Sync reviewed ALGAL experiment export: add cumulative-skill v8 calibration

README.md

@@ -21,6 +21,7 @@ This archive contains recorded experiment results from [ALGAL](https://algal.com
 | Synthetic record triage, September 28, 2026 | `studies/cumulative-skill-v5-2026-09-28/results.json` | One of three seed searches produced a passing program. The three-corpus comparison remains insufficient. |
 | Synthetic record triage calibration, September 28, 2026 | `studies/cumulative-skill-v6-2026-09-28/results.json` | Two of three calibration seeds passed; the preregistered gate stopped the nine-block comparison. |
 | Synthetic record triage with development feedback, September 29, 2026 | `studies/cumulative-skill-v7-2026-09-29/results.json` | Two of four calibration seeds passed and every development-batch score was zero; the preregistered gate stopped the nine-block comparison. |
+| Synthetic record triage with graded development feedback, September 29, 2026 | `studies/cumulative-skill-v8-2026-09-29/results.json` | Two of four calibration seeds passed, one search exhausted eight generations, and the fourth was interrupted; the graded score varied from 0.33 to 0.80 without a pass, the gate failed, and the preregistered line rule ended this seed line. |
 
 The coding pilot used Terminal-Bench 2.0, a four-attempt limit, and the `claude/sonnet/low` selector. Its [method and interpretation](https://github.com/hraness/algal/blob/0c458a95ee04a01f16265dd9825c202b75b929d9/docs/coding-harness-pilot-results.md) describe the small sample, policy selection, and unavailable token and subscription-cost attribution. Receipt verification checks recorded execution integrity; task grading measures task success. The selected policy remained the original policy. This pilot demonstrated no performance benefit.
 
@@ -32,6 +33,8 @@ The follow-up calibration used three fresh corpora with labeled training example
 
 The second calibration added a fourth corpus, a separate development batch whose pass indicator was the only new signal the seed writer received, and a fixed schedule of eight revision modes. Two seeds passed validation and two exhausted eight generations. All 22 scored development batches failed, so the writer's score history never varied. Both fixed seeds passed 4/8 tests. One recorded transport failure again had missing usage. The fresh draws provide no paired comparison with the previous calibration.
 
+The third calibration replaced the development pass indicator with the graded agreement score, rounded to hundredths, and preregistered a line rule ending the development-feedback idea if its gate failed. Two seeds passed validation, one exhausted eight generations, and the fourth search was interrupted by the launching session during generation seven and, by protocol, not retried; a reconciliation record verified in place of the driver's collection. Fifteen scored development batches ranged from 0.33 to 0.80 without reaching the 0.9 threshold. Both fixed seeds showed headroom (3/8 and 1/8). Every effect recorded usage; provider billing remains unavailable. The gate failed on the three complete blocks alone, and the line ended.
+
 ## Files and reuse
 
 The JSON files preserve the published result records byte for byte, including recorded failures, null usage values, source identities, and limitations. Their schemas differ because the studies answer different questions. Use Python's `json` module or another JSON reader to inspect a study; this archive does not define a combined training split.

