Round shaping-model-1

Writing the specification inside the attempt: a stronger model did not help

Status · DESCRIPTIVEReviewed status is transcribed from rounds/STATUS.md. The original narrative and its qualifications remain below.

Source-reported narrativeRound README and results text are source material. Numeric values that are not tied to a claim ledger record remain source-reported.

Round source README · rounds/shaping-model-1/README.md

Descriptive run in Kogen Lab. Kogen first writes a specification (short prose, executable acceptance scenarios and their step code) and then builds against it. This run compared a supplied specification with one written inside each attempt, by a cheap model and by a stronger one. With the supplied specification, 4 of 5 finished attempts were correct. With a specification written in the attempt, none of the finished attempts were: 0 of 8 with the cheap model and 0 of 5 with the stronger one. The records suggest why for the Sol-written specifications. In most of them the step code was broken, so the acceptance check failed whatever the build did. Most Luna-written attempts never produced a specification that passed readiness. Kogen's readiness check accepted it, because before a build the scenarios are expected to fail anyway. The builder is not allowed to change the acceptance files, so those attempts could not succeed and ran until the repair limit. A Lab crash ended the run after 22 attempts. Its decision was already reached, so the six unfinished cells (four lost in the crash, two never started) were not rerun.

Question and scope

When the specification is written inside each attempt from the task's request, does GPT-6.1 Sol at low effort as specification writer reach the correctness of a supplied specification, where GPT-6 Luna at low effort does not? GPT-6 Luna at low effort plans and builds; GPT-6.1 Sol at low effort reviews after every passing verification. The run used Kogen Lab, which reproduces Kogen's specification, readiness, file protection and build loop with its own engine; its equivalence to Kogen is not measured.

  • Tasks: r70-5-elixir, r70-7-elixir, r70-5-go, r70-7-go, as in the second mechanism screen.
  • Configurations: the supplied specification used in the earlier rounds; a specification written by Luna; a specification written by Sol. Kogen's defaults for writing it: one session, feasibility probes allowed, a formal model only when warranted.
  • Two repetitions per task; each task and repetition ran its three configurations on one host.
  • Every attempt carried the experimental stops used in the screens: at most six repairs and 100,000 output tokens.
  • The hidden tests grade each attempt independently of the workflow's own checks and review.

Results

"Stopped" means the six-repair stop ended the attempt; "specification failed" means the writer never produced a specification that passed readiness; "lost" means a Lab crash ended the attempt before an outcome; "not run" means the cell was never started.

Task, repetition Supplied Written by Luna Written by Sol
r70-5-elixir 1 lost stopped stopped
r70-5-elixir 2 stopped specification failed stopped
r70-7-elixir 1 correct specification failed stopped
r70-7-elixir 2 correct specification failed stopped
r70-5-go 1 not run stopped lost
r70-5-go 2 correct specification failed stopped
r70-7-go 1 correct stopped lost
r70-7-go 2 not run specification failed lost
Correct, of finished attempts 4 of 5 0 of 8 0 of 5

Of the 7 Sol-written attempts that reached verification (the five stopped ones and two of the three lost ones), 6 failed the acceptance check on every verification. The seventh passed verification every time and was sent back by the review each time. One broken step file, for example, parsed table cells as JSON and failed on line breaks inside them.

A Lab crash during writing a specification ended the run after 22 attempts (since fixed). The remaining cells were not completed because the decision rule fixed before the run was already met: the Sol-written specification was behind the supplied one in three of the four pairs where both finished, and the Luna-written one in four of five. The rule fixed before the run was that a configuration behind in at least three pairs was enough to stop. The run used about 8.1M input and 224k output tokens.

Limits

  • Few cells, four tasks of one family, and one style of specification writing (Kogen's prompt and defaults).
  • The six-repair stop is experimental; Kogen has no such stop, and these attempts would have kept going.
  • Results describe Kogen Lab's reproduction of Kogen; equivalence with Kogen itself is not measured.
  • Lab's evaluation is not an official kogen-bench grade, so the Standard records declare the official outcome as missing (see "What's missing and why").

