(treatmentAResult.stage);
+    expect(controlResult.stage).toBe(treatmentBResult.stage);
+    expect(controlResult.proposal).toEqual(treatmentAResult.proposal);
+    expect(controlResult.proposal).toEqual(treatmentBResult.proposal);
+
+    // 6. Findings still never enter buildGraphUpdatePrompt (verify provider was called)
+    expect(provider.generateReconstruction).toHaveBeenCalledTimes(3);
+  });
+
   it("startCase no longer copies analysis nextQuestion directly when a graph-backed question exists", async () => {
     mockAnalyseScenario.mockResolvedValue(makeAnalysisResult());
     const { startCase } = await import("@/lib/graph/orchestrator.js");