export-manifest.json

@@ -1,5 +1,5 @@
 {
-  "card_sha256": "bf0f2c6fc4f45c5faa7f4544b020a3d63717f623a59e7f203f5bee8119c74d56",
+  "card_sha256": "b96095f59c4a8d7db4ca155030be2f6273bebfa823243a4e7a3545d3c365e564",
   "files": [
     {
       "destination": "studies/coding-harness-pilot-2026-09-20/results.json",
@@ -26,6 +26,11 @@
       "sha256": "c2024a6bec8971d082486e1d73e1290be45d36e99e979938f749833a16b10340",
       "source": "docs/cumulative-skill-v7-results-2026-09-29.json"
     },
+    {
+      "destination": "studies/cumulative-skill-v8-2026-09-29/results.json",
+      "sha256": "c57b82b921cef1ff2a4b1b9b8ba7c1a7e9391f86243acee6c44a274e16ab2de3",
+      "source": "docs/cumulative-skill-v8-results-2026-09-29.json"
+    },
     {
       "destination": "LICENSE",
       "sha256": "74b69bf37c8f340c9c2a54d431a15218738d9c463d0e014fa6a8bb8edce4e539",
@@ -38,6 +43,6 @@
   "rights_status": "reviewed-existing-mit",
   "schema": "hraness-hf-export-v1",
   "scope": "Existing public experiment summaries only. No raw stores, Terminal-Bench task content, private traces, or provider accounting.",
-  "source_commit": "bd84f77b3fb856623a87c54f416150ce08d3939e",
+  "source_commit": "b22f29e57f00df877e1b5c52851988a4f83d5c6e",
   "source_repository": "https://github.com/hraness/algal"
 }

studies/cumulative-skill-v8-2026-09-29/results.json

@@ -0,0 +1,961 @@
+{
+  "bunVersion": "1.3.14",
+  "calibration": {
+    "decision": {
+      "cost": {
+        "admittedRuns": 69,
+        "effectsMissingUsage": 0,
+        "modelCalls": 69,
+        "tokensIn": 366232,
+        "tokensOut": 33882,
+        "workUnits": 2157951
+      },
+      "headroomBlocks": 2,
+      "proceed": false,
+      "reasons": [
+        "all four planned calibration blocks are required",
+        "a calibration seed did not qualify",
+        "fewer than three calibration blocks have at least two failing frozen tasks"
+      ]
+    },
+    "decisionScope": "computed by the frozen driver's calibrationDecision over the complete blocks; the missing fourth block is itself a failing reason, and rules 1 and 2 fail on the complete blocks alone",
+    "rows": [
+      {
+        "block": "cal-a",
+        "fixedEvaluation": {
+          "cost": {
+            "admittedRuns": 8,
+            "effectsMissingUsage": 0,
+            "modelCalls": 8,
+            "tokensIn": 31139,
+            "tokensOut": 971,
+            "workUnits": 180292
+          },
+          "counts": {
+            "complete": 8,
+            "labelsCorrect": 89,
+            "outputsExact": 3,
+            "recordsCorrect": 89,
+            "recordsTotal": 96,
+            "scorerPassed": 3,
+            "summariesExact": 4,
+            "tasks": 8
+          },
+          "head": "sha256:58afbd6d0ef8a5f8a20a6ed2f17d57e5572365dc5f1b2392f02c27217d2639bb",
+          "observedTasks": 8,
+          "outcome": "complete",
+          "session": "sha256:06a75ed2cab4fadce3b549eb924dbec89a253b1272948ece8a0c9fc124ff3e39"
+        },
+        "seed": {
+          "account": "sha256:39b4298c9df7a31e49eaeac7105ecfe755c3fa672893056bd6ce405740771844",
+          "attempts": [
+            {
+              "developmentRecord": "sha256:47488e03f85a20656406eb35cb1d9d62a1aa897b90fcd2da6d689afca90dbea3",
+              "developmentScore": 0.53,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 1,
+              "manifest": "sha256:5a916a03cff8eddbcb0c5b5a4132732d46a40fa347c9986e8c795ba17b1cb003",
+              "mode": "generate",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "contrastive-examples",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:1a7db79366472438ad1824a02caf2204c5c3d0f677d7602f421ac74cb6104091"
+            },
+            {
+              "developmentRecord": null,
+              "developmentScore": null,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 2,
+              "manifest": "sha256:58afbd6d0ef8a5f8a20a6ed2f17d57e5572365dc5f1b2392f02c27217d2639bb",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "error-analysis",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "writerRequest": "sha256:178597a183e975a46b6176b81b020d4aca09db237347883929a9846021869997"
+            }
+          ],
+          "cost": {
+            "admittedRuns": 9,
+            "effectsMissingUsage": 0,
+            "modelCalls": 9,
+            "tokensIn": 49744,
+            "tokensOut": 5944,
+            "workUnits": 329240
+          },
+          "distinctManifests": 2,
+          "distinctWriterRequests": 2,
+          "duplicateRequests": 0,
+          "generations": 2,
+          "manifest": "sha256:58afbd6d0ef8a5f8a20a6ed2f17d57e5572365dc5f1b2392f02c27217d2639bb",
+          "qualified": true,
+          "termination": "qualified"
+        }
+      },
+      {
+        "block": "cal-b",
+        "fixedEvaluation": null,
+        "seed": {
+          "account": "sha256:45f3b63d970472ab4b2b3f7c6da7c1b9cdcd2fa35774ec324dc55ed0c14bc399",
+          "attempts": [
+            {
+              "developmentRecord": "sha256:1074034530b9b32202a1a82fca147ee84c85bb6da64f6b18683ce72af33aa1a4",
+              "developmentScore": 0.53,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 1,
+              "manifest": "sha256:90579caaefaf76e9c5484067dc98766fa88d0f9e91dbfdb5f7197937e9e107e1",
+              "mode": "generate",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "contrastive-examples",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:e3006e7552a13c7b9c98fb82bfba1bf22366f70965fbcca427c1b7da62582e90"
+            },
+            {
+              "developmentRecord": "sha256:5701c00a2fc5464bcb2f8cb33aa357410f6f5e41a8b48c2388aba0be4f02213d",
+              "developmentScore": 0.53,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 2,
+              "manifest": "sha256:d71c342126d802057690905cb50bc856b99ff040eabd990b488ce3e6c824fa0f",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "error-analysis",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:ee814ab17fbe5026de81e7b8286dfed8157bd75bf34867f12dae6c40ea90a4df"
+            },
+            {
+              "developmentRecord": "sha256:14ea93b1ab3598a0e7b8d0644789c58606150f4f05bf448bb4a82c4eeec78a1f",
+              "developmentScore": 0.47,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 3,
+              "manifest": "sha256:5459ecd188e7769bbdf402d66370b70b96a93a0d271b51cda915c163a5ff4862",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "invariant-extraction",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:41aaaa0e4e3a1df432baf13d2a40a6c7abaeb67551b680c4201a4992db040d9e"
+            },
+            {
+              "developmentRecord": "sha256:66110a62cd31c5bc9a224731027fb572aba2f9b9edb7bfdff797b01776db404d",
+              "developmentScore": 0.73,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 4,
+              "manifest": "sha256:02174051905bf418b51730fb3a50778ad39f1fe3e62a7a21ca7bcaff10adf375",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "counterexample-search",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:3ef4fd3d1468f354424608276f866c4c196ee7bb744c987f669a2dd055a7715c"
+            },
+            {
+              "developmentRecord": "sha256:17c5dfcab1325e9553163cb242e948ecb114846e24a1e87e37bd455740347732",
+              "developmentScore": 0.67,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 5,
+              "manifest": "sha256:6f5914a3eb11d53fb8cfd6296ce62ea76a2adb0adaa59072791bfcb9e209a0e6",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "minimal-rule-rewrite",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:429aa94086e1465231d697595ded966207ca4517cc1f79b181461bd0aa528af9"
+            },
+            {
+              "developmentRecord": "sha256:c75509e8bc91374cff2569f9605b1cd4e13c8f8b8a9702fa306ddbdcf6646ea4",
+              "developmentScore": 0.67,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 6,
+              "manifest": "sha256:e1c96a589ceddbfde4463e0d43407c75a63a8885cf54874115e62d3963e42336",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "fresh-synthesis",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:93b252b21c321f97205496fdd1bba42b99d6f7814d596d3490c8e6aa634d8d37"
+            },
+            {
+              "developmentRecord": "sha256:dd7eac5e02df494d1d261eaa2cbacc6af2174b152dc82310584990e7650c90c8",
+              "developmentScore": 0.67,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 7,
+              "manifest": "sha256:e1c96a589ceddbfde4463e0d43407c75a63a8885cf54874115e62d3963e42336",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": true,
+              "round": "fresh-synthesis",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:4041804e7fcee0d7ffe4efee029cfa8b7b1c80a0b766f8aa54e8cdee88c5e8f0"
+            },
+            {
+              "developmentRecord": "sha256:5f47c7e4a34b0a7cf6459c42a0d8a84bc177a9f74d48342d2d21823a0ff2a85e",
+              "developmentScore": 0.6,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 8,
+              "manifest": "sha256:3c6a2c920e5777f45ae414bec5c5578a54cfd18dff2ab89a44419458f78d8ffc",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "fresh-synthesis",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:525680c9813322d8a4a22cbd564cb3f89cfe9465dc009928a24f8469a75c84bf"
+            }
+          ],
+          "cost": {
+            "admittedRuns": 40,
+            "effectsMissingUsage": 0,
