# Contribution Novelty Ledger

Last updated: 2026-07-25

This ledger audits the book's signature ideas and coined terms the same way the
non-core evidence ledger audits measured side lanes: it names the contribution
claim, closest source-noted comparators, current delta, strongest objection,
and next artifact. It is not proof of novelty, not an exhaustive prior-art
review, not external review, and not support-state movement.

## Ledger Boundary

- Source of record: `docs/contribution_novelty_ledger.json`
- Validator: `python3 scripts/validate_contribution_novelty_ledger.py`
- Rows: 9 signature ideas
- Chapter-core support effect: none; all 87 chapter core claims remain
  `argument`.
- Novelty claim effect: none; the ledger permits disciplined positioning, not
  defended novelty language.
- Current support-state discipline: historical negative labels rehabilitated
  without core movement.

## Signature-Idea Rows

| Idea | Current confidence | Closest comparator IDs | Delta | Next artifact |
|---|---|---|---|---|
| Verification bandwidth | `capacity_model_backed_not_empirical` | `ext_lost_in_middle_2023`; `ext_longbench_2023`; `ext_ruler_2024`; `ext_helm_2022`; `ext_livebench_2024` | Frames verification capacity as an architectural budget that must be routed, escalated, and recorded before claims can move; the first record-level capacity model exists, but contradiction-rate and distractor-resistance evidence remains open. | Contradiction-rate and distractor-resistance benchmark or replayed adequacy-classifier trace with logged prompts, baselines, negative controls, and independent review. |
| Pattern language for governed cognition | `pattern_local_delta_broadened_not_novelty_proven` | `ext_model_cards_2019`; `ext_datasheets_datasets_2021`; `ext_factsheets_ai_services_2019`; `ext_drexler_cais_2019`; `ext_nist_ai_rmf_1_0_2023`; `ext_frontier_ai_regulation_2023` | CAIS already frames broad AI as composed services and R&D automation. The narrower ASI Stack target is an interface contract binding authority ceilings, claim movement, belief continuity, artifact reality, effect leases, routability, evidence receipts, residual custody, rollback, and release status; the opener names it, Integrated Reference Architecture applies it at trace scale, Living Book Methodology applies it to book changes, and non-owner chapters must state a local delta rather than reuse it as a novelty claim. | Preserve local-delta discipline in future chapter edits and pursue source-noted prior-art or external review before stronger novelty language; no standalone audit report by default. |
| Verification tax and governance economics | `bounded_evidence_exists_non_core` | `ext_frugalgpt_2023`; `ext_hybrid_llm_2024`; `ext_routellm_2024`; `ext_sparse_moe_2017`; `ext_switch_transformer_2021`; `ext_helm_2022` | Adds authority cost, verification burden, fallback cost, and residual custody to cost/quality routing. | Larger public-safe workload with route quality, hidden cost, fallback, and reviewer burden measurements. |
| Residual honesty | `residual_storage_replay_backed_not_deployed` | `ext_helm_2022`; `ext_goodhart_variants_2018`; `ext_benchmark_contamination_2023`; `ext_model_evaluation_extreme_risks_2023`; `ext_nist_ai_rmf_1_0_2023` | Makes displaced burden a ledgered artifact that can block promotion even when visible metrics improve; the residual-conservation fixture at `experiments/residual_honesty_conservation/results/2026-07-03-local.json` accepts visible accepted/deferred/discharged residual records and rejects hidden, erased, unowned, support-promoting, and zero-residual-overclaim controls; `experiments/residual_ledger_trace/results/2026-07-03-local.json` traces residualized deferrals, displaced costs, repair residuals, readiness escrow, rejected hidden burdens, and no-promotion decisions across current Resource, Compact GVR, and Readiness artifacts; and `experiments/residual_ledger_storage_replay/results/2026-07-04-local.json` replays a bounded append-only residual ledger event log with owner handoff, discharge review, workload context, digest-chain construction, and five rejected invalid controls. | Live or externally reviewable residual-ledger storage/replay with real or public-safe workload context, owner handoff, discharge review, tamper/adversarial replay behavior, and independent review. |
| Support-state discipline | `historical_negative_labels_rehabilitated_without_core_movement` | `ext_model_cards_2019`; `ext_datasheets_datasets_2021`; `ext_ml_reproducibility_program_2021`; `ext_proof_carrying_code_1997`; `ext_agm_belief_revision_1985`; `ext_truth_maintenance_system_1979`; `ext_assumption_based_tms_1986` | Joins reporting, reproducibility, proof receipts, and belief revision into public claim-state transition discipline. The post-v2.3 transition files preserve two historical QCSA `refuted` labels, while the competence audit classifies both as N2 proxy/regime evidence with no exact, broad, parent, or chapter-core refutation. | Earn N3 or stronger negative inference only through a prospectively competent claim-matched test; do not manufacture one. Current artifacts include the immutable transition records and `evidence_quality/negative_result_rehabilitation.json`. |
| Stable capability fields | `promising_not_defended` | `ext_semver_2_0_0`; `ext_slsa_v1_0`; `ext_feature_toggles_fowler`; `ext_argo_rollouts_docs`; `ext_capability_based_computer_systems_1984` | Applies identity-under-replacement discipline to AI capability fields with evidence-state and authority-preservation gates. | Executable field-transition model over replacement sequences. |
| Bounded self-improvement | `safety_critical_not_defended` | `ext_corrigibility_2015`; `ext_off_switch_game_2016`; `ext_optimal_policies_power_2019`; `ext_model_evaluation_extreme_risks_2023`; `ext_frontier_ai_regulation_2023`; `ext_slsa_v1_0` | Packages improvement as a gated artifact lifecycle rather than unconstrained optimization. | Clean Theseus replay or archived public fixture plus transition-system invariant. |
| Human oversight degradation | `finite_fixture_backed_not_deployed` | `ext_humans_automation_1997`; `ext_ironies_automation_1983`; `ext_levels_automation_2000`; `ext_complacency_bias_automation_2010` | Turns approval fatigue, rubber-stamping, alarm fatigue, and automation bias into adapter record fields, reviewer-rotation routes, independent-evidence checks, and no-promotion boundaries. The first bounded fixture exists at `experiments/human_oversight_degradation/results/2026-07-03-local.json`, and the accepted no-promotion decision at `evidence_transitions/v1_x_measured/human_oversight_degradation_no_change.json` blocks approval-workflow, deployed-human-factors, reviewer-correctness, alert-quality, and Runtime Adapters chapter-core promotion. | Deployed or externally reviewable approval-workflow trace with public-safe reviewer-load fields, independent-evidence checks, reviewer-rotation behavior, alert-quality evidence, privacy boundaries, and independent review before any upward support movement. |
| Record-reality gap | `public_site_record_reality_attestation_backed_not_external_review` | `ext_proof_carrying_code_1997`; `ext_ai_safety_debate_2018`; `ext_llm_as_judge_mt_bench_2023`; `ext_model_evaluation_extreme_risks_2023`; `ext_nist_zero_trust_architecture_2020` | Treats receipt faithfulness and the epistemic trusted computing base as first-class architecture problems: records must correspond to reality, and verifier trust must bottom out in named roots rather than ambient confidence. Current bounded backing state: `repository_receipt_audit_backed_not_open_world`, `repository_receipt_challenge_backed_not_open_world`, `live_artifact_attestation_backed_not_open_world`, `randomized_artifact_attestation_backed_not_open_world`, `github_pages_ci_attestation_backed_not_external_review`, and `public_site_record_reality_attestation_backed_not_external_review`. The first bounded receipt fixture exists at `experiments/receipt_faithfulness/results/2026-07-03-local.json`, the first repository receipt audit exists at `experiments/receipt_repository_audit/results/2026-07-03-local.json`, the deterministic repository challenge exists at `experiments/receipt_repository_audit/results/2026-07-04-challenge.json`, the local live artifact attestation probe exists at `experiments/artifact_live_attestation/results/2026-07-04-local.json`, the deterministic pseudo-random multi-artifact audit exists at `experiments/artifact_randomized_attestation/results/2026-07-04-local.json`, the GitHub Pages CI attestation exists at `experiments/artifact_github_pages_ci_attestation/results/2026-07-05-local.json`, the public deployed-site attestation exists at `experiments/artifact_public_site_attestation/results/2026-07-05-live.json`, the first bounded epistemic trusted computing base fixture exists at `experiments/epistemic_tcb/results/2026-07-03-local.json`, and the Artifact Graphs owner-section now states the record-reality objection, bounded answer, weakening condition, and no-deployed-faithfulness boundary in live and curated reader prose. | Deployed or externally reviewable attestation/audit beyond served-page and CI-service status, verifier-quality evidence, trust-base storage/replay evidence, or a randomized/external challenge outside this local repository; no standalone chapter unless later evidence makes the owner section duplicative or too narrow; explicit boundary that current fixtures, repository audit, deterministic challenge, live artifact attestation probe, randomized artifact attestation audit, GitHub Pages CI attestation, public deployed-site attestation, and owner-section prose do not prove open-world receipt faithfulness, verifier correctness, external project truth, independent external human review, reader release approval, deployed attestation behavior, or deployed trust-base behavior. |

## Current Decision

The book may use these ideas as signature framing, but it should not call them
defended contributions unless the row's next artifact exists and the relevant
review, proof, source, or evidence gate passes. The most important still-open
row is `receipt_faithfulness_gap`: it now has bounded receipt-faithfulness and
epistemic trusted computing base fixtures, live and randomized local repository
attestation audits, CI-service and public served-page attestations, plus an
Artifact Graphs owner-section treatment in both live and curated reader prose,
but still needs deployed or externally reviewable attestation, audit,
trust-base storage/replay, and verifier-quality evidence before the book can
claim a complete answer to record or verifier-trust laundering.

## Non-Claims

- This ledger does not prove novelty.
- This ledger does not prove that any ASI Stack mechanism works.
- This ledger is not accepted external review.
- This ledger is not a source-derived support-state transition.
- This ledger does not approve reader, release, ebook, PDF, DOCX, audio, DOI,
  archive, implementation, model-quality, benchmark, safety, or deployment
  claims.