Round r70

r70

Status · INCOMPLETEReviewed 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/r70/README.md

Round 70 compared hidden-test outcomes for tasks 5–7 across Rust, Go, TypeScript/Bun, Elixir, and Gleam with Codex gpt-6-luna/max and gpt-6.1-sol/medium. The recovered observed cohort has 97 full passes among 107 scored cells; one additional cell is invalid. Status: INCOMPLETE (study); all 108 observed cells now have per-cell outcomes, but the full planned task set was not admitted. No stack-level conclusion was reached.

Round date: HISTORICAL (before 2026-10-09) Publication badge: INCOMPLETE; KEPT FOR AUDIT Why not VALID: The historical inventory retains INCOMPLETE because of INCOMPLETE_EXECUTION and NO_PREREG. See the round evidence and limitations below. Recomputation status: OBSERVED SOURCE ONLY

Recomputation update: The mined raw records now provide per-cell outcome, manifest token-counter, and wall-time recomputation for the observed cohort. This does not reconstruct missing planned cells, a full intention-to-treat denominator, or model execution.

Recovered per-cell results

The table below keeps model/effort arms separate. cells includes the one invalid outcome; full passes count official complete passes. The detailed task, repetition, test-count, timeout, and gate-fail rows follow below; the machine-readable rows are in cells.jsonl.

Stack Codex gpt-6-luna / max Codex gpt-6.1-sol / medium
Rust 11/11 passes (0 fail, 0 invalid) 10/11 passes (0 fail, 1 invalid)
Go 10/11 passes (1 fail, 0 invalid) 11/11 passes (0 fail, 0 invalid)
TypeScript/Bun 9/9 passes (0 fail, 0 invalid) 9/9 passes (0 fail, 0 invalid)
Elixir 10/12 passes (2 fail, 0 invalid) 8/12 passes (4 fail, 0 invalid)
Gleam 9/11 passes (2 fail, 0 invalid) 10/11 passes (1 fail, 0 invalid)

The original decision rule required an equal-task pooled intention-to-treat result at least 10 percentage points above every other stack; a tie within 10 points would use median wall time and then tokens, with a declared top-up to seven repetitions for tied stacks. The observed results do not meet this condition. The prespecified tie procedure and seven-repetition top-up did not produce a qualifying stack-level conclusion. The full seven-task plan remains incomplete; this is an interim study with no cross-stack conclusion.

The 138 grading records cover 108 run IDs; we kept the previously recorded outcome and test counts for each ID, choosing the latest matching timestamp; where records conflict, that is stated, so these totals describe the kept observations. Four IDs had conflicting grades: task 5 Rust/Luna rep 1 (23/24 FAIL or 24/24 PASS), task 6 Rust/Luna rep 2 (23/24 FAIL or 24/24 PASS), task 7 Rust/Luna rep 1 (0/24 FAIL, 23/24 FAIL, or 24/24 PASS), and task 7 Rust/Sol rep 1 (0/24 FAIL or 24/24 PASS). For all four, the kept record was 24/24 PASS; the latest matching timestamps were 2026-10-05 23:51:34Z, 2026-10-06 00:26:13Z, 2026-10-08 18:10:29Z, and 2026-10-05 23:53:47Z, respectively. This gives 97 PASS, 10 FAIL, and 1 INVALID kept observations.

Per-cell outcomes

Each row is joined by its exact cell ID to the recomputed observed cohort.

Cell ID Task Stack Model / effort Rep Outcome Tests passed / total Timeout Gate fail
codex__gpt-6-luna__max__default__r70-5-elixir__r1 r70-5-elixir Elixir gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-elixir__r2 r70-5-elixir Elixir gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-elixir__r3 r70-5-elixir Elixir gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-elixir__r1 r70-5-elixir Elixir gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-elixir__r2 r70-5-elixir Elixir gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-elixir__r3 r70-5-elixir Elixir gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-gleam__r1 r70-5-gleam Gleam gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-gleam__r2 r70-5-gleam Gleam gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-gleam__r3 r70-5-gleam Gleam gpt-6-luna / max 3 FAIL 23/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-gleam__r1 r70-5-gleam Gleam gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-gleam__r2 r70-5-gleam Gleam gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-gleam__r3 r70-5-gleam Gleam gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-go__r1 r70-5-go Go gpt-6-luna / max 1 FAIL 23/24 no yes
codex__gpt-6-luna__max__default__r70-5-go__r2 r70-5-go Go gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-go__r3 r70-5-go Go gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-go__r4 r70-5-go Go gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-go__r1 r70-5-go Go gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-go__r2 r70-5-go Go gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-go__r3 r70-5-go Go gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-go__r4 r70-5-go Go gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-rust__r1 r70-5-rust Rust gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-rust__r2 r70-5-rust Rust gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-rust__r3 r70-5-rust Rust gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-rust__r4 r70-5-rust Rust gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-rust__r1 r70-5-rust Rust gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-rust__r2 r70-5-rust Rust gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-rust__r3 r70-5-rust Rust gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-rust__r4 r70-5-rust Rust gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 r70-5-ts-bun TypeScript/Bun gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-ts-bun__r2 r70-5-ts-bun TypeScript/Bun gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-5-ts-bun__r3 r70-5-ts-bun TypeScript/Bun gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 r70-5-ts-bun TypeScript/Bun gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r2 r70-5-ts-bun TypeScript/Bun gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r3 r70-5-ts-bun TypeScript/Bun gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-elixir__r1 r70-6-elixir Elixir gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-elixir__r2 r70-6-elixir Elixir gpt-6-luna / max 2 FAIL 23/24 no no
codex__gpt-6-luna__max__default__r70-6-elixir__r3 r70-6-elixir Elixir gpt-6-luna / max 3 FAIL 22/24 no no
codex__gpt-6-luna__max__default__r70-6-elixir__r5 r70-6-elixir Elixir gpt-6-luna / max 5 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-elixir__r1 r70-6-elixir Elixir gpt-6.1-sol / medium 1 FAIL 23/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-elixir__r2 r70-6-elixir Elixir gpt-6.1-sol / medium 2 FAIL 23/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-elixir__r3 r70-6-elixir Elixir gpt-6.1-sol / medium 3 FAIL 23/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-elixir__r5 r70-6-elixir Elixir gpt-6.1-sol / medium 5 FAIL 23/24 no no
codex__gpt-6-luna__max__default__r70-6-gleam__r1 r70-6-gleam Gleam gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-gleam__r2 r70-6-gleam Gleam gpt-6-luna / max 2 FAIL 23/24 no no
codex__gpt-6-luna__max__default__r70-6-gleam__r3 r70-6-gleam Gleam gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-gleam__r4 r70-6-gleam Gleam gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 r70-6-gleam Gleam gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-gleam__r2 r70-6-gleam Gleam gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-gleam__r3 r70-6-gleam Gleam gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-gleam__r4 r70-6-gleam Gleam gpt-6.1-sol / medium 4 FAIL 23/24 no no
codex__gpt-6-luna__max__default__r70-6-go__r1 r70-6-go Go gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-go__r2 r70-6-go Go gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-go__r3 r70-6-go Go gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-go__r4 r70-6-go Go gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-go__r1 r70-6-go Go gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-go__r2 r70-6-go Go gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-go__r3 r70-6-go Go gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-go__r4 r70-6-go Go gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-rust__r1 r70-6-rust Rust gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-rust__r2 r70-6-rust Rust gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-rust__r3 r70-6-rust Rust gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-rust__r4 r70-6-rust Rust gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-rust__r1 r70-6-rust Rust gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-rust__r2 r70-6-rust Rust gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-rust__r3 r70-6-rust Rust gpt-6.1-sol / medium 3 INVALID 0/0 no no
codex__gpt-6.1-sol__medium__default__r70-6-rust__r4 r70-6-rust Rust gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-ts-bun__r1 r70-6-ts-bun TypeScript/Bun gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-ts-bun__r2 r70-6-ts-bun TypeScript/Bun gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-6-ts-bun__r3 r70-6-ts-bun TypeScript/Bun gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-ts-bun__r1 r70-6-ts-bun TypeScript/Bun gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-ts-bun__r2 r70-6-ts-bun TypeScript/Bun gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-6-ts-bun__r3 r70-6-ts-bun TypeScript/Bun gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-elixir__r1 r70-7-elixir Elixir gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-elixir__r2 r70-7-elixir Elixir gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-elixir__r3 r70-7-elixir Elixir gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-elixir__r4 r70-7-elixir Elixir gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-elixir__r5 r70-7-elixir Elixir gpt-6-luna / max 5 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-elixir__r1 r70-7-elixir Elixir gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-elixir__r2 r70-7-elixir Elixir gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-elixir__r3 r70-7-elixir Elixir gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-elixir__r4 r70-7-elixir Elixir gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-elixir__r5 r70-7-elixir Elixir gpt-6.1-sol / medium 5 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-gleam__r1 r70-7-gleam Gleam gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-gleam__r2 r70-7-gleam Gleam gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-gleam__r3 r70-7-gleam Gleam gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-gleam__r4 r70-7-gleam Gleam gpt-6-luna / max 4 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-gleam__r1 r70-7-gleam Gleam gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-gleam__r2 r70-7-gleam Gleam gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-gleam__r3 r70-7-gleam Gleam gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-gleam__r4 r70-7-gleam Gleam gpt-6.1-sol / medium 4 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-go__r1 r70-7-go Go gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-go__r2 r70-7-go Go gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-go__r3 r70-7-go Go gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-go__r1 r70-7-go Go gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-go__r2 r70-7-go Go gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-go__r3 r70-7-go Go gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-rust__r1 r70-7-rust Rust gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-rust__r2 r70-7-rust Rust gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-rust__r3 r70-7-rust Rust gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 r70-7-rust Rust gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-rust__r2 r70-7-rust Rust gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-rust__r3 r70-7-rust Rust gpt-6.1-sol / medium 3 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 r70-7-ts-bun TypeScript/Bun gpt-6-luna / max 1 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-ts-bun__r2 r70-7-ts-bun TypeScript/Bun gpt-6-luna / max 2 PASS 24/24 no no
codex__gpt-6-luna__max__default__r70-7-ts-bun__r3 r70-7-ts-bun TypeScript/Bun gpt-6-luna / max 3 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-ts-bun__r1 r70-7-ts-bun TypeScript/Bun gpt-6.1-sol / medium 1 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-ts-bun__r2 r70-7-ts-bun TypeScript/Bun gpt-6.1-sol / medium 2 PASS 24/24 no no
codex__gpt-6.1-sol__medium__default__r70-7-ts-bun__r3 r70-7-ts-bun TypeScript/Bun gpt-6.1-sol / medium 3 PASS 24/24 no no

Rust / Sol medium / task 6 / rep 3 is invalid, no tests ran; its cause isn't documented here; it's excluded from the 107 scored cells.