Required reproduction metadata

  • Kogen commit: fd0cdcfb422f6edef5ec69146cdecdb3e3c1317a, as reproduced by Kogen Lab.
  • Harness commit: not recorded as a public Git commit. Kogen Lab's source is not published yet; its build fingerprint is in the bundle's manifest.json.
  • Model and effort: planner and builder gpt-6-luna at low; reviewer gpt-6.1-sol at low; specification writer gpt-6-luna at low or gpt-6.1-sol at low, as stated per configuration.
  • Task IDs: r70-5-elixir, r70-7-elixir, r70-5-go, r70-7-go.
  • Command: kogen-lab freeze, then kogen-lab run (ended by a Lab crash), then kogen-lab export and Lab's export_bench_round.py to write these records.
  • Raw records: `records.jsonl` and the Lab export in `bundles/`.

Metadata

Round date: 2026-10-18
Publication badge: DESCRIPTIVE; KEPT FOR AUDIT
Recomputation status: OBSERVED SOURCE ONLY
Why not VALID: DESCRIPTIVE_ONLY, PROTOCOL_DEVIATION, INCOMPLETE_CAPTURE — Lab development run stopped by a Lab crash with its decision reached, six cells not completed, no official kogen-bench grade and declared missing Standard fields.
STATUS: DESCRIPTIVE

Execution record

  • One experiment, frozen before its first attempt and run from a frozen Lab development build over six host slots (three work folders on one macOS workstation, two on one Ubuntu 24.04 server, one Ubuntu 24.04 laptop).
  • Records: `records.jsonl` holds one Standard 1.2 record per attempt (22), including the four the crash left without an outcome. `MISSING.md` and `MEASURED.md` declare what the records cannot carry.
  • `bundles/` holds Lab's public export; its EXCLUDED.json lists what was left out and why. The written specifications themselves are workflow artifacts and stay out of the export.