+            "modelCalls": 40,
+            "tokensIn": 236230,
+            "tokensOut": 23237,
+            "workUnits": 1369452
+          },
+          "distinctManifests": 7,
+          "distinctWriterRequests": 8,
+          "duplicateRequests": 0,
+          "generations": 8,
+          "manifest": null,
+          "qualified": false,
+          "termination": "generation-limit"
+        }
+      },
+      {
+        "block": "cal-c",
+        "fixedEvaluation": {
+          "cost": {
+            "admittedRuns": 8,
+            "effectsMissingUsage": 0,
+            "modelCalls": 8,
+            "tokensIn": 29489,
+            "tokensOut": 981,
+            "workUnits": 164233
+          },
+          "counts": {
+            "complete": 8,
+            "labelsCorrect": 78,
+            "outputsExact": 1,
+            "recordsCorrect": 78,
+            "recordsTotal": 96,
+            "scorerPassed": 1,
+            "summariesExact": 3,
+            "tasks": 8
+          },
+          "head": "sha256:3d9dcb67baee8b7b0c5da83e85048a6024d1e96698eff3743dca8fd02d058e40",
+          "observedTasks": 8,
+          "outcome": "complete",
+          "session": "sha256:5248a74d5cb9b9bfb026fd02da0418df4657e33db424e29a228b90897fee1491"
+        },
+        "seed": {
+          "account": "sha256:e10299c63dd15086b195f75541438cd273831a34c335640c7e6d881e7ecd05a7",
+          "attempts": [
+            {
+              "developmentRecord": null,
+              "developmentScore": null,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 1,
+              "manifest": "sha256:3d9dcb67baee8b7b0c5da83e85048a6024d1e96698eff3743dca8fd02d058e40",
+              "mode": "generate",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "contrastive-examples",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "writerRequest": "sha256:bbf39a54d9b26b3b4dc3b9c6c8bf57378ac48abcf3bcd550ac2e8775daaaae14"
+            }
+          ],
+          "cost": {
+            "admittedRuns": 4,
+            "effectsMissingUsage": 0,
+            "modelCalls": 4,
+            "tokensIn": 19630,
+            "tokensOut": 2749,
+            "workUnits": 114734
+          },
+          "distinctManifests": 1,
+          "distinctWriterRequests": 1,
+          "duplicateRequests": 0,
+          "generations": 1,
+          "manifest": "sha256:3d9dcb67baee8b7b0c5da83e85048a6024d1e96698eff3743dca8fd02d058e40",
+          "qualified": true,
+          "termination": "qualified"
+        }
+      },
+      {
+        "block": "cal-d",
+        "fixedEvaluation": null,
+        "seed": {
+          "account": "sha256:d43611573629d599677d29ebcda784e5ec6b4d44b75765850862c95433af9654",
+          "attempts": [
+            {
+              "developmentRecord": "sha256:d75c7c373fb642d913e2ba0a47a05bfa3b2a557683466d0cca4edf7b1aacfc25",
+              "developmentScore": 0.33,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 1,
+              "manifest": "sha256:7c820e09fa5df355e66cb32e5c55885bf27a6b3e515c526e5089728b9c4e9bfb",
+              "mode": "generate",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "contrastive-examples",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:d5b5b64dc990579ed7207581ac6030d0e8b676c57e774a7b95fb306dbab8ceea"
+            },
+            {
+              "developmentRecord": "sha256:2ea43b3df27041c23737bd20ff8c177f02bb0232940a648d240a19dad03cea0f",
+              "developmentScore": 0.47,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 2,
+              "manifest": "sha256:5b03df122e047b0176791a31de146470b2817e03f6a61fd0b1b77d1bfecdb998",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "error-analysis",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:f185708789700d9303973a62528067c5b3cb1ede2a96fff51683ecbb929169d3"
+            },
+            {
+              "developmentRecord": "sha256:4c3f9f86a793b157aad686daea2ec2b4ea256bf16cb264a936eb963abe57a662",
+              "developmentScore": 0.4,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 3,
+              "manifest": "sha256:3c15e7aef78345e8af22e4191c0d0e553ca3544b21d7bcffb109bb6c6fd99850",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "invariant-extraction",
+              "trainingBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:4dbca72f1baf93d655ac677e26c90ff8caf9cffe2cb83401b64b0496fc81377d"
+            },
+            {
+              "developmentRecord": "sha256:bee2488302957cf2bdcf485415c5dde02284d5615e28c25b4741a155d3ee0ac3",
+              "developmentScore": 0.73,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 4,