Status

INCOMPLETE

  • NO_PREREG — The README records no pre-registration.
  • INCOMPLETE_EXECUTION — The full seven-task plan was not admitted across stacks and repetitions; recovered outcomes cover only the observed task-5/6/7 cells.

Required reproduction metadata

  • Kogen commit: not recorded in the cited round source.
  • Harness commit: not recorded in the cited round source as a Git commit; any adapter or runtime fingerprints are not Git commits.
  • Raw records: the full per-cell table appears above, and machine-readable result rows publish the 108 observed outcomes. A complete request and Standard run-record bundle remains unavailable; the original delivery records and every missing Standard field are documented separately.

Terminology: public round glossary.

STATUS: INCOMPLETE

Standard capture completeness: 37.8% (6 original delivery records; PASS WITH DECLARED GAPS). Measurement contract; declared gaps.

Pre-registered: no

Lifecycle: Historical core study; the recovered outcome supplement now scores the observed 108 cells, while the original planned task set remains incomplete.

Question: What hidden-suite pass outcomes were recorded across implementation stacks on the Round 70 tasks?

How and where it ran: Runs were assigned to Hetzner Linux hosts in the US and Europe; hardware details are not available in the records; the recovered grading records (dated 5–8 October 2026) were assembled on a MacBook. The surviving record does not identify a harness Git commit or kit version; the Standard capture gaps are declared in MISSING.md.

Design: 7 tasks × 5 stacks × 2 models × 5 reps = 350 planned if all tasks were admitted. The recovered supplement covers observed task-5/6/7 cells, not the full planned task, stack, and repetition set.

Decision rule: A stack-level result required an equal-task pooled ITT difference of at least 10 pp over every other stack; ties within 10 pp used median wall then tokens, with one declared top-up to 7 reps for tied stacks.

Arms: Rust, Go, TypeScript/Bun, Elixir, Gleam; Codex Luna max and Sol medium.

Planned task IDs in the surviving round note include the following task-2 and task-5 variants: r70-2-rust, r70-2-go, r70-2-ts-bun, r70-2-elixir, r70-2-gleam, r70-5-rust, r70-5-go, r70-5-ts-bun, r70-5-elixir, r70-5-gleam.

Other task IDs listed in the note:

Task reconciliation: The surviving plan lists 35 task-stack combinations. The original Standard run-record export contains five task IDs, all for task 7 (one per stack); none of the 30 task 1–6 combinations appears in that tagged export. The recovered grade supplement adds observed outcomes for task 5, 6, and 7 cell IDs without changing or reattributing those delivery records. Verdict: The frozen core study remains INTERIM and has no stack-level conclusion. The seven-task plan was not completed; only the observed task-5/6/7 cells have recovered outcomes. The original rule was not met. Official outcomes use the documented grading pipeline.

Round 70 implementation-stack extension

Post-hoc decision dated 2026-10-06. Task 1 is CONFOUNDED: the three original Elixir cells returned 24/25 and the three FE2 rerun cells returned 25/25. This is consistent with an encoding confound; the code-level mechanism is not verified in this snapshot. Task 4 remains a separate, source-reported exit-code issue whose code-level explanation is unverified. Public cell-level outcomes, controls, and test counts are linked from r70-rve-rerun.

Task Status Confound
1 CONFOUNDED Original Elixir 0/3 at 24/25 each; FE2 rerun 3/3 at 25/25. Consistent with an encoding confound; the code-level mechanism is unverified.
4 CONFOUNDED Exit-code handling remains a source-reported issue; its code-level explanation is unverified. Public outcomes and admission controls are linked from the rerun ledger.

Original 40-cell RvE language comparison

These are the official as-graded observations from tasks 1, 3, 4, and 6. The measure is hidden-suite full pass; each stack's own checks are secondary. The comparison used one model, direct Codex with gpt-6-luna at max. The original n is 3 on tasks 1 and 4 and 2 on tasks 3 and 6 for each stack. Rust and Elixir ran in Europe; Go and TypeScript/Bun ran in the US. Wall time is comparable only within a host. Tasks 1 and 4 are confounded by skeleton front-end traps; repetition selection was partly strategic. This table is descriptive; no general language ranking follows. It does not change the frozen r70 core verdict.

The supplemental receipts record the requested model and effort for the 44 rerun and extension cells, but do not preserve per-cell effective model/effort receipts or a verified harness version. The original Rust and Elixir resource records do preserve effective values and the Codex CLI version; the original Go and TypeScript/Bun grades do not have matching resource records in this snapshot.

The summary and 40 per-cell rows below are recomputed from the public official test-count ledger and own-check ledger. Own checks remain separate from the hidden-suite outcome.

