Publish Codex Live quality evidence

Update Live Lab and the Live reference with the default Sol worker, full-task quality gate, Spark control, cold readiness, and production architecture.\n\nAI-assisted: OpenAI Codex.

site/data/live-codex-worker-result.ts

@@ -1,9 +1,12 @@
 export const liveCodexWorkerResult = {
-  generatedAt: '2026-07-13T01:05:15.178Z',
+  generatedAt: '2026-07-13T03:03:13.899Z',
   codexVersion: '0.144.0-alpha.4',
   transport: 'stdio JSONL',
-  model: 'GPT-5.3-Codex-Spark',
-  effort: 'low',
+  model: 'GPT-5.6-Sol',
+  effort: 'medium',
+  profile: 'quality',
+  defaultInCodex: true,
+  coldWorkerReadyMs: 5600,
   timings: {
     coldHandshakeMs: 36.83,
     coldThreadStartMs: 400.19,
@@ -11,7 +14,24 @@ export const liveCodexWorkerResult = {
     wakeToTurnStartedMs: 1.47,
     trivialTurnCompletedMs: 2060.18,
   },
-  result: 'READY',
+  protocolProbe: 'GPT-5.3-Codex-Spark · low',
+  quality: {
+    tasks: ['bolder', 'polish'],
+    runs: 4,
+    passed: 4,
+    medianGenerationMs: 34339,
+    p95GenerationMs: 42780.1,
+    averageScores: {
+      commandFidelity: 8,
+      brandAndSystemFidelity: 8.5,
+      frontendQuality: 7.5,
+      taskCompletion: 7.75,
+    },
+    sparkControlPassed: 1,
+    sparkControlRuns: 2,
+    sparkMedianGenerationMs: 16470.5,
+  },
+  result: '4/4 full-task quality gate',
   passed: true,
-  architecture: 'A dedicated persistent Live app-server connection and worker thread. Do not attach a second app-server to the desktop task.',
+  architecture: 'A dedicated persistent Live app-server connection and worker thread with native Impeccable skill, sub-command, project context, and annotation-image inputs. Do not attach a second app-server to the desktop task.',
 } as const;

site/data/live-harnesses.ts

@@ -3,8 +3,8 @@ export const harnessPaths = [
     name: 'Dedicated app-server worker',
     harness: 'Codex · isolated persistent thread',
     pickup: '1.47 ms turn wake',
-    tradeoff: 'A separate Live-owned thread can generate while the control supervisor keeps polling. Cold setup is 437 ms; a persistent connection amortizes it.',
-    status: 'Protocol probe passed',
+    tradeoff: 'A separate Live-owned Sol thread generates while the control supervisor keeps polling. Full context protects design quality; the 5.6 s cold worker startup happens during Live initialization.',
+    status: 'Codex default · gated',
     statusTone: 'ready',
     source: 'https://learn.chatgpt.com/docs/app-server',
   },
@@ -136,11 +136,11 @@ export const liveExperiments = [
   },
   {
     rank: 13,
-    title: 'Wake a dedicated Codex worker — protocol proven',
-    evidence: 'A cold app-server handshake took 36.83 ms, thread creation 400.19 ms, and turn/start reached turn/started in 1.47 ms. A low-effort Spark probe completed in 2.06 s.',
-    move: 'Keep one Live-owned app-server connection and isolated worker thread warm. Feed it durable events and structured output; never resume the desktop task from a second server.',
-    expected: 'Removes terminal-output pickup and main-turn availability from generation startup while preserving transactional publication fencing.',
-    confidence: 'Official protocol + live probe',
+    title: 'Wake a dedicated Codex worker — shipped',
+    evidence: 'The production Sol/medium path passed 4/4 full bolder + polish tasks after fidelity tuning (34.3 s median, 42.8 s p95). The Spark control passed only 1/2 despite a 16.5 s median. Turn wake itself is 1.47 ms.',
+    move: 'Keep one Live-owned app-server connection and isolated quality-model thread warm. Attach the real skill, sub-command, resolved project context, source neighborhood, and annotation image; never resume the desktop task from a second server.',
+    expected: 'Removes main-turn availability from generation and allows independent progressive delivery without trading away brand fidelity.',
+    confidence: 'Full-task model gate + protocol smoke',
   },
 ];
 