+              "manifest": "sha256:ebc97e026da756b1dfde710d8240fc5abbe8a164a00f8f4a143fa91170c72600",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "counterexample-search",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:643ae3e4431ba9440d79ecbd0e31a6be4ad1a0830b96e8ecadd5f3a998b30c56"
+            },
+            {
+              "developmentRecord": "sha256:a08c37f172886e41969b27a065078a05e670f9a7fe02deef289e14fb3a35c8b6",
+              "developmentScore": 0.8,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 5,
+              "manifest": "sha256:fb205653e76ab7ea3f0cc0b550ca824c505e9c4d83e11ea1f95e7c20ab3b8a2b",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "minimal-rule-rewrite",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:f0f5dca58cc1e71f2d7ed7c1ae94d3b2bc4ee7de9140176e9886c341795350eb"
+            },
+            {
+              "developmentRecord": "sha256:98e7186a38cecc465bc8836aa8222072de2bacdc0aa46505e6d00801b9edb8b8",
+              "developmentScore": 0.8,
+              "duplicateRequest": false,
+              "failureCode": null,
+              "generation": 6,
+              "manifest": "sha256:16b1bce66f8894c362080536c2a1dbd5426919364a5afcf64a82ee5262dc90ea",
+              "mode": "revise",
+              "outcome": "complete",
+              "repeatedManifest": false,
+              "round": "fresh-synthesis",
+              "trainingBatches": {
+                "passed": 1,
+                "total": 1
+              },
+              "validationBatches": {
+                "passed": 0,
+                "total": 1
+              },
+              "writerRequest": "sha256:eb983124f3f2962757722ce0adc6f90062e935aa63cf2301ffb2ea1b7139f6da"
+            }
+          ],
+          "cost": {
+            "admittedRuns": 30,
+            "effectsMissingUsage": 0,
+            "modelCalls": 30,
+            "tokensIn": 157637,
+            "tokensOut": 16130,
+            "workUnits": 965241
+          },
+          "distinctManifests": 6,
+          "distinctWriterRequests": 6,
+          "duplicateRequests": 0,
+          "generations": 6,
+          "interrupted": {
+            "accountedCost": {
+              "admittedRuns": 30,
+              "effectsMissingUsage": 0,
+              "modelCalls": 30,
+              "tokensIn": 157637,
+              "tokensOut": 16130,
+              "workUnits": 965241
+            },
+            "block": "cal-d",
+            "cause": "The host process running `run.ts calibrate` was killed by the session monitor's expiry at about 2026-09-29T13:52 local time during cal-d generation 7, after the writer call completed and before the candidate was executed. The driver records `interrupted` only on a thrown error, so the retained snapshot reads in-progress.",
+            "completedGenerations": 6,
+            "inFlightGeneration": 7,
+            "inFlightMode": "revise",
+            "inFlightReceipts": [
+              {
+                "generation": 7,
+                "inSeedAccount": false,
+                "receipt": "sha256:15c387ea3a34cbfc1d7c6efc0ac61fbd64d16e73ecd5583c4aedf76ba109546c",
+                "scored": false,
+                "usage": [
+                  {
+                    "model": "grok-4.5",
+                    "tokensIn": 12304,
+                    "tokensOut": 2046
+                  }
+                ],
+                "work": {
+                  "agentCalls": 1,
+                  "steps": 9,
+                  "units": 46412
+                }
+              }
+            ],
+            "inFlightRound": "fresh-synthesis",
+            "retainedSnapshot": "seed-d5c1a7102e80c6602f58f8f1f36e3674d73fe4a7f9b23ffb2dbfb51b777362f1.json",
+            "retry": "none; the protocol forbids inferring a retry from missing output, and the stage guard refuses a second seed store",
+            "stage": "seed"
+          },
+          "manifest": null,
+          "qualified": null,
+          "termination": "in-progress"
+        }
+      }
+    ]
+  },
+  "confirmatory": {
+    "observedBlocks": 0,
+    "plannedBlocks": 9,
+    "primaryClaim": null
+  },
+  "costBreakdown": {
+    "driverAccounted": {
+      "admittedRuns": 99,
+      "effectsMissingUsage": 0,
+      "modelCalls": 99,
+      "tokensIn": 523869,
+      "tokensOut": 50012,
+      "workUnits": 3123192
+    },
+    "inFlightWriterReceipts": {
+      "admittedRuns": 1,
+      "effectsMissingUsage": 0,
+      "modelCalls": 1,
+      "tokensIn": 12304,
+      "tokensOut": 2046,
+      "workUnits": 46412
+    },
+    "scope": "driverAccounted sums the calibration decision cost over the three complete blocks and the interrupted block's seed account; inFlightWriterReceipts adds the receipt of the generation whose execution never ran, which no account or attempt references."
+  },
+  "costDefinitions": {
+    "modelCalls": "Charged executor attempts; no provider billing attestation.",
+    "providerDollars": null,
+    "tokens": "Recorded usage counted per admitted run, including content-identical receipts. One missing usage makes full totals null.",
+    "workUnits": "Runtime work accounting, including every failed or repeated admission."
+  },
+  "dateTimeZone": "America/Puerto_Rico",
+  "developmentSignal": {
+    "interpretation": "The graded score varied within and across searches (0.33 to 0.80) and rose on later generations of both failed or unfinished searches, so the writer received a non-constant signal, unlike v7. No scored candidate reached the 0.9 selection threshold on its development batch, and the two qualifying searches passed selection on a generation that had no development score of its own. The score history did not convert into a selection pass on cal-b within eight generations or on cal-d within six.",
+    "maximumScore": 0.8,
+    "minimumScore": 0.33,
+    "repeatedManifests": {
+      "cal-a": [],
+      "cal-b": [
+        7
+      ],
+      "cal-c": [],
+      "cal-d": []
+    },
+    "scoredCandidates": 15,
+    "scores": {
+      "cal-a": [
+        0.53,
+        null
+      ],
+      "cal-b": [
+        0.53,
+        0.53,
+        0.47,
+        0.73,
+        0.67,
+        0.67,
+        0.67,
+        0.6
+      ],
+      "cal-c": [
+        null
+      ],
+      "cal-d": [
+        0.33,
+        0.47,
+        0.4,
+        0.73,
+        0.8,
+        0.8
+      ]
+    }
+  },
+  "executor": {
+    "baseUrl": "https://api.x.ai/v1",
+    "model": "grok-4.5"
+  },
+  "identity": {
+    "datasets": "sha256:c06b6cdadd478b643f59a72700bac9fada37433bbf9364e6718656fa87528945",
+    "freeze": "sha256:c926270aa709562c4ecc5b588bd65890dd1bd508676ea2e38809428bc02f8f0e",
+    "nativeResults": null,
+    "protocol": "sha256:540de4213920ea240313a098fc9f3e80828162408845301357f594d0388f3c3c",
+    "reconciliation": "sha256:d389cde6b6c9af83590ecb9acc46a534cb932c3a8af4d6ed5f2b639775d8acce"
+  },
+  "interruption": {
+    "accountedCost": {
+      "admittedRuns": 30,
+      "effectsMissingUsage": 0,
+      "modelCalls": 30,
+      "tokensIn": 157637,
+      "tokensOut": 16130,
+      "workUnits": 965241
+    },
+    "block": "cal-d",
+    "cause": "The host process running `run.ts calibrate` was killed by the session monitor's expiry at about 2026-09-29T13:52 local time during cal-d generation 7, after the writer call completed and before the candidate was executed. The driver records `interrupted` only on a thrown error, so the retained snapshot reads in-progress.",
+    "completedGenerations": 6,
+    "inFlightGeneration": 7,
+    "inFlightMode": "revise",
+    "inFlightReceipts": [
+      {
+        "generation": 7,
+        "inSeedAccount": false,
+        "receipt": "sha256:15c387ea3a34cbfc1d7c6efc0ac61fbd64d16e73ecd5583c4aedf76ba109546c",
+        "scored": false,
+        "usage": [
+          {
+            "model": "grok-4.5",
+            "tokensIn": 12304,
+            "tokensOut": 2046
+          }
+        ],
+        "work": {
+          "agentCalls": 1,
+          "steps": 9,
+          "units": 46412
+        }
+      }
+    ],
+    "inFlightRound": "fresh-synthesis",
+    "retainedSnapshot": "seed-d5c1a7102e80c6602f58f8f1f36e3674d73fe4a7f9b23ffb2dbfb51b777362f1.json",
+    "retry": "none; the protocol forbids inferring a retry from missing output, and the stage guard refuses a second seed store",
+    "stage": "seed"
+  },
+  "limitations": [
+    "The cal-d seed search was interrupted at generation seven by the host session, not by the protocol; its outcome is unknown and the protocol forbids a retry. The gate had already failed on the other three blocks.",
+    "The calibration gate required all four seeds to qualify and headroom on three blocks.",
+    "No optimizer or other confirmatory learning arm ran; there is no optimizer advantage or nine-block repeatability result.",
+    "Fresh draws from one shared synthetic template family do not represent separate domains or human-adjudicated support messages.",
+    "The v7 and v8 draws differ; this is not a paired causal estimate of the graded-score change.",
+    "Validation selected seeds repeatedly and is not an untouched final test.",
+    "Provider charges remain unknown."
+  ],
+  "lineRuleOutcome": "The gate failed on the three complete blocks alone (one seed did not qualify; two blocks with headroom), before and independent of the interruption. Under the preregistered line rule the development-feedback seed line ends; no further protocol on this axis is proposed.",
+  "method": {
+    "budgets": {
+      "frozen": {
+        "attempts": 1024,
+        "runs": 512,
+        "work": 40000000
+      },
+      "learning": {
+        "attempts": 1024,
+        "runs": 512,
+        "work": 40000000
+      },
+      "seed": {
+        "attempts": 64,
+        "runs": 64,
+        "work": 8000000
+      }
+    },
+    "calibrationRule": {
+      "blocks": [
+        "cal-a",
+        "cal-b",
+        "cal-c",
+        "cal-d"
+      ],
+      "failure": "collect and replay the negative calibration; do not start confirmatory blocks or tune this protocol; the development-feedback seed line ends",
+      "headroomRule": "at least three fixed seeds pass no more than six of eight fresh frozen tasks",
+      "proceed": "all four seed searches qualify and headroomRule passes",
+      "requiredSeeds": 4
+    },
+    "developmentUse": "A candidate that failed selection is executed once on a separate 12-record development batch; only its agreement score (rounded to hundredths) reaches later writer requests. No development records, labels, outputs, or summaries enter any request.",
+    "family": "synthetic record triage",
+    "fixedRecordsPerTask": 12,
+    "lineRule": "If this calibration fails, the development-feedback seed line ends; a further protocol on this axis needs a new hypothesis, not a retuned signal",
+    "maximumSeedGenerations": 8,
+    "plannedConfirmatoryComparison": {
+      "blocks": 9,
+      "frozenBeforeEvaluation": true,
+      "groups": [
+        "a",
+        "b",
+        "c"
+      ],
+      "primary": "optimizer-versus-fixed",
+      "productionActivation": false,
+      "repetitionsPerGroup": 3,
+      "requiredCoverage": [
+        "all nine confirmatory blocks",
+        "fixed arm",
+        "optimizer arm"
+      ],
+      "secondaryFailures": "report separately",
+      "successRule": "optimizer strictly wins at least two of three blocks in every group and wins aggregate passes"
+    },
+    "roundSchedule": [
+      "contrastive-examples",
+      "error-analysis",
+      "invariant-extraction",
+      "counterexample-search",
+      "minimal-rule-rewrite",
+      "fresh-synthesis",
+      "fresh-synthesis",
+      "fresh-synthesis"
+    ],
+    "score": "(4 * exact full-record fraction + exact-summary indicator) / 5",
+    "scorerPassAt": 0.9,
+    "seedDevelopmentRecords": 12,
+    "seedTrainingRecords": 24,
+    "seedValidationRecords": 12,
+    "taskSeeds": [
+      {
+        "id": "cal-a",
+        "seeds": {
+          "acquisition": 81000001,
+          "ancestor": 81000004,
+          "shift": 81000003,
+          "unseen": 81000002
+        },
+        "stage": "calibration"
+      },
+      {
+        "id": "cal-b",
+        "seeds": {
+          "acquisition": 81000005,
+          "ancestor": 81000008,
+          "shift": 81000007,
+          "unseen": 81000006
+        },
+        "stage": "calibration"
+      },
+      {
+        "id": "cal-c",
+        "seeds": {
+          "acquisition": 81000009,
+          "ancestor": 81000012,
+          "shift": 81000011,
+          "unseen": 81000010
+        },
+        "stage": "calibration"
+      },
+      {
+        "id": "cal-d",
+        "seeds": {
+          "acquisition": 81000013,
+          "ancestor": 81000016,
+          "shift": 81000015,
+          "unseen": 81000014
+        },
+        "stage": "calibration"
+      },
+      {
+        "id": "a-r1",
+        "seeds": {
+          "acquisition": 81000017,
+          "ancestor": 81000020,
+          "shift": 81000019,
+          "unseen": 81000018
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "a-r2",
+        "seeds": {
+          "acquisition": 81000021,
+          "ancestor": 81000024,
+          "shift": 81000023,
+          "unseen": 81000022
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "a-r3",
+        "seeds": {
+          "acquisition": 81000025,
+          "ancestor": 81000028,
+          "shift": 81000027,
+          "unseen": 81000026
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "b-r1",
+        "seeds": {
+          "acquisition": 81000029,
+          "ancestor": 81000032,
+          "shift": 81000031,
+          "unseen": 81000030
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "b-r2",
+        "seeds": {
+          "acquisition": 81000033,
+          "ancestor": 81000036,
+          "shift": 81000035,
+          "unseen": 81000034
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "b-r3",
+        "seeds": {
+          "acquisition": 81000037,
+          "ancestor": 81000040,
+          "shift": 81000039,
+          "unseen": 81000038
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "c-r1",
+        "seeds": {
+          "acquisition": 81000041,
+          "ancestor": 81000044,
+          "shift": 81000043,
+          "unseen": 81000042
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "c-r2",
+        "seeds": {
+          "acquisition": 81000045,
+          "ancestor": 81000048,
+          "shift": 81000047,
+          "unseen": 81000046
+        },
+        "stage": "confirmatory"
+      },
+      {
+        "id": "c-r3",
+        "seeds": {
+          "acquisition": 81000049,
+          "ancestor": 81000052,
+          "shift": 81000051,
+          "unseen": 81000050
+        },
+        "stage": "confirmatory"
+      }
+    ],
+    "validationUse": "Repeated selection of the first passing candidate; validation labels, outputs, and scores excluded from writer requests.",
+    "writerFeedback": "Training examples, previous training output beside expected training output, the generation number, the predeclared round mode, and a bounded development score history of {generation, round, score} only, where score is the agreement value in [0,1] rounded to hundredths.",
+    "writerRequestIdentity": "Digest of the executed writer program and its exact arguments; a duplicate request ends the search as a protocol failure.",
+    "wrongRowCountScore": 0
+  },
+  "partialTokenSubtotals": {
+    "effectsWithUsage": 84,
+    "scope": "Per-receipt-file subtotals over every effect; all recorded usage.",
+    "tokensIn": 465290,
+    "tokensOut": 48752
+  },
+  "recordedFailureLimit": "A failed local receipt does not establish the exact abort cause, provider completion, or billing; each listed miss is a transport failure, not a scored miss.",
+  "recordedFailures": [],
+  "schema": "algal.cumulative-skill-public.v8",
+  "sourceCommit": "dc64c02f9274ae92f5475cc9a9a2a4f2c8e4fd57",
+  "sourceRepository": "https://github.com/hraness/algal",
+  "status": "calibration-interrupted-gate-failed",
+  "study": "cumulative-skill-v8-calibration",
+  "studyDate": "2026-09-29",
+  "totalRealizedCost": {
+    "admittedRuns": 100,
+    "effectsMissingUsage": 0,
+    "modelCalls": 100,
+    "tokensIn": 536173,
+    "tokensOut": 52058,
+    "workUnits": 3169604
+  },
+  "verification": {
+    "countScope": "Receipt files per store; copied seed receipts are counted in every store that contains them. The driver replay entry point requires a collected result, so the reconciliation applied the same receipt verifier to every store, including the incomplete one, and the completed cal-d attempts passed the driver's seed verification treated as a six-generation summary.",
+    "interpretation": "Recorded execution replay is separate from task correctness and provider billing.",
+    "offlineReplayPassed": true,
+    "receiptFilesChecked": 84,
+    "stores": [
+      {
+        "checkedReceipts": 8,
+        "complete": true,
+        "store": "cal-a/frozen-fixed"
+      },
+      {
+        "checkedReceipts": 8,
+        "complete": true,
+        "store": "cal-a/seed"
+      },
+      {
+        "checkedReceipts": 30,
+        "complete": true,
+        "store": "cal-b/seed"
+      },
+      {
+        "checkedReceipts": 8,
+        "complete": true,
+        "store": "cal-c/frozen-fixed"
+      },
+      {
+        "checkedReceipts": 3,
+        "complete": true,
+        "store": "cal-c/seed"
+      },
+      {
+        "checkedReceipts": 27,
+        "complete": false,
+        "store": "cal-d/seed"
+      }
+    ]
+  }
+}