Stack Official full passes n Per-task n (1 / 3 / 4 / 6) Own checks passing Region
Rust 8/10 10 3 / 2 / 3 / 2 10/10 Europe
Go 7/10 10 3 / 2 / 3 / 2 9/10 US
TypeScript/Bun 7/10 10 3 / 2 / 3 / 2 10/10 US
Elixir 2/10 10 3 / 2 / 3 / 2 7/10 Europe

Per-cell as-graded observations:

Task Stack Rep Region Hidden tests passed/total Official result Own check (secondary)
1 Rust 6 Europe 25/25 pass pass
1 Rust 12 Europe 25/25 pass pass
1 Rust 15 Europe 25/25 pass pass
1 Go 16 US 24/25 fail pass
1 Go 22 US 25/25 pass pass
1 Go 25 US 25/25 pass pass
1 TypeScript/Bun 16 US 25/25 pass pass
1 TypeScript/Bun 22 US 25/25 pass pass
1 TypeScript/Bun 25 US 25/25 pass pass
1 Elixir 6 Europe 24/25 fail pass
1 Elixir 12 Europe 24/25 fail pass
1 Elixir 15 Europe 24/25 fail pass
3 Rust 7 Europe 28/28 pass pass
3 Rust 14 Europe 28/28 pass pass
3 Go 17 US 28/28 pass pass
3 Go 24 US 28/28 pass pass
3 TypeScript/Bun 17 US 28/28 pass pass
3 TypeScript/Bun 24 US 28/28 pass pass
3 Elixir 7 Europe 27/28 fail pass
3 Elixir 14 Europe 28/28 pass pass
4 Rust 9 Europe 18/18 pass pass
4 Rust 10 Europe 16/18 fail pass
4 Rust 13 Europe 16/18 fail pass
4 Go 19 US 18/18 pass fail
4 Go 20 US 17/18 fail pass
4 Go 23 US 16/18 fail pass
4 TypeScript/Bun 19 US 15/18 fail pass
4 TypeScript/Bun 20 US 16/18 fail pass
4 TypeScript/Bun 23 US 16/18 fail pass
4 Elixir 9 Europe 15/18 fail fail
4 Elixir 10 Europe 15/18 fail pass
4 Elixir 13 Europe 17/18 fail pass
6 Rust 8 Europe 24/24 pass pass
6 Rust 11 Europe 24/24 pass pass
6 Go 18 US 24/24 pass pass
6 Go 21 US 24/24 pass pass
6 TypeScript/Bun 18 US 24/24 pass pass
6 TypeScript/Bun 21 US 24/24 pass pass
6 Elixir 8 Europe 22/24 fail fail
6 Elixir 11 Europe 24/24 pass fail

Original RvE cost and usage

Twenty original Rust and Elixir cells have versioned source-reported resource receipts. The cost is a Standard API equivalent, not invoice spend. Reasoning tokens are a subset of output and are not added a second time. The price-table and calculator hashes, counter definitions, and coverage limits are in the cost-time metadata; per-cell values are in the cost-time ledger.

The five ungraded deliveries caused by a pre-grade selector fault are listed separately and excluded from scored medians. Cost source metadata for these five deliveries is incomplete; no cost or usage is inferred for Go or TypeScript/Bun original cells without receipts.

The 20 scored records do not preserve a cache-write token counter; cache_write_tokens is null, not zero. The displayed token medians and totals sum the recorded uncached-input, cached-input, and output components. Reasoning is included within output and is not added a second time.

Scored original RvE resource receipts:

Stack n Cost USD median [min–max] Median uncached / cached / output / reasoning tokens
Rust 10 0.0248 [0.0119–0.0678] 62,285 / 549,376 / 26,216 / 18,892
Elixir 10 0.0395 [0.0136–0.0844] 77,724 / 1,255,680 / 38,669 / 29,096

Five ungraded deliveries excluded before scoring (individual resource spend only):

Delivery Stack Cost USD Wall s Uncached / cached / cache-write / output / reasoning tokens
excluded-01 Elixir 0.076200 1383.125 126,817 / 3,205,376 / 0 / 62,929 / 42,616
excluded-02 Elixir 0.046885 1040.836 89,769 / 1,437,696 / 0 / 47,063 / 32,930
excluded-03 Rust 0.043027 740.365 82,601 / 1,367,296 / 0 / 42,188 / 30,421
excluded-04 Rust 0.040104 668.040 112,373 / 1,283,584 / 0 / 32,061 / 20,474
excluded-05 Rust 0.050431 935.774 93,081 / 1,834,496 / 0 / 45,555 / 26,596

Scored plus excluded all-attempt totals by stack:

Stack Scored cost USD total Excluded deliveries Excluded cost USD total All-attempt cost USD total All-attempt tokens
Rust 0.275191 3 0.133561 0.408752 12,616,424
Elixir 0.448316 2 0.123085 0.571402 23,079,495

Existing r70 agent-wall attribution

This read-only transcript snapshot covers 108 timed original r70 cells. Command classification is approximate; overlapping timestamped intervals are deduplicated and merged, and mixed shell commands contribute their full elapsed interval. Residual time is the remainder after command intervals and includes calls, waits, editing, and uninstrumented overhead; it is not model latency. The sample was interrupted and is not balanced; no causal stack comparison follows from this sample.

