# Historical HFA source audit — 2026-10-10

The historical discovery inventory is useful for locating related work, but it does not establish HFA's first/only status or submission readiness. This audit checks the **35 actual bibliography entries** (the header says 34) and the four September 2 reports. It records concrete metadata corrections and limits the interpretations the retrieved sources can support. This audit does not edit the historical files; the reports already carry October 10 supersession notices above their retained bodies.

This complements [the October 10 literature correction](literature-correction-2026-10-10.md); it does not replace the correction concerning Positive Alignment v1 §5.3. [The JSON ledger](source-audit-2026-10-10.json) contains original fields, observed metadata, versions, primary URLs, input hashes and individual boundaries.

## Scope and method

Access date for every source below: **2026-10-10**. The input corpus is the historical BibTeX file and `hfa-blue-ocean-verification`, `hfa-competitive-landscape-complete`, `hfa-competitive-landscape-deep-read`, and `hfa-gate-close-report`, all dated 2026-09-02. Sources were resolved through supplied identifiers, targeted title searches, versioned arXiv records, publisher/author/institutional pages, and public DataCite/Crossref records. No experiments, production walkthroughs or independent proof checks were performed.

**Verified** means the specified primary text supports the bounded statement; it does not mean the source's scientific results are true or all metadata is complete. **Metadata-only** means a primary abstract/registry description was retrieved but the relevant full text, implementation or complete series was not audited. **Mismatched** identifies a concrete field or version-attribution conflict; it can coexist with limited full-text access. **Inaccessible** would mean no adequate primary evidence was obtained, not that a source is false. Some full texts were inaccessible even where metadata fallback succeeded. These are audit dispositions, not research quality scores.

## Changes required in a living literature review

1. Retain the existing Positive Alignment correction: institutional auditing, local-norm middleware and continuing governance are already discussed in v1 §5.3. Neither a model-training-only characterization nor a claim that no relevant follow-up exists is warranted. [Versioned source](https://arxiv.org/html/2605.10310v1).
2. Narrow the metrics/economics-versus-governance distinction. Ruan and colleagues' v2 abstract includes institutional appraisal, assurance, safeguards and post-deployment monitoring; FVT v2 explicitly includes AI governance. These statements do not establish an implemented action-boundary gate, but they defeat a blanket exclusion from governance. [RoF v2](https://arxiv.org/abs/2608.00151v2), [FVT v2](https://arxiv.org/abs/2608.01432v2).
3. Compare EAF and HFA at the same evidence level. The EAF repository at `cbb9556ea4d47b8ee8d9473e353cf6d705132147` (2026-07-05) connects flourishing to governance constraints and receipts. Its public boundary and spec explicitly limit the material to conceptual/toy examples and exclude production proof. HFA cannot establish operational superiority merely by contrasting those declared limits with its own unverified deployment counts. [Pinned repository](https://github.com/tombudd/eudaimonic-alignment/tree/cbb9556ea4d47b8ee8d9473e353cf6d705132147).
4. Treat FAR as a partially resolved series. Three identified December 2025 records support appraisal axioms, directional dynamics and viability. This audit does not establish the claimed seven-paper inventory or that the entire series lacks runtime/governance content. [Axioms](https://api.datacite.org/dois/10.5281/zenodo.17813646), [Curvature](https://api.datacite.org/dois/10.5281/zenodo.17852704), [Viability](https://api.datacite.org/dois/10.5281/zenodo.17894203).
5. Separate proposed mechanisms, synthetic evaluations, prototypes and author-reported deployments. In particular, RiskGate leaves quantitative evaluation to future work; AEGIS draft v0.7.1 disclaims production readiness; LATTICE's cited planner comparison uses a conservative point with no automatically allowed action. These sources cannot be pooled as equivalent deployment evidence. [RiskGate](https://arxiv.org/abs/2604.24686v1), [AEGIS registry](https://api.datacite.org/dois/10.5281/zenodo.19545991), [LATTICE](https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1800407/full).

A defensible contribution statement is that HFA **proposes a particular synthesis and translation method**. Its individual mechanisms and measurement assumptions can then be compared with identified prior work. No documented search-completeness argument in this audit establishes priority for that synthesis. Test totals, round-the-clock operation and production claims require their own pinned evidence; even verified engineering tests would not establish an effect on flourishing.