What's missing and why

  • circumstances.cap_end (22 missing): The Lab export does not carry this Standard field. Boundary telemetry not retained. Reason: Boundary telemetry not retained (m08).
  • circumstances.cap_start (22 missing): The Lab export does not carry this Standard field. Boundary telemetry not retained. Reason: Boundary telemetry not retained (m08).
  • circumstances.concurrent_cells_end (22 missing): The Lab export does not carry this Standard field. Host-wide concurrent cell count was not captured. Reason: Host-wide concurrent cell count was not captured (m66).
  • circumstances.concurrent_cells_start (22 missing): The Lab export does not carry this Standard field. Host-wide concurrent cell count was not captured. Reason: Host-wide concurrent cell count was not captured (m66).
  • circumstances.dispatcher_id (22 missing): The Lab export does not carry this Standard field. Exact dispatcher revision is not present in the public receipt extract. Reason: Exact dispatcher revision is not present in the public receipt extract (m63).
  • circumstances.incidents (22 missing): The Lab export does not carry this Standard field. Per-cell incident receipt was not captured. Reason: Per-cell incident receipt was not captured (m78).
  • circumstances.load1_end (22 missing): The Lab export does not carry this Standard field. Host boundary load was not captured in the public round data. Reason: Host boundary load was not captured in the public round data (m65).
  • circumstances.load1_start (22 missing): The Lab export does not carry this Standard field. Host boundary load was not captured in the public round data. Reason: Host boundary load was not captured in the public round data (m65).
  • circumstances.load_samples (22 missing): The Lab export does not carry this Standard field. Timestamped host-wide load samples were not captured. Reason: Timestamped host-wide load samples were not captured (m93).
  • circumstances.queue (22 missing): The Lab export does not carry this Standard field. Queue receipt was not captured. Reason: Queue receipt was not captured (m91).
  • cost.accounting (22 missing): The Lab export does not carry this Standard field. Per-cell cost and pricing inputs are not present in the public round data. Reason: Per-cell cost and pricing inputs are not present in the public round data (m70).
  • cost.calculator_version (22 missing): The Lab export does not carry this Standard field. Cost calculator version not recorded. Reason: Cost calculator version not recorded (m61).
  • cost.long_context_reconciled (22 missing): The Lab export does not carry this Standard field. Boolean receipt not recorded. Reason: Boolean receipt not recorded (m07).
  • cost.price_table_version (22 missing): The Lab export does not carry this Standard field. Price table version not recorded. Reason: Price table version not recorded (m89).
  • cost.usd (22 missing): The Lab export does not carry this Standard field. Per-cell cost and pricing inputs are not present in the public round data. Reason: Per-cell cost and pricing inputs are not present in the public round data (m70).
  • environment.account_class (22 missing): The Lab export does not carry this Standard field. No owner-approved class-only account receipt is available for the cell. Reason: No owner-approved class-only account receipt is available for the cell (m100).
  • environment.network.allowlist_hosts (22 missing): The Lab export does not carry this Standard field. Per-cell network allowlist not present in the public round data. Reason: Per-cell network allowlist not present in the public round data (m79).
  • environment.network.profile (22 missing): The Lab export does not carry this Standard field. Per-cell network profile not present in the public round data. Reason: Per-cell network profile not present in the public round data (m80).
  • environment.toolchains.elixir (12 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • environment.toolchains.erlang (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • environment.toolchains.node (4 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • environment.toolchains.other_inventory (22 missing): The Lab export does not carry this Standard field. Full per-cell toolchain inventory not recorded. Reason: Full per-cell toolchain inventory not recorded (m64).
  • environment.toolchains.python (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • environment.toolchains.ruby (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • environment.toolchains.rust (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • grade.grader (22 missing): The Lab export does not carry this Standard field. Per-cell grader software version is not in the public data. Reason: Per-cell grader software version is not in the public data (m77).
  • grade.timestamp (22 missing): The Lab export does not carry this Standard field. No official grade in the public snapshot. Reason: No official grade in the public snapshot (m32).
  • host.id (22 missing): The Lab export does not carry this Standard field. Not recorded in available public metadata. Reason: Not recorded in available public metadata (m39).
  • itt.class (22 missing): The Lab export does not carry this Standard field. No evidence-backed ITT classification. Reason: No evidence-backed ITT classification (m28).
  • itt.cohort (22 missing): The Lab export does not carry this Standard field. No evidence-backed ITT classification. Reason: No evidence-backed ITT classification (m28).
  • model.effective (12 missing): The Lab export does not carry this Standard field. Per-cell effective-model receipt not published. Reason: Per-cell effective-model receipt not published (m74).
  • model.requested (22 missing): The Lab export does not carry this Standard field. Not recorded in available public metadata. Reason: Not recorded in available public metadata (m39).
  • outcome (22 missing): The Lab export does not carry this Standard field. Not available for ungraded delivery. Reason: Not available for ungraded delivery (m38).
  • sandbox.egress_allow (22 missing): The Lab export does not carry this Standard field. Per-cell egress allowlist is not present in the public round data. Reason: Per-cell egress allowlist is not present in the public round data (m75).
  • sandbox.egress_profile (22 missing): The Lab export does not carry this Standard field. Per-cell egress profile is not present in the public round data. Reason: Per-cell egress profile is not present in the public round data (m76).
  • sandbox.profile (22 missing): The Lab export does not carry this Standard field. Per-cell sandbox profile is not present in the public round data. Reason: Per-cell sandbox profile is not present in the public round data (m87).
  • sandbox.profile_sha256 (22 missing): The Lab export does not carry this Standard field. Per-cell sandbox profile fingerprint is not present in the public round data. Reason: Per-cell sandbox profile fingerprint is not present in the public round data (m86).
  • setup.adapter_harness_sha (22 missing): The Lab export does not carry this Standard field. Not recorded in available public metadata. Reason: Not recorded in available public metadata (m39).
  • setup.deps_source (22 missing): The Lab export does not carry this Standard field. Per-cell dependency source is not recorded. Reason: Per-cell dependency source is not recorded (m71).
  • setup.sandbox_mode (22 missing): The Lab export does not carry this Standard field. Per-cell sandbox mode is not present in the public round data. Reason: Per-cell sandbox mode is not present in the public round data (m85).
  • setup.sandbox_profile_sha256 (22 missing): The Lab export does not carry this Standard field. Per-cell sandbox fingerprint is not present in the public round data. Reason: Per-cell sandbox fingerprint is not present in the public round data (m84).
  • setup.task_base.hash (22 missing): The Lab export does not carry this Standard field. Original base commit/tree hash absent; fresh_base_commit is not a base hash. Reason: Original base commit/tree hash absent; fresh_base_commit is not a base hash (m42).
  • setup.task_base.kind (22 missing): The Lab export does not carry this Standard field. Original base revision type not recorded per cell. Reason: Original base revision type not recorded per cell (m43).
  • stop_reason (5 missing): The Lab export does not carry this Standard field. No normalized stop receipt. Reason: No normalized stop receipt (m31).
  • task.base_repo (22 missing): The Lab export does not carry this Standard field. Not recorded in available public metadata. Reason: Not recorded in available public metadata (m39).
  • task.base_revision.hash (22 missing): The Lab export does not carry this Standard field. Original base commit/tree hash absent; fresh_base_commit is not a base hash. Reason: Original base commit/tree hash absent; fresh_base_commit is not a base hash (m42).
  • task.base_revision.kind (22 missing): The Lab export does not carry this Standard field. Original base revision type not recorded per cell. Reason: Original base revision type not recorded per cell (m43).
  • timestamps.attempts (4 missing): The Lab export does not carry this Standard field. Attempt-boundary receipts are not in the public round data. Reason: Attempt-boundary receipts are not in the public round data (m60).
  • timestamps.cell.end_utc (4 missing): The Lab export does not carry this Standard field. Per-cell contestant end timestamp not recorded in the public round data. Reason: Per-cell contestant end timestamp not recorded in the public round data (m68).
  • timestamps.phases.develop.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.develop.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.gate.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.gate.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.grade.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.grade.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.plan.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.plan.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.review.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.review.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.setup.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.setup.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.shape.end_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timestamps.phases.shape.start_utc (22 missing): The Lab export does not carry this Standard field. Phase timestamp not present in the public per-cell data. Reason: Phase timestamp not present in the public per-cell data (m88).
  • timing.phases_s.develop (22 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.phases_s.gate (22 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.phases_s.plan (9 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.phases_s.review (15 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.phases_s.setup (22 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.phases_s.shape (9 missing): The Lab export does not carry this Standard field. Phase wall not emitted or not separable. Reason: Phase wall not emitted or not separable (m45).
  • timing.timeout_cap_s (22 missing): The Lab export does not carry this Standard field. Timeout receipt unavailable. Reason: Timeout receipt unavailable (m53).
  • timing.total_wall_s (4 missing): The Lab export does not carry this Standard field. Wall counter unavailable. Reason: Wall counter unavailable (m58).
  • tools.grader (22 missing): The Lab export does not carry this Standard field. Per-cell grader software version is not in the public data. Reason: Per-cell grader software version is not in the public data (m77).
  • tools.runner (22 missing): The Lab export does not carry this Standard field. Per-cell runner version not recorded in the public round data. Reason: Per-cell runner version not recorded in the public round data (m83).
  • tools.toolchains.elixir (12 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • tools.toolchains.erlang (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • tools.toolchains.node (4 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • tools.toolchains.other_inventory (22 missing): The Lab export does not carry this Standard field. Full per-cell toolchain inventory not recorded. Reason: Full per-cell toolchain inventory not recorded (m64).
  • tools.toolchains.python (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • tools.toolchains.ruby (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).
  • tools.toolchains.rust (22 missing): The Lab export does not carry this Standard field. Toolchain version not recorded. Reason: Toolchain version not recorded (m95).

Captured run records

Standard-schema records assigned to this round in the indexed snapshot.

No per-round file is listed. Check the run-record index for the complete partition list.

Reproduction inventory

Published analysis replay and full experiment replay are reported separately.

Replay type
analysis-only
Script
No per-round analysis script is published.
Inputs
Full execution replay
Unavailable
Scope
Published Standard records and the Lab public exports (frozen configuration, per-attempt outcomes, public test states, stage times and usage) are available; raw conversations, tool output and candidate code are not included.

View the complete per-round reproduction inventory · Repository reproduction guide