{
  "boundaries": [
    "Protocol classes are disconnected and cannot be blended.",
    "Legacy strict, bounded, and frontier records are audit-only; historical top1 did not prove strict N=1.",
    "The outcome-ranked historical atlas is not a prospective win-rate study.",
    "Focal cases are post-hoc descriptive records and carry no causal attribution.",
    "Foundation comparator wall-clock was not receipted; candidate timing is not a measured foundation-training speedup.",
    "The billion-scale value is a forward-looking company estimate, unobserved, and not a finding of this study."
  ],
  "claims": [
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "strict",
      "company_reported": true,
      "id": "strict-elo",
      "label": "Strict first move",
      "limitations": [
        "Applies only to the sealed field and disclosed strict protocol."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "One frozen first executable artifact, before task-time search.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/strict_first_move/elo",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "1108.8 Elo"
    },
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "strict",
      "company_reported": true,
      "id": "strict-rank",
      "label": "Strict field rank",
      "limitations": [
        "Rank depends on this field, evaluator, and frozen artifact."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Position in the sealed field under the strict protocol.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/strict_first_move/field_rank",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "14/98"
    },
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "bounded",
      "company_reported": true,
      "id": "bounded-elo",
      "label": "Bounded probe",
      "limitations": [
        "Cannot be substituted for the strict result."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Bounded task-time selection, not a strict first move.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/bounded_probe/elo",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "1212.2 Elo"
    },
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "bounded",
      "company_reported": true,
      "id": "bounded-rank",
      "label": "Bounded field rank",
      "limitations": [
        "Rank depends on the bounded protocol and disclosed field."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Position in the sealed field under bounded selection.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/bounded_probe/field_rank",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "7/98"
    },
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "frontier",
      "company_reported": true,
      "id": "frontier-elo",
      "label": "Frontier diagnostic",
      "limitations": [
        "A diagnostic upper boundary, separate from strict and bounded modes."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Diagnostic retained frontier; not a deployable first move.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/frontier_diagnostic/elo",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "1331.1 Elo"
    },
    {
      "active_paper_boundary": "Historical top1 did not prove strict N=1; a prospective frozen one-proposal study is still required.",
      "admission_status": "audit_only",
      "class": "frontier",
      "company_reported": true,
      "id": "frontier-rank",
      "label": "Frontier field rank",
      "limitations": [
        "Rank does not describe one frozen first artifact."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Diagnostic position kept separate from deployable modes.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "sealed_blind_ledger.json",
        "pointer": "/frontier_diagnostic/field_rank",
        "sha256": "b427cd5e796243ffd0b4bd8bded2962bc0564b58652ebe35ad49aa5e19383828"
      },
      "value": "6/98"
    },
    {
      "class": "atlas",
      "company_reported": true,
      "id": "atlas-cases",
      "label": "Outcome-ranked atlas",
      "limitations": [
        "Historical and outcome-ranked; not a prospective win-rate study."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "All recorded wins plus outcome-ranked losses in the released discovery set.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "atlas_analysis.json",
        "pointer": "/case_count",
        "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
      },
      "value": "600 cases"
    },
    {
      "class": "historical",
      "company_reported": true,
      "id": "cohort-cases",
      "label": "Full matched cohort",
      "limitations": [
        "Historical cohort; no prospective sampling protocol was declared."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Complete eligible matched historical record before atlas selection.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "build_manifest.json",
        "pointer": "/expected_locked_counts/eligible_rows",
        "sha256": "e5e5f695cce8ebd39467bfc1819838c2b0433ec3f37ac190ccef9f618a59175b"
      },
      "value": "1,649 cases"
    },
    {
      "class": "campaign",
      "company_reported": true,
      "id": "campaign-generations",
      "label": "Training campaign",
      "limitations": [
        "The raw top-one trajectory is not monotonic."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Recorded self-play campaign length.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "campaign_ledger.json",
        "pointer": "/generations",
        "sha256": "7f34b6f6b695bc70e068b7bbb4d7ed563ebf22c12edb7c7983eb05c774e3349b"
      },
      "value": "36 generations"
    },
    {
      "class": "campaign",
      "company_reported": true,
      "id": "campaign-promotions",
      "label": "Validated promotions",
      "limitations": [
        "A promotion count is not a causal attribution."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Retained-frontier promotions in the recorded campaign.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "campaign_ledger.json",