## Bibliography ledger

Each row uses the citation key from the preserved historical file. Exact source metadata and additional source links are in the JSON record. Dates below are source dates, not this audit's access date.

| Historical key | Disposition | Primary source and observed version | Correction or boundary |
|---|---|---|---|
| `land_dark_enlightenment` | metadata-only | [Primary record](https://www.thedarkenlightenment.com/the-dark-enlightenment-by-nick-land/); unversioned web essay | Title and byline identify the essay. Use its direct URL, not only the site homepage. The precise publication date was not resolved from the inspected primary text. |
| `buterin_dacc` | mismatched | [Primary record](https://vitalik.eth.limo/general/2023/11/27/techno_optimism.html); dated author essay | The bibliography says 2024; the author page is dated November 27, 2023. |
| `buterin_dacc_one_year` | verified | [Primary record](https://vitalik.eth.limo/general/2025/01/05/dacc2.html); dated author essay | Title, author and January 2025 date match. |
| `life_reacc` | verified | [Primary record](https://omniharmonic.substack.com/p/a-regenerative-accelerationist-manifesto); dated author manifesto | Title, author and January 2026 date match; subtitle is Navigating the Technocapitalist Endgame. |
| `cosmos_institute` | mismatched | [Primary record](https://blog.cosmos-institute.org/p/introducing-the-cosmos-institute); dated institutional announcement | The entry gives 2025 and adds an AI and Human Flourishing title suffix. The primary heading and date are different. |
| `anarcho_accelerationism` | mismatched | [Primary record](https://arxiv.org/abs/2602.13154v1); v1 title inspected; latest v9 has a different title and author-disavowal comment | Anonymous is incorrect even for v1. The historical title matches v1, whereas v9 is Anarchist Automation: A Sociotechnical Framework for Decentralization and Universal Care. Preserve the withdrawal/version history. |
| `macaskill_viatopia` | verified | [Primary record](https://www.forethought.org/research/viatopia); dated author essay | Title, author and January 2026 date match. |
| `macaskill_better_futures_4` | mismatched | [Primary record](https://www.forethought.org/research/persistent-path-dependence); Better Futures part 4 of 6 | The bibliography uses a descriptive invented title and 2026. The page heading is Persistent Path-Dependence (browser title adds Why Our Actions Matter Long-Term), dated August 3, 2025. |
| `macaskill_better_futures_5` | mismatched | [Primary record](https://www.forethought.org/research/how-to-make-the-future-better); Better Futures part 5 of 6 | Replace the descriptive Policy and Regulatory Actions title and 2026 date with the primary title and 2025 date. |
| `finnveden_agi_lockin` | verified | [Primary record](https://www.forethought.org/research/agi-and-lock-in); dated author essay with update notice | Title, author list and stated update period match. |
| `kaptein_runtime_governance` | metadata-only | [Primary record](https://arxiv.org/abs/2603.16586v1); arXiv v1 | Title, authors and date match the primary abstract record. |
| `rhodes_poe` | metadata-only | [Primary record](https://arxiv.org/abs/2607.05397v1); arXiv v1 | Title and authors match. The source displays April 26 despite the 2607 identifier prefix; do not silently replace the displayed date by an inferred month. |
| `bhardwaj_abc` | metadata-only | [Primary record](https://arxiv.org/abs/2602.22302v1); arXiv v1 | Title, author and date match the primary record. |
| `tannenbaum_aegis` | metadata-only | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.19545991); draft v0.7.1 | Title/author match. The deposited version is draft v0.7.1. DOI 19545991 is a version of concept DOI 19542056, not a separate paper. |
| `kenney_governance_stack` | metadata-only | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.19477446); None | The title and Noah M. Kenney attribution match; the missing identifier can be resolved to DOI 10.5281/zenodo.19477446. The author site also identifies the 2026 open-access book. |
| `lattice_governance` | mismatched | [Primary record](https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2026.1800407/full); Frontiers published article, volume 9 | Frontiers in Artificial Intelligence is the venue, not the author. The primary byline is Elias Calboreanu. |
| `aegis_action_boundary` | mismatched | [Primary record](https://arxiv.org/abs/2608.16891v1); arXiv v1 | Aegis Team is a placeholder; primary attribution is Adam Mazzocchetti. Preserve the displayed date rather than deriving one from the identifier prefix. |
| `shipkowski_persistence_kernel` | metadata-only | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.19600943); None | Title matches; registry creator is James Shipkowski (the extra middle initial in the bibliography is not resolved here). Version DOI 19600943 belongs to concept DOI 18501601. |
| `jennifer_runtime` | mismatched | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.18967870); None | The title identifies the March 11 version. The note imports a directional-layer account present in later DOI 19300260, titled Jennifer: A Constraint-Governed AI Runtime with Viability Windows and Deterministic Collapse (March 28). Both are versions of concept DOI 18718591. The creator is James Shipkowski; ref impl is not an author name. |
