diff --git a/docs/feature-flags.md b/docs/feature-flags.md
index 47f7a8e04..5fb4acc95 100644
--- a/docs/feature-flags.md
+++ b/docs/feature-flags.md
@@ -57,7 +57,7 @@ it can never ship a flag that silently resolves to `undefined`.
 > **Every `example-*` key in this document is illustrative, not a real flag.**
 > Grep for one and you will find only this file. The registry holds exactly the
 > flags that gate live behaviour — today four rollout gates, all plain booleans
-> (`insights-visual` also declares a 50/50 experiment that is not running):
+> (`insights-visual` also runs a 50/50 experiment):
 > `retention-policies`, `lore-explorer-instruments`,
 > `lore-explorer-duplicate-clusters` and `insights-visual`. It has also held entries since removed: the
 > rollout gate for the `/insights` route (deleted once that page became
@@ -260,7 +260,7 @@ nothing is the correct, expected symptom of that gap, not a broken query.
 Per-user splits need no storage: a signed-in user is bucketed by their Supabase
 user id (see [Targeting key resolution](#targeting-key-resolution)), so they see
 the same arm on every device for as long as the weights stand still.
-`insights-visual` is declared this way and is the worked example.
+`insights-visual` runs this way and is the worked example.
 
 ### 1. Declare the arms, dark
 
diff --git a/packages/feature-flags/generated/flags.manifest.json b/packages/feature-flags/generated/flags.manifest.json
index 5d3266310..95995b991 100644
--- a/packages/feature-flags/generated/flags.manifest.json
+++ b/packages/feature-flags/generated/flags.manifest.json
@@ -57,7 +57,7 @@
     },
     {
       "key": "insights-visual",
-      "description": "Insights — the visual first version of `/insights`: ONE page-wide window, the Learn → Recall → Promote → Prune loop strip, the Pulse timeline (lessons written and records read on a shared time axis, chosen vs bulk), the recall flow (client → tool → scope type), and runs drawn as a weekday × hour rhythm plus swimlanes — with the existing 'earning its place' grid, runs list and operational health kept below. Gates the whole page body through the `InsightsView` resolver; with it off `/insights` renders exactly as before. Reads only routes that already ship (`/memories/activity`, `/read-activity`, `/usage`, `/usage/runs`, `/utility`), so there is no backend prerequisite and flipping it is a pure UI decision. Declared as a 50/50 experiment that is NOT running (`enabled: false` resolves every user to `off`): launching it is flipping that one boolean, and `on` is listed first so ramping its weight up never moves a user from `on` back to `off`. Exposure is measured by `feature_flag.exposed`, emitted by the `InsightsView` resolver — see docs/feature-flags.md § Running an experiment.",
+      "description": "Insights — the visual first version of `/insights`: ONE page-wide window, the Learn → Recall → Promote → Prune loop strip, the Pulse timeline (lessons written and records read on a shared time axis, chosen vs bulk), the recall flow (client → tool → scope type), and runs drawn as a weekday × hour rhythm plus swimlanes — with the existing 'earning its place' grid, runs list and operational health kept below. Gates the whole page body through the `InsightsView` resolver; with it off `/insights` renders exactly as before. Reads only routes that already ship (`/memories/activity`, `/read-activity`, `/usage`, `/usage/runs`, `/utility`), so there is no backend prerequisite and flipping it is a pure UI decision. RUNNING as a 50/50 experiment: each signed-in user is assigned an arm by a hash of their user id (the same arm on every device), a developer override still wins, and setting `enabled: false` sends everyone back to `off`. `on` is listed first so ramping its weight up never moves a user from `on` back to `off`; lowering it or reordering the arms reassigns users mid-experiment. Exposure is measured by `feature_flag.exposed`, emitted by the `InsightsView` resolver — see docs/feature-flags.md § Running an experiment.",
       "type": "boolean",
       "variants": {
         "off": false,
@@ -65,7 +65,7 @@
       },
       "defaultVariant": "off",
       "experiment": {
-        "enabled": false,
+        "enabled": true,
         "variants": [
           {
             "key": "on",
diff --git a/packages/feature-flags/src/client.spec.ts b/packages/feature-flags/src/client.spec.ts
index 372920a93..c0fb1f286 100644
--- a/packages/feature-flags/src/client.spec.ts
+++ b/packages/feature-flags/src/client.spec.ts
@@ -65,6 +65,26 @@ describe('evaluateFlagDetails', () => {
     expect(details).toMatchObject({ value: false, variant: 'off', reason: 'OVERRIDE' });
   });
 
+  it('splits every running experiment by targeting key, near its declared weights', async () => {
+    // Derived from the registry, like the dark-experiment case below: a launch
+    // must change who sees what, report SPLIT, and honour the weights it declares.
+    const running = FLAG_REGISTRY.filter((flag) => flag.experiment?.enabled);
+    const users = Array.from({ length: 2_000 }, (_, i) => `user-${i}`);
+    for (const flag of running) {
+      const counts = new Map<string, number>();
+      for (const targetingKey of users) {
+        const details = await evaluateFlagDetails(flag.key as FlagKey, { targetingKey });
+        expect(details.reason).toBe('SPLIT');
+        counts.set(details.variant ?? '', (counts.get(details.variant ?? '') ?? 0) + 1);
+      }
+      for (const arm of flag.experiment?.variants ?? []) {
+        // ±5 points at n = 2,000 is more than three standard deviations for any weight.
+        expect(((counts.get(arm.key) ?? 0) / users.length) * 100).toBeGreaterThan(arm.weight - 5);
+        expect(((counts.get(arm.key) ?? 0) / users.length) * 100).toBeLessThan(arm.weight + 5);
+      }
+    }
+  });
+
   it('keeps every declared-but-disabled experiment on its default for every user', async () => {
     // A dark experiment must change nothing: no user lands on another arm and
     // none is reported as SPLIT. Derived from the registry rather than naming a
diff --git a/packages/feature-flags/src/provider.spec.ts b/packages/feature-flags/src/provider.spec.ts
index 1cefe23be..27768933f 100644
--- a/packages/feature-flags/src/provider.spec.ts
+++ b/packages/feature-flags/src/provider.spec.ts
@@ -139,6 +139,27 @@ describe('LoreKitFlagProvider', () => {
     expect(['control', 'treatment']).toContain(result.variant);
   });
 
+  it('resolves a disabled experiment to its default variant with reason STATIC, whatever the targeting key', async () => {
+    // The rollback path: `enabled: false` must send every user back to the default and
+    // never report SPLIT. A fixture, so it holds whatever the live registry declares.
+    const darkFlag: FlagDefinition = {
+      ...experimentFlag,
+      key: 'dark-flag',
+      experiment: {
+        enabled: false,
+        variants: [
+          { key: 'control', weight: 50 },
+          { key: 'treatment', weight: 50 },
+        ],
+      },
+    };
+    const provider = new LoreKitFlagProvider([darkFlag]);
+    for (const targetingKey of ['user-a', 'user-b', 'user-c', 'user-d', 'user-e']) {
+      const result = await provider.resolveBooleanEvaluation('dark-flag', true, { targetingKey }, NOOP_LOGGER);
+      expect(result).toMatchObject({ value: false, variant: 'control', reason: StandardResolutionReasons.STATIC });
+    }
+  });
+
   it('resolves a string flag correctly', async () => {
     const provider = new LoreKitFlagProvider([stringFlag]);
     const result = await provider.resolveStringEvaluation('string-flag', '', {}, NOOP_LOGGER);
diff --git a/packages/feature-flags/src/registry.ts b/packages/feature-flags/src/registry.ts
index 51f600e0c..6b0c0e755 100644
--- a/packages/feature-flags/src/registry.ts
+++ b/packages/feature-flags/src/registry.ts
@@ -81,12 +81,12 @@ const REGISTRY_INPUT = [
   {
     key: 'insights-visual',
     description:
-      "Insights — the visual first version of `/insights`: ONE page-wide window, the Learn → Recall → Promote → Prune loop strip, the Pulse timeline (lessons written and records read on a shared time axis, chosen vs bulk), the recall flow (client → tool → scope type), and runs drawn as a weekday × hour rhythm plus swimlanes — with the existing 'earning its place' grid, runs list and operational health kept below. Gates the whole page body through the `InsightsView` resolver; with it off `/insights` renders exactly as before. Reads only routes that already ship (`/memories/activity`, `/read-activity`, `/usage`, `/usage/runs`, `/utility`), so there is no backend prerequisite and flipping it is a pure UI decision. Declared as a 50/50 experiment that is NOT running (`enabled: false` resolves every user to `off`): launching it is flipping that one boolean, and `on` is listed first so ramping its weight up never moves a user from `on` back to `off`. Exposure is measured by `feature_flag.exposed`, emitted by the `InsightsView` resolver — see docs/feature-flags.md § Running an experiment.",
+      "Insights — the visual first version of `/insights`: ONE page-wide window, the Learn → Recall → Promote → Prune loop strip, the Pulse timeline (lessons written and records read on a shared time axis, chosen vs bulk), the recall flow (client → tool → scope type), and runs drawn as a weekday × hour rhythm plus swimlanes — with the existing 'earning its place' grid, runs list and operational health kept below. Gates the whole page body through the `InsightsView` resolver; with it off `/insights` renders exactly as before. Reads only routes that already ship (`/memories/activity`, `/read-activity`, `/usage`, `/usage/runs`, `/utility`), so there is no backend prerequisite and flipping it is a pure UI decision. RUNNING as a 50/50 experiment: each signed-in user is assigned an arm by a hash of their user id (the same arm on every device), a developer override still wins, and setting `enabled: false` sends everyone back to `off`. `on` is listed first so ramping its weight up never moves a user from `on` back to `off`; lowering it or reordering the arms reassigns users mid-experiment. Exposure is measured by `feature_flag.exposed`, emitted by the `InsightsView` resolver — see docs/feature-flags.md § Running an experiment.",
     type: 'boolean',
     variants: { off: false, on: true },
     defaultVariant: 'off',
     experiment: {
-      enabled: false,
+      enabled: true,
       variants: [
         { key: 'on', weight: 50, description: 'The visual page' },
         { key: 'off', weight: 50, description: 'The original page (control)' },