        "pointer": "/promotions",
        "sha256": "7f34b6f6b695bc70e068b7bbb4d7ed563ebf22c12edb7c7983eb05c774e3349b"
      },
      "value": "7 promotions"
    },
    {
      "class": "campaign",
      "company_reported": true,
      "id": "campaign-rows",
      "label": "Processed rows",
      "limitations": [
        "A work total is not an independent evaluation result."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Summed processed rows across the recorded campaign.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "campaign_ledger.json",
        "pointer": "/processed_rows",
        "sha256": "7f34b6f6b695bc70e068b7bbb4d7ed563ebf22c12edb7c7983eb05c774e3349b"
      },
      "value": "375,522 rows"
    },
    {
      "class": "campaign",
      "company_reported": true,
      "id": "campaign-hours",
      "label": "Generation-hours",
      "limitations": [
        "Not critical-path wall time or deployment time."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Summed generation-hours across the recorded campaign.",
      "record_type": "legacy empirical claim",
      "site_version": "0.2.0",
      "source": {
        "artifact": "campaign_claims.json",
        "pointer": "/claims/23/value",
        "sha256": "debd62e215a0bf05a1b1028f19838d9dbbb19804d77450471f15135be485df5c"
      },
      "value": "36.42 hours"
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "aggregate_selection.selected_wins",
      "id": "historical-ensemble-wins",
      "label": "Programs beating TabFM Ensemble",
      "limitations": [
        "Historical comparison under the recorded referee and comparator snapshot."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Standalone executable programs in the complete eligible historical cohort.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "aggregate_selection_receipt.json",
          "id": "aggregate-selection",
          "sha256": "5a2d5b8c54324fb70c0ff2078fcf0255e9520cf48f1cbc751458e68600e66c96"
        }
      ],
      "value": {
        "count": 353,
        "eligible_historical_denominator": 1649
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "all_four.row_count_over_denominator",
      "id": "historical-all-four-wins",
      "label": "Programs beating all four foundations",
      "limitations": [
        "The joint count is reported over the eligible historical denominator, not a selected-row rate."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Joint wins against every exact foundation configuration stored for the row.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "all_four_joint_comparison_receipt.json",
          "id": "all-four",
          "sha256": "9422b5f30eb5bfc24746bef3c465e20a5bbbd47c4669032f5fda50eb1b32e86d"
        },
        {
          "artifact": "current_snapshot_receipt.json",
          "id": "snapshot",
          "sha256": "1d6a9db2356a75c870706043a1aea9e55255382555a6fbf8c5d7a1ebabd8c25e"
        }
      ],
      "value": {
        "count": 228,
        "eligible_historical_denominator": 1649
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "atlas.outcome_counts",
      "id": "atlas-composition",
      "label": "Released atlas composition",
      "limitations": [
        "Outcome-ranked historical discovery set; not a prospective win-rate study."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "All recorded wins plus outcome-ranked losses in the released interpretive set.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "atlas_analysis.json",
          "id": "atlas",
          "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
        },
        {
          "artifact": "aggregate_selection_receipt.json",
          "id": "aggregate-selection",
          "sha256": "5a2d5b8c54324fb70c0ff2078fcf0255e9520cf48f1cbc751458e68600e66c96"
        }
      ],
      "value": {
        "cases": 600,
        "recorded_losses": 247,
        "recorded_wins": 353
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "atlas.wins.fit_predict_median",
      "id": "winning-fit-median",
      "label": "Winning median fit and predict time",
      "limitations": [
        "Comparator wall-clock was not receipted; this is candidate runtime only."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Median recomputed across every recorded winning atlas row.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "atlas_analysis.json",
          "id": "atlas",
          "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
        }
      ],
      "value": {
        "seconds": 0.706549
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "atlas.wins.workflow_threshold",
      "id": "winning-workflow-threshold",
      "label": "Winning workflows within twenty seconds",
      "limitations": [
        "Comparator wall-clock was not receipted; no measured training-speedup claim is made."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Search, finalist evaluation, fit, prediction, and recorded overhead within the threshold.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "atlas_analysis.json",
          "id": "atlas",
          "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
        }
      ],
      "value": {
        "count": 228,
        "threshold_seconds": 20
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "atlas.wins.fit_predict_thresholds",
      "id": "winning-fit-thresholds",
      "label": "Winning fit and predict thresholds",
      "limitations": [
        "Candidate runtime only; hardware matching to foundation training was not receipted."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Counts recomputed from the complete released row ledger.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "atlas_analysis.json",
          "id": "atlas",
          "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
        }
      ],
      "value": {
        "lte_1_second": 239,
        "lte_2_seconds": 320
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "atlas.wins.factor_thresholds",
      "id": "winning-quality-thresholds",
      "label": "Winning RMSE-factor thresholds",
      "limitations": [
        "Historical recorded comparisons do not establish universal superiority."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Comparator RMSE divided by candidate RMSE at fixed factor thresholds.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "atlas_analysis.json",
          "id": "atlas",
          "sha256": "06a8a3f345e117e85b803989211d6e9f30da95a9914cd98fe0901d314babd120"
        }
      ],
      "value": {
        "at_least_10x": 12,
        "at_least_2x": 39,
        "at_least_5x": 18
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "case_studies.cpu",
      "id": "focal-cpu",
      "label": "CPU focal invention",
      "limitations": [
        "Post-hoc focal case with a small held-out split; no causal or frequency claim."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Exact recorded program, clocks, dimensions, strongest-foundation result, and four comparisons.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "case_studies.json",
          "id": "case-studies",
          "sha256": "130386a4c460d989b69fe411583d6d239ce3bddb73dc8ae87aafd7f9b303a8d3"
        }
      ],
      "value": {
        "all_four_foundation_comparisons": {
          "tabfm_ensemble_regressor": {
            "baseline_rmse": 0.0136233718239636,
            "candidate_rmse": 0.002653883792755252,
            "outcome": "win",
            "rmse_factor": 5.133371649939453
          },
          "tabfm_regressor": {
            "baseline_rmse": 0.01767330524645911,
            "candidate_rmse": 0.002653883792755252,
            "outcome": "win",
            "rmse_factor": 6.659411875796849
          },
          "tabpfn_v3_regressor": {
            "baseline_rmse": 0.030087212535387813,
            "candidate_rmse": 0.002653883792755252,
            "outcome": "win",
            "rmse_factor": 11.33704972972889
          },
          "tabpfn_v3_regressor_default": {
            "baseline_rmse": 0.029721850537278143,
            "candidate_rmse": 0.002653883792755252,
            "outcome": "win",
            "rmse_factor": 11.199379045312693
          }
        },
        "audit": {
          "declared_features": 7,
          "declared_test_rows": 35,
          "declared_train_rows": 174
        },
        "claim_boundary": "The ablation is post-hoc and the small held-out split does not establish causal mechanism or prospective frequency.",
        "dataset_hash": "87f1d673934c2f39",
        "genome_hash": "026a44f577eb9541291f255476d3085f5accec5707c8ae13de2370b89bdcd3d4",
        "legal_genome": "(pipe X (poly_interact append=true) (sigmoid append=true) (poly_interact append=true) (sigmoid append=false) (sigmoid) (sigmoid) (ridge))",
        "source_file": "cpu.csv",
        "strongest_foundation": {
          "candidate_rmse": 0.002653883792755252,
          "factor": 5.133371649939453,
          "identity": "tabfm_ensemble_regressor",
          "rmse": 0.0136233718239636
        },
        "timing": {
          "finalist_cv_secs": 1.183175,
          "fit_predict_secs": 0.479872,
          "search_secs": 17.578435,
          "total_secs": 19.527443
        }
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "case_studies.greenhouse",
      "id": "focal-greenhouse",
      "label": "Greenhouse focal invention",
      "limitations": [
        "One encoded table; not a climate-modeling or generalization claim."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Exact recorded program, clocks, dimensions, strongest-foundation result, and four comparisons.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "case_studies.json",
          "id": "case-studies",
          "sha256": "130386a4c460d989b69fe411583d6d239ce3bddb73dc8ae87aafd7f9b303a8d3"
        }
      ],
      "value": {
        "all_four_foundation_comparisons": {
          "tabfm_ensemble_regressor": {
            "baseline_rmse": 0.023325622628200413,
            "candidate_rmse": 0.005004426154663911,
            "outcome": "win",
            "rmse_factor": 4.660998465620665
          },
          "tabfm_regressor": {
            "baseline_rmse": 0.017019119789940106,
            "candidate_rmse": 0.005004426154663911,
            "outcome": "win",
            "rmse_factor": 3.4008134527230487
          },
          "tabpfn_v3_regressor": {
            "baseline_rmse": 0.025166207715146084,
            "candidate_rmse": 0.005004426154663911,
            "outcome": "win",
            "rmse_factor": 5.0287899026529255
          },
          "tabpfn_v3_regressor_default": {
            "baseline_rmse": 0.025286796885451792,
            "candidate_rmse": 0.005004426154663911,
            "outcome": "win",
            "rmse_factor": 5.052886405744159
          }
        },
        "audit": {
          "declared_features": 2,
          "declared_test_rows": 50,
          "declared_train_rows": 250
        },
        "claim_boundary": "This is an inductive-bias match on one encoded table, not a climate-modeling claim.",
        "dataset_hash": "174a532440eec96c",
        "genome_hash": "7626e65256423d6fc27e8cb5c0e7074fc692f7e2ec8088b9b1634b73d17ec504",
        "legal_genome": "(pipe X (gpr kernel=rbf*periodic m=300 steps=40))",
        "source_file": "rdatasets-dslabs-greenhouse-gases-concentration.csv",
        "strongest_foundation": {
          "candidate_rmse": 0.005004426154663911,
          "factor": 3.4008134527230487,
          "identity": "tabfm_regressor",
          "rmse": 0.017019119789940106
        },
        "timing": {
          "finalist_cv_secs": 1.676858,
          "fit_predict_secs": 1.223867,
          "search_secs": 15.95511,
          "total_secs": 19.614276
        }
      }
    },
    {
      "class": "discovery",
      "company_reported": true,
      "derivation": "case_studies.spectrometer",
      "id": "focal-spectrometer",
      "label": "Spectrometer quality outlier",
      "limitations": [
        "Post-hoc calibration record; no universal spectral-learning claim."
      ],
      "observed": true,
      "paper_version": "v1",
      "qualifier": "Quality outlier with exact recorded program, clocks, dimensions, and four comparisons.",
      "record_type": "derived empirical finding",
      "site_version": "0.2.0",
      "sources": [
        {
          "artifact": "case_studies.json",
          "id": "case-studies",
          "sha256": "130386a4c460d989b69fe411583d6d239ce3bddb73dc8ae87aafd7f9b303a8d3"
        }
      ],
      "value": {
        "all_four_foundation_comparisons": {
          "tabfm_ensemble_regressor": {
            "baseline_rmse": 0.01183941204578691,
            "candidate_rmse": 2.7068358298285417e-6,
            "outcome": "win",
            "rmse_factor": 4373.893649300796
          },
          "tabfm_regressor": {
            "baseline_rmse": 0.014710401633106869,
            "candidate_rmse": 2.7068358298285417e-6,
            "outcome": "win",
            "rmse_factor": 5434.537799079844
          },
          "tabpfn_v3_regressor": {
            "baseline_rmse": 0.012263736129838386,
            "candidate_rmse": 2.7068358298285417e-6,
            "outcome": "win",
            "rmse_factor": 4530.653833784668
          },
          "tabpfn_v3_regressor_default": {
            "baseline_rmse": 0.012247530436579227,
            "candidate_rmse": 2.7068358298285417e-6,
            "outcome": "win",
            "rmse_factor": 4524.6668828656
          }
        },
        "audit": {
          "declared_features": 101,
          "declared_test_rows": 89,
          "declared_train_rows": 442
        },
        "claim_boundary": "Diagnostics are consistent with a low-dimensional calibration structure; they do not establish a universal spectral-learning mechanism.",
        "dataset_hash": "9e5ba1af84e3972c",
        "genome_hash": "c924becb3ed0bcf9bfa18bbab69430e65ed3a5cde0fa0ea7171fa832697bb28b",
        "legal_genome": "(pipe X (ridge_gcv))",
        "source_file": "openml-spectrometer-313-target-blue-band-flux-1.csv",
        "strongest_foundation": {
          "candidate_rmse": 2.7068358298285417e-6,
          "factor": 4373.893649300796,
          "identity": "tabfm_ensemble_regressor",
          "rmse": 0.01183941204578691
        },
        "timing": {
          "finalist_cv_secs": 19.012488,
          "fit_predict_secs": 0.563333,
          "search_secs": 20.439741,
          "total_secs": 100.180149
        }
      }
    },
    {
      "class": "estimate",
      "classification": "forward-looking company estimate",
      "company_reported": true,
      "definition": "A receipt is a canonical program executed under a declared referee and producing a finite result or an explicit failure.",
      "empirical_source": null,
      "id": "billion-receipts",
      "label": "Evaluation-scale forecast",
      "observed": false,
      "paper_version": "v1",
      "qualifier": "Not a finding of this study. Generated or merely compiled programs do not count.",
      "record_type": "forward-looking estimate",
      "site_version": "0.2.0",
      "value": {
        "approximate": true,
        "count": 1000000000,
        "unit": "evaluated program receipts"
      }
    }
  ],
  "paper_sha256": "7a80bbfac4f20a9db7b94bd46111fa6262d72bc65dbcf33c3605a484f68dabce",
  "paper_version": "v1",
  "schema": "move42.evidence.v2",
  "site_version": "0.2.0",
  "status": "company-reported"
}