| `sarc_composition` | mismatched | [Primary record](https://arxiv.org/abs/2608.18360v1); arXiv v1 | Replace SARC composition authors with Gaston Besanson. |
| `reconstructive_authority` | mismatched | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.19699460); 1.1 | RAM authors is a placeholder. The registry creator is Marcelo Patricio Fernandez and version is v1.1. |
| `invariant_preserving_values` | mismatched | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.19452989); 2.0 | Replace the placeholder with the deposited creators John Richard Smith and SHAI / HATI / Deepseek, preserving their stated provenance. Registry version is v2.0; this is a concept DOI with version DOI 19452990. |
| `mchenry_civitae` | verified | [Primary record](https://www.clawrxiv.io/abs/2604.00713); clawRxiv 2604.00713, no version identifier displayed | The cited identifier resolves to the expected title and Deric J. McHenry byline. Another search result has 2604.00830; that does not make the cited 00713 record wrong. |
| `cais_mdpi` | mismatched | [Primary record](https://api.crossref.org/works/10.3390/make8050125); published article, Machine Learning and Knowledge Extraction 8(5), 125 | The historical label is a descriptive placeholder, not the article title. Primary registry identifies the author and DOI 10.3390/make8050125. Publisher search text corroborates the mechanism; direct publisher opening failed. |
| `mazzocchetti_aegis_crypto` | metadata-only | [Primary record](https://arxiv.org/abs/2603.16938v1); arXiv v1 | Title and Adam Massimo Mazzocchetti attribution match. |
| `marin_riskgate` | metadata-only | [Primary record](https://arxiv.org/abs/2604.24686v1); arXiv v1 | Title and author pair match. |
| `besanson_sarc` | metadata-only | [Primary record](https://arxiv.org/abs/2605.07728v1); arXiv v1 | Title, author and date match. |
| `salmon_nine_laws` | metadata-only | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.20617735); None | Title and Sheldon K. Salmon creator match. v2.2 is in the title; the registry version field is null. Description also credits ALBEDO (AionSystem), which should not be silently collapsed into a human byline. |
| `laukkonen_positive_alignment` | metadata-only | [Primary record](https://arxiv.org/abs/2605.10310v1); arXiv v1 used for the existing manuscript correction | The title and 16 named authors match. Bibliographic identity is sound; the historical reports narrow the intellectual scope too far. |
| `abrahams_far_series` | metadata-only | [Primary record](https://api.datacite.org/dois/10.5281/zenodo.17813646); one version DOI and two concept DOIs; see component records | The bibliography has no item-level identifiers and says 7+ papers. Three records named in the reports were resolved. This does not verify the entire series or the listed seven-paper count. |
| `ruan_flourishing_metrics` | metadata-only | [Primary record](https://arxiv.org/abs/2608.00151v2); arXiv v2 | Title and three authors match. July is the original submission date, while the inspected revision is August 4. |
| `ruan_fvt` | metadata-only | [Primary record](https://arxiv.org/abs/2608.01432v2); arXiv v2 | Title, author and August date match. |
| `vanderweele_flourishing_considerations` | metadata-only | [Primary record](https://api.crossref.org/works/10.3390/info17010088); Information 17(1), article 88 | Crossref confirms the title, author pair, year, volume and issue; publisher PDF link is supplied by the registry. Direct publisher retrieval failed in this audit. |
| `harvard_flourishing_initiative` | verified | [Primary record](https://hfh.fas.harvard.edu/post/ai-and-flourishing-initiative); unversioned institutional page inspected 2026-10-10 | The initiative title and partnership match. The bibliography uses an institutional author label rather than the page byline. The visible dateline alone does not verify 2026. |
| `ostrom_polycentric` | mismatched | [Primary record](https://www.aeaweb.org/articles?id=10.1257/aer.100.3.641); American Economic Review 100(3), pages 641-672 | The 2010 source is a journal article, not a book. Use DOI 10.1257/aer.100.3.641 and the AER venue/pages; distinguish the earlier Nobel lecture if citing that version. |

Inventory dispositions: metadata-only: 16; mismatched: 13; verified: 6. These counts are generated from the ledger, not inferred from the historical category headings.


## Supplementary sources and version relationships

- **FAR:** the three inspected DOI records are not interchangeable with a complete series bibliography. Creator strings in the registry include organization/citation text; preserve the deposited wording and resolve authorship before a formal citation cleanup. Concept DOIs 17852704 and 17894203 point to version DOIs 17852705 and 17894204. The audit retrieved the concept metadata, not those version PDFs.
- **Persistence Kernel:** version DOI 19600943 is related to concept DOI 18501601. They are not independent research outputs for counting purposes. [Primary relation metadata](https://api.datacite.org/dois/10.5281/zenodo.19600943).
- **Jennifer:** March 11 DOI 18967870 and March 28 DOI 19300260 are versions of concept 18718591 with different titles. The latter describes the directional layer and qualifies itself as a prototype. [Later-version metadata](https://api.datacