{
  "schema": "agent-art-lab.document/v1",
  "id": "projects-thought-readme",
  "title": "THOUGHT: an annotated worked case",
  "page": "/projects/thought/",
  "revision": "sha256:afac3ee9cbb96882cf16672b527a793f23a7bc5bc8fc995d71aad65c543e376c",
  "source": {
    "path": "projects/thought/README.md",
    "url": "https://github.com/agent-art-collective/Agent-Art-Lab/blob/main/projects/thought/README.md",
    "sha256": "afac3ee9cbb96882cf16672b527a793f23a7bc5bc8fc995d71aad65c543e376c",
    "byteLength": 7654
  },
  "contentFormat": "markdown",
  "content": "# THOUGHT: an annotated worked case\n\nAn exact human prompt and one Agent response become the candidate THOUGHT work.\nAgent Art means art in which an Agent participates at the level of intention;\na response or protocol pass alone does not establish that participation.\n\nThis collection examines technical initiation and completion. It neither grades\nartworks nor proves intentional participation. Read [Guidance](../../GUIDANCE.md).\nTHOUGHT implementation, runners, protocols and product repairs remain owned by\nApplications; they are not the central Lab's next priority.\n\n## Collection index\n\n| Record | Status/evidence class | Retained here | Missing or private evidence |\n| --- | --- | --- | --- |\n| Claude CLI pilot, 2026-09-17 | Completed observation with conflicting outcomes | Reviewed summary below and reported artifact identity | Raw JSON/report and transcript unavailable here. |\n| Claude CLI pilot, 2026-09-18 | Completed strict-fixture observation | Reviewed summary below and reported artifact identity | Raw report unavailable; not a real-App canary. |\n| [Codex diagnostic](studies/2026-09-20-model-acquisition.md) | Completed retrospective diagnosis | Sanitized analysis, excerpt, method review | Original run privately retained; exact canary build/handoff identity unknown. |\n| [Boundary corrections and canary follow-up](studies/2026-09-24-boundaries-and-canary-follow-up.md) | Retrospective update; OPS reports production delivered | Sanitized observations, scoped lessons and source references | Raw evidence excluded; Lab did not independently rerun or verify canaries/deployments. |\n| [Native tools and explicit execution prerequisites](studies/2026-10-01-native-explicit-execution.md) | Retrospective contribution authorized by the operator on 2026-10-01 | Sanitized OPS account and provisional connection to P-04–P-07 | Private runtime evidence unavailable here; no Lab rerun or controlled reliability/timing comparison. |\n| Initiation-reliability comparison | Proposed, unfrozen, unrun | Scope/status note below | No comparison results. |\n| [Practice-led artwork study](studies/2026-09-20-intentional-participation-proposal.md) | Proposal prepared; study unrun | Question, evidence inventory and proposed reading method | No candidate selected or artwork exported for interpretation. |\n\n## Provenance and access\n\nThese are portable derivatives of Applications research notes. The operator\nauthorized publication of this reviewed Lab edition on 2026-09-21; private\noriginals remain outside the published file set.\nThe source checkout contained uncommitted documents; its Git HEAD alone does\nnot identify them or the tested implementation. [Provenance](../../PROVENANCE.json)\nrecords the observed source-document hashes.\n\nThe 2026-09-24 follow-up was initially a local-only addition from sanitized OPS\nreports. The operator subsequently authorized its commit and push that day.\n\nThe 2026-10-01 native/explicit execution study was initially prepared locally\nwithout publication authority. The operator subsequently authorized publication\nof this sanitized contribution that day; the connecting guidance stays provisional.\n\nRaw reports and run evidence are privately retained and unavailable here.\nTheir recorded hashes do not make the evidence independently inspectable.\nPrivate identifiers, machine paths, credentials and raw handoffs are omitted.\nNo exact deployed commit/handoff hash was established for the failed Codex\ncanary; current source correspondence must not be labeled immutable build proof.\n\n## Claude CLI pilot, 2026-09-17\n\nOne trial used configured `claude-sonnet-5`, Medium effort and Auto permissions\nagainst a permissive local synthetic fixture. The configured CLI was\n`2.1.269`. Limits included one trial, two minutes, eight turns and a $1 client\nbudget setting—not a hard provider billing cap.\n\nThe child exited `1` with `error_max_turns` and `isError: true`.\nSeparately, the fixture reached `returned` with receipt acceptance.\nThat acceptance did not establish clean completion or protocol correctness.\nThe original runner's batch-success label/exit `0` were reporting defects,\npreserved by the original analysis. Later corrections do not change this result.\n\nReported original artifact: JSON 16,915 bytes,\nSHA-256 `6bc6fe2f4c6643b660f62682413a59459211b6fa10e1f33dad1adfe5bf8babed`,\nrunner `agent-lab-runner-v1`.\n\nThe retained report omitted tool arguments/results; the complete handshake\ncannot be reconstructed from that report. This is not Desktop, real-App,\nart-quality or acceptance-rate evidence.\n\n## Claude CLI strict-fixture pilot, 2026-09-18\n\nA separately authorized trial used `2.1.269 (Claude Code)`, configured\n`claude-sonnet-5`, Medium/Auto, two minutes, eight turns, a $1 client-budget\nsetting and 256 KiB child output. No repeat trial occurred.\n\nThe recorded outcomes were separate:\n\n1. Runner ended without a batch stop condition.\n2. Child exited `0`, with `isError: false`.\n3. Independent synthetic observer accepted ordered claim → ready → start →\n   result, reached returned state and issued a fixture receipt.\n\nOnly the observer evidence established the strict synthetic protocol pass;\nassistant prose and exit code did not.\n\nReported original artifact: JSON 21,162 bytes,\nSHA-256 `799c89b2773ef952760125cb356744e94bfb67ef4f1f57b1224fc4a08ea7b921`,\nrunner `agent-lab-runner-v3`.\n\nThe runtime reported `claude-sonnet-5`; it was not provider attestation.\nRuntime-reported effort/permission remained unknown. Runner and fixture changed,\nso this was not a controlled comparison with pilot 1.\n\n## Codex failure and retrospective diagnosis\n\nOne real run reportedly claimed successfully then failed with\n`AGENT_START_FAILED` / `Model unavailable`, without creative start or receipt.\nOPS inspected commands that searched App claim data for a host model, with no\nobserved host-metadata lookup. Stored host values existed, but agent access was\nnot established.\n\nThe [diagnostic](studies/2026-09-20-model-acquisition.md) explains why a synthetic\nfailure-handling PASS did not contradict that observation. Supported acquisition\nand product repair remain unresolved.\n\n## Follow-up, 2026-09-24\n\nThe [later boundary study](studies/2026-09-24-boundaries-and-canary-follow-up.md)\nrecords hash/worker corrections, two successful fresh manual staging canaries,\nremaining Codex compliance gaps and completed production delivery as reported by\nOPS. Production delivery checks included no production Agent run. These later\nobservations preserve the earlier failed cases; they do not establish the cause\nof the historical HTTP failure, complete compliance or general reliability.\n\n## Draft comparison and artistic gap\n\nThe proposed initiation-reliability comparison remains **unfrozen and unrun**.\nNo treatment effect, replicated finding or general acceptance rate is claimed.\nThe original draft remains project-owned; its live settings are not copied here.\n\nNo artwork is retained here for curatorial assessment. Fixture output and\nreceipt acceptance cannot establish Agent intention or artistic quality.\nThe [practice-led proposal](studies/2026-09-20-intentional-participation-proposal.md)\nsets out how to examine documented choices and interpretations once an existing\ncandidate and permitted material are identified. It contains no artistic findings.\n\n## Follow-up ownership\n\nApplications owns the product investigation and any authorized regression or\ncanary. Agent-Art-Lab has prepared the practice-led proposal; its next study step\nneeds operator-selected material and review scope. This collection grants no\nlive execution, spending,\nprivate-source access, deployment or policy-change authority.\n",
  "assets": []
}