Stack Timed cells n Build/check/test wall share, median [Q1–Q3] Residual share, median [Q1–Q3]
Rust 22 1.4% [1.1%–5.0%] 98.6% [95.0%–98.9%]
Go 22 10.6% [6.1%–14.8%] 89.4% [85.2%–93.9%]
TypeScript/Bun 18 0.5% [0.3%–0.7%] 99.5% [99.3%–99.7%]
Elixir 24 40.0% [21.4%–49.1%] 59.9% [50.9%–78.5%]
Gleam 22 3.5% [1.2%–5.4%] 96.5% [94.5%–98.8%]

The public cell-level test-count and control ledgers publish the original and rerun cell IDs, official outcomes, hidden-test counts, runner status, grade timestamps, and available control patch hashes. The task-4 control failures remain a limitation on causal interpretation; the recorded task-4 outcomes remain descriptive. See RESULTS.md and RERUN-RESULTS.md.

Separate compile-only timing from full native checks and hidden-wrapper timing. The latter uses a distinct n=3 diagnostic, includes all five stacks on tasks 1, 4, and 6, and is reported in the compile diagnostic.

Public delivery and outcome reconciliation

This captured-delivery summary uses the public run-record export, grouped by audit round, task ID, assigned region, and the exact exported arm label. Captured counts rows; graded counts rows marked graded=true; pass, fail, and other are the run-record outcome fields. Official outcome states are in cells.jsonl, whose classifications can differ; see the register source crosswalk. Planned n is shown only when source-reported; unknown allocations are not inferred.

Task ID Assigned region Arm label in export Planned n Captured Graded Pass Fail Other graded outcome
r70-7-elixir Europe not recorded 5 1 0 0 0 0
r70-7-gleam Europe not recorded 5 1 0 0 0 0
r70-7-go Europe not recorded 5 1 0 0 0 0
r70-7-rust Europe not recorded 5 2 0 0 0 0
r70-7-ts-bun Europe not recorded 5 1 0 0 0 0

Reproduction record and source reconciliation

This section documents exact replay limits and keeps the page lifecycle and earlier statuses unchanged.

  • Kogen source commit: not applicable (Harness does not use Kogen) in the per-cell setup.kogen_sha field.
  • Harness source commit: not recorded. Per-cell tools.harness SHA-256 fingerprint(s): 08019d8aab7c15481f6d14126645eebd7b2c2c98025f94118baeee2167e4a300. Only wrapper fingerprints are public; the underlying source commit remains unrecorded.
  • Models and effort (requested → effective): codex: model requested/effective gpt-6-luna → gpt-6-luna; effort requested/effective max → max; codex: model requested/effective gpt-6.1-sol → gpt-6.1-sol; effort requested/effective medium → medium.
  • Observed task IDs: r70-7-elixir, r70-7-gleam, r70-7-go, r70-7-rust, r70-7-ts-bun.
  • Original execution command: not recorded in the public snapshot; the benchmark cells cannot be replayed from this repository.
  • Rebuild the public metadata from the repository root with python3 reproduce/export_results.py and python3 reproduce/build_records.py; check this declaration with python3 reproduce/validate_round.py --round r70. These commands regenerate or validate the archived public records. An original cell-launch command is not available.
  • Raw public records: r70.jsonl filtered by audit_round/round_id, and cells.jsonl filtered by round. Separate RvE evidence is in test-counts.jsonl, controls.jsonl and cost-time.jsonl.

Public snapshot rows for this tag: 6 captured deliveries (0 pass, 0 fail, 6 ungraded/unknown in results/run-records/r70.jsonl); official outcome export has 0 rows: 0 pass, 0 fail in cells.jsonl. These are separate source totals. The public files do not provide a complete joined intention-to-treat result.

Round-specific reconciliation: The six public records tagged r70 are ungraded delivery records for the frozen stack study. Keep them distinct from the original RvE, rerun and extension cohorts represented by separate supplemental ledgers/pages; their results do not close the frozen stack study.

Recomputed from raw records

`recomputed.json`, mined cell records, and the token audit back observed official outcomes, per-arm pass counts/rates, manifest token counters, and wall medians. The source contains 108 rows (fail 10, invalid 1, pass 97); overall pass rate is 90.7% (97/107) among pass/fail grades. Per-arm detail and coverage counts are in the JSON.

No shared cell IDs are present in the committed public grade export; the source-only cohort is checked against the page's reported figures below. Recovered-source discrepancies: 128 field mismatches across 10 cell IDs (experiment 8, official_grade.classification 2, official_grade.outcome 2, official_grade.pass_fail 2, official_grade.test_counts.tests_passed 3, patch.archive_member.28a64f305781e409eacffde8d072785d2df2a5d21bec421d5f32901ad7df1d99 1, patch.archive_member.362f057f326640c2cf4e374e7d802f66271114bc23e42dfe5de90f0581feb710 1, patch.archive_member.7174c706f2b613835fb9bfedbebd2a3a358b1dbe637ad24a25d153bafd8c03ff 1, patch.archive_member.7f6d66e03155eeae947915388a3af6087d50525d3cc67b38711bd9cf9d791a83 1, patch.archive_member.b1673af73621115abf6e4cb46daaa2c461a203d823d51b7f086a39336c35786f 1, patch.archive_member.bab0ed96a55291698e619a5e55275e34b570e568df73e337e597a6c768d5c9d2 1, patch.archive_member.bd7258cea5afcc7daf7221d6b06a6be9b51482db06102acaeec64dd116860be2 1, patch.archive_member.c2409a3799e9d40b28836b76da0e4a2655eaf6f6aef2e34df3ed684e1390e5f1 1, patch.archive_member.d537fd80ccec570bf0614ef61004850747bc04d4eb5e7eecdd11ba75af6388db 1, patch.attempt-1 12, usage.cached_input_tokens 18, usage.input_tokens 18, usage.output_tokens 18, usage.reasoning_tokens 18, wall_s 18). Existing values were preserved; each cell, field, kept value, and recovered value is listed below.

  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field patch.attempt-1: kept {"sha256": "3a65028ee36b6baf4dd9b6af3263a2558d243ff324aa69ddad51aa64faef238a", "size_bytes": 8290}, recovered {"sha256": "f90736d8939c7a56a738fb785fd64ad64e928ca6cb4600ee0be44f5384d71ac3", "size_bytes": 10078}.
  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field usage.cached_input_tokens: kept 320000, recovered 638208.
  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field usage.input_tokens: kept 47517, recovered 61921.
  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field usage.output_tokens: kept 21832, recovered 22967.
  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field usage.reasoning_tokens: kept 16951, recovered 16432.
  • codex__gpt-6-luna__max__default__r70-5-ts-bun__r1 field wall_s: kept 421.595, recovered 440.637.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field patch.archive_member.c2409a3799e9d40b28836b76da0e4a2655eaf6f6aef2e34df3ed684e1390e5f1: kept {"archive_member": null, "sha256": "c2409a3799e9d40b28836b76da0e4a2655eaf6f6aef2e34df3ed684e1390e5f1", "size_bytes": 11918}, recovered null.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field patch.attempt-1: kept {"sha256": "b5579bf42680c1a175d6efd9ff956e94650b19f91eaf5db5328ff5e66c547806", "size_bytes": 11975}, recovered {"sha256": "c2409a3799e9d40b28836b76da0e4a2655eaf6f6aef2e34df3ed684e1390e5f1", "size_bytes": 11918}.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field usage.cached_input_tokens: kept 2047488, recovered 1401088.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field usage.input_tokens: kept 101265, recovered 86918.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field usage.output_tokens: kept 33628, recovered 28178.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field usage.reasoning_tokens: kept 22378, recovered 19441.
  • codex__gpt-6-luna__max__default__r70-6-gleam__r1 field wall_s: kept 670.51, recovered 596.572.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field experiment: kept "r70-eu-7-elixir-gpt-6-luna-r1", recovered "r70-eu-7-elixir-gpt-6-luna-r1-smoke".
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field patch.attempt-1: kept {"sha256": "9f98c8fc866e9039c8744fff8c183cae19304297e7abd03a235ff413c3173aaa", "size_bytes": 12146}, recovered {"sha256": "cd3a0a27e7e36ae70f80e079c579415f8fbd2301d41dfcb12d27d3b742e93cd8", "size_bytes": 11188}.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.cached_input_tokens: kept 1113856, recovered 261632.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.cached_input_tokens: kept 1113856, recovered 261632.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.input_tokens: kept 73534, recovered 45022.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.input_tokens: kept 73534, recovered 45022.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.output_tokens: kept 36752, recovered 18580.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.output_tokens: kept 36752, recovered 18580.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.reasoning_tokens: kept 29232, recovered 13674.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field usage.reasoning_tokens: kept 29232, recovered 13674.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field wall_s: kept 807.354, recovered 362.553.
  • codex__gpt-6-luna__max__default__r70-7-elixir__r1 field wall_s: kept 807.354, recovered 362.553.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field experiment: kept "r70-eu-7-gleam-gpt-6-luna-r1-smoke", recovered "r70-eu-7-gleam-gpt-6-luna-r1".
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field patch.archive_member.b1673af73621115abf6e4cb46daaa2c461a203d823d51b7f086a39336c35786f: kept {"archive_member": null, "sha256": "b1673af73621115abf6e4cb46daaa2c461a203d823d51b7f086a39336c35786f", "size_bytes": 14866}, recovered null.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field patch.attempt-1: kept {"sha256": "048604acf6560276fedcce16407b7d14a073e5cece7094b178cd9afca1e688d0", "size_bytes": 12718}, recovered {"sha256": "b1673af73621115abf6e4cb46daaa2c461a203d823d51b7f086a39336c35786f", "size_bytes": 14866}.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.cached_input_tokens: kept 996864, recovered 1592064.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.cached_input_tokens: kept 996864, recovered 1592064.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.input_tokens: kept 89289, recovered 79354.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.input_tokens: kept 89289, recovered 79354.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.output_tokens: kept 23068, recovered 30635.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.output_tokens: kept 23068, recovered 30635.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.reasoning_tokens: kept 14066, recovered 19313.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field usage.reasoning_tokens: kept 14066, recovered 19313.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field wall_s: kept 1003.39, recovered 604.46.
  • codex__gpt-6-luna__max__default__r70-7-gleam__r1 field wall_s: kept 1003.39, recovered 604.46.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field experiment: kept "r70-us-7-go-gpt-6-luna-r1", recovered "r70-eu-7-go-gpt-6-luna-r1-smoke".
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field patch.archive_member.bab0ed96a55291698e619a5e55275e34b570e568df73e337e597a6c768d5c9d2: kept {"archive_member": null, "sha256": "bab0ed96a55291698e619a5e55275e34b570e568df73e337e597a6c768d5c9d2", "size_bytes": 8659}, recovered null.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field patch.attempt-1: kept {"sha256": "c5cdc19fb4440b738a66a4cd5469dad16c8193d92a5a0445d346e79c3c120b5b", "size_bytes": 9193}, recovered {"sha256": "bab0ed96a55291698e619a5e55275e34b570e568df73e337e597a6c768d5c9d2", "size_bytes": 8659}.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.cached_input_tokens: kept 222976, recovered 150016.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.cached_input_tokens: kept 222976, recovered 150016.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.input_tokens: kept 31137, recovered 31055.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.input_tokens: kept 31137, recovered 31055.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.output_tokens: kept 13869, recovered 10896.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.output_tokens: kept 13869, recovered 10896.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.reasoning_tokens: kept 9129, recovered 6570.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field usage.reasoning_tokens: kept 9129, recovered 6570.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field wall_s: kept 280.945, recovered 311.115.
  • codex__gpt-6-luna__max__default__r70-7-go__r1 field wall_s: kept 280.945, recovered 311.115.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field experiment: kept "r70-eu-7-rust-gpt-6-luna-r1", recovered "r70-eu-7-rust-gpt-6-luna-r1-smoke".
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field official_grade.classification: kept "pass", recovered "fail".
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field official_grade.outcome: kept "pass", recovered "fail".
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field official_grade.pass_fail: kept "pass", recovered "fail".
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field official_grade.test_counts.tests_passed: kept 24, recovered 0.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field official_grade.test_counts.tests_passed: kept 24, recovered 23.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field patch.archive_member.d537fd80ccec570bf0614ef61004850747bc04d4eb5e7eecdd11ba75af6388db: kept {"archive_member": null, "sha256": "d537fd80ccec570bf0614ef61004850747bc04d4eb5e7eecdd11ba75af6388db", "size_bytes": 11250}, recovered null.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field patch.attempt-1: kept {"sha256": "abd6d6e16050e9533940b0d5d8c57b043119cbdccd98bcdeb31337c063c2f361", "size_bytes": 15986}, recovered {"sha256": "d537fd80ccec570bf0614ef61004850747bc04d4eb5e7eecdd11ba75af6388db", "size_bytes": 11250}.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.cached_input_tokens: kept 476416, recovered 255488.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.cached_input_tokens: kept 476416, recovered 255488.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.input_tokens: kept 47929, recovered 62021.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.input_tokens: kept 47929, recovered 62021.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.output_tokens: kept 18970, recovered 12636.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.output_tokens: kept 18970, recovered 12636.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.reasoning_tokens: kept 10517, recovered 7641.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field usage.reasoning_tokens: kept 10517, recovered 7641.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field wall_s: kept 161.452, recovered 288.762.
  • codex__gpt-6-luna__max__default__r70-7-rust__r1 field wall_s: kept 161.452, recovered 288.762.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field experiment: kept "r70-us-7-ts-bun-gpt-6-luna-r1", recovered "r70-eu-7-ts-bun-gpt-6-luna-r1-smoke".
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field experiment: kept "r70-us-7-ts-bun-gpt-6-luna-r1", recovered "r70-us-7-ts-bun-gpt-6-luna-r1-fairness-v2-smoke".
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field experiment: kept "r70-us-7-ts-bun-gpt-6-luna-r1-fairness-v2-smoke", recovered "r70-eu-7-ts-bun-gpt-6-luna-r1-smoke".
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field patch.archive_member.362f057f326640c2cf4e374e7d802f66271114bc23e42dfe5de90f0581feb710: kept {"archive_member": null, "sha256": "362f057f326640c2cf4e374e7d802f66271114bc23e42dfe5de90f0581feb710", "size_bytes": 8844}, recovered null.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field patch.archive_member.bd7258cea5afcc7daf7221d6b06a6be9b51482db06102acaeec64dd116860be2: kept {"archive_member": null, "sha256": "bd7258cea5afcc7daf7221d6b06a6be9b51482db06102acaeec64dd116860be2", "size_bytes": 7289}, recovered null.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field patch.attempt-1: kept {"sha256": "11dbf0173ea66a75465e3e2da52171c4f44b410dd90ef456f567327e4491281e", "size_bytes": 7952}, recovered {"sha256": "362f057f326640c2cf4e374e7d802f66271114bc23e42dfe5de90f0581feb710", "size_bytes": 8844}.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field patch.attempt-1: kept {"sha256": "11dbf0173ea66a75465e3e2da52171c4f44b410dd90ef456f567327e4491281e", "size_bytes": 7952}, recovered {"sha256": "bd7258cea5afcc7daf7221d6b06a6be9b51482db06102acaeec64dd116860be2", "size_bytes": 7289}.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field patch.attempt-1: kept {"sha256": "362f057f326640c2cf4e374e7d802f66271114bc23e42dfe5de90f0581feb710", "size_bytes": 8844}, recovered {"sha256": "bd7258cea5afcc7daf7221d6b06a6be9b51482db06102acaeec64dd116860be2", "size_bytes": 7289}.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.cached_input_tokens: kept 202752, recovered 181504.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.cached_input_tokens: kept 202752, recovered 570368.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.cached_input_tokens: kept 202752, recovered 570368.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.cached_input_tokens: kept 570368, recovered 181504.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.input_tokens: kept 23746, recovered 26751.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.input_tokens: kept 23746, recovered 67777.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.input_tokens: kept 23746, recovered 67777.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.input_tokens: kept 67777, recovered 26751.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.output_tokens: kept 11517, recovered 11772.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.output_tokens: kept 11517, recovered 20234.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.output_tokens: kept 11517, recovered 20234.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.output_tokens: kept 20234, recovered 11772.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.reasoning_tokens: kept 14359, recovered 8622.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.reasoning_tokens: kept 7958, recovered 14359.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.reasoning_tokens: kept 7958, recovered 14359.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field usage.reasoning_tokens: kept 7958, recovered 8622.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field wall_s: kept 226.325, recovered 320.028.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field wall_s: kept 226.325, recovered 395.616.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field wall_s: kept 226.325, recovered 395.616.
  • codex__gpt-6-luna__max__default__r70-7-ts-bun__r1 field wall_s: kept 395.616, recovered 320.028.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field patch.archive_member.28a64f305781e409eacffde8d072785d2df2a5d21bec421d5f32901ad7df1d99: kept {"archive_member": null, "sha256": "28a64f305781e409eacffde8d072785d2df2a5d21bec421d5f32901ad7df1d99", "size_bytes": 9583}, recovered null.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field patch.attempt-1: kept {"sha256": "5fde2b92d1353f40683bc12cf698951bf9b93124c1f4bfcc4fa1b87699369e75", "size_bytes": 10944}, recovered {"sha256": "28a64f305781e409eacffde8d072785d2df2a5d21bec421d5f32901ad7df1d99", "size_bytes": 9583}.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field usage.cached_input_tokens: kept 127104, recovered 87552.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field usage.input_tokens: kept 13622, recovered 11611.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field usage.output_tokens: kept 4983, recovered 4674.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field usage.reasoning_tokens: kept 212, recovered 349.
  • codex__gpt-6.1-sol__medium__default__r70-5-ts-bun__r1 field wall_s: kept 118.009, recovered 188.683.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field patch.archive_member.7f6d66e03155eeae947915388a3af6087d50525d3cc67b38711bd9cf9d791a83: kept {"archive_member": null, "sha256": "7f6d66e03155eeae947915388a3af6087d50525d3cc67b38711bd9cf9d791a83", "size_bytes": 16173}, recovered null.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field patch.attempt-1: kept {"sha256": "cf49e8027b45756b5456759ae6c0912d60a5a2d43af828d02432cd61c70cd628", "size_bytes": 10233}, recovered {"sha256": "7f6d66e03155eeae947915388a3af6087d50525d3cc67b38711bd9cf9d791a83", "size_bytes": 16173}.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field usage.cached_input_tokens: kept 186624, recovered 149632.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field usage.input_tokens: kept 17440, recovered 30284.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field usage.output_tokens: kept 5447, recovered 7374.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field usage.reasoning_tokens: kept 447, recovered 865.
  • codex__gpt-6.1-sol__medium__default__r70-6-gleam__r1 field wall_s: kept 141.211, recovered 305.009.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field experiment: kept "r70-eu-7-rust-gpt-6.1-sol-r1-smoke", recovered "r70-eu-7-rust-gpt-6.1-sol-r1".
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field official_grade.classification: kept "pass", recovered "fail".
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field official_grade.outcome: kept "pass", recovered "fail".
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field official_grade.pass_fail: kept "pass", recovered "fail".
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field official_grade.test_counts.tests_passed: kept 24, recovered 0.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field patch.archive_member.7174c706f2b613835fb9bfedbebd2a3a358b1dbe637ad24a25d153bafd8c03ff: kept {"archive_member": null, "sha256": "7174c706f2b613835fb9bfedbebd2a3a358b1dbe637ad24a25d153bafd8c03ff", "size_bytes": 15970}, recovered null.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field patch.attempt-1: kept {"sha256": "3cdd6eee05d96a567d58da932ee90ba8ec3628996ed8f7e5d43810e7152abb64", "size_bytes": 16382}, recovered {"sha256": "7174c706f2b613835fb9bfedbebd2a3a358b1dbe637ad24a25d153bafd8c03ff", "size_bytes": 15970}.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.cached_input_tokens: kept 65280, recovered 83712.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.cached_input_tokens: kept 65280, recovered 83712.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.input_tokens: kept 14982, recovered 14074.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.input_tokens: kept 14982, recovered 14074.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.output_tokens: kept 5260, recovered 4876.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.output_tokens: kept 5260, recovered 4876.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.reasoning_tokens: kept 183, recovered 49.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field usage.reasoning_tokens: kept 183, recovered 49.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field wall_s: kept 207.218, recovered 111.25.
  • codex__gpt-6.1-sol__medium__default__r70-7-rust__r1 field wall_s: kept 207.218, recovered 111.25. No shared public-export cell IDs are available for a paired outcome comparison. No complete published metadata/manifest token pairs are available for this round (0 missing manifest usage; 108 missing published metadata). Missing counters remain unknown, not zero. These 108 outcomes are now public in rounds/r70/cells.jsonl but remain separate from the repository-wide outcome export.

Captured run records

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

6 records · 34,834 bytesDownload r70.jsonl · View immutable source ↗

Reproduction inventory

Published analysis replay and full experiment replay are reported separately.

Replay type
unavailable
Script
No per-round analysis script is published.
Inputs
No identified public input bundle.
Full execution replay
Unavailable
Scope
No public per-round analysis script with a complete, identified input bundle is published. See the round README, MEASURED.md, and MISSING.md for recorded scope and gaps; historical execution inputs/toolchains/grading are not bundled.

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