site/pages/live-lab/index.astro

@@ -404,9 +404,9 @@ const providerColumns = [
         <p>{currentHarnessProbe.surface}: {currentHarnessProbe.result}</p>
       </div>
       <div class="live-performance-finding live-finding-card--quiet" role="note">
-        <strong>Dedicated Codex worker probe</strong>
+        <strong>Dedicated Codex worker</strong>
         <p>
-          {liveCodexWorkerResult.model} at {liveCodexWorkerResult.effort} effort started a dedicated turn in {displayMs(liveCodexWorkerResult.timings.wakeToTurnStartedMs)} and completed a trivial structured response in {displayMs(liveCodexWorkerResult.timings.trivialTurnCompletedMs)}. Cold app-server plus thread setup cost {displayMs(liveCodexWorkerResult.timings.coldHandshakeMs + liveCodexWorkerResult.timings.coldThreadStartMs)} and can be amortized. The safe path owns a separate Live thread; it never resumes the desktop task from a competing app-server.
+          Codex now defaults to {liveCodexWorkerResult.model} at {liveCodexWorkerResult.effort} effort in a separate Live-owned thread. It passed {liveCodexWorkerResult.quality.passed}/{liveCodexWorkerResult.quality.runs} full <code>/bolder</code> and <code>/polish</code> tasks at a {displayMs(liveCodexWorkerResult.quality.medianGenerationMs)} median; the Spark control passed {liveCodexWorkerResult.quality.sparkControlPassed}/{liveCodexWorkerResult.quality.sparkControlRuns}. Protocol wake is {displayMs(liveCodexWorkerResult.timings.wakeToTurnStartedMs)}; the measured cold worker reached ready in {displayMs(liveCodexWorkerResult.coldWorkerReadyMs)} during Live initialization.
         </p>
       </div>
     </section>
@@ -470,9 +470,10 @@ const providerColumns = [
               <header><h2>Dedicated Codex worker</h2></header>
               <dl>
                 <div><dt>Wake</dt><dd>{displayMs(liveCodexWorkerResult.timings.wakeToTurnStartedMs)}</dd></div>
-                <div><dt>Cold setup</dt><dd>{displayMs(liveCodexWorkerResult.timings.coldHandshakeMs + liveCodexWorkerResult.timings.coldThreadStartMs)}</dd></div>
+                <div><dt>Full-task median</dt><dd>{displayMs(liveCodexWorkerResult.quality.medianGenerationMs)}</dd></div>
+                <div><dt>Quality gate</dt><dd>{liveCodexWorkerResult.quality.passed}/{liveCodexWorkerResult.quality.runs}</dd></div>
               </dl>
-              <p>Live owns the worker thread. It never resumes the desktop task.</p>
+              <p>Default in Codex; other harnesses keep the foreground path.</p>
             </section>
           </aside>
         </div>

skill/reference/live.md

@@ -99,9 +99,9 @@ node {{scripts_path}}/live-complete.mjs --id SESSION_ID
 
 Server restart rule: start `live-server.mjs` again, then poll. Startup requeues unacknowledged pending events from the journal, so do not ask the user to click Go again unless `live-resume.mjs` says no active session exists.
 
-### Experimental dedicated Codex worker
+### Dedicated Codex worker
 
-Codex can opt into a Live-owned persistent app-server supervisor instead of using the desktop task as the poll supervisor:
+Codex uses a Live-owned persistent app-server supervisor instead of using the desktop task as the poll supervisor:
 
 ```bash
 node {{scripts_path}}/live.mjs

skill/scripts/live/codex-worker-supervisor.mjs

@@ -148,7 +148,7 @@ export class CodexLiveWorkerSupervisor {
           id: event.id,
           type: 'error',
           sourceEventType: event.type,
-          message: `Experimental Codex worker does not handle ${event.type}; disable IMPECCABLE_LIVE_CODEX_WORKER for the portable foreground path.`,
+          message: `Dedicated Codex worker does not handle ${event.type}; disable IMPECCABLE_LIVE_CODEX_WORKER for the portable foreground path.`,
         });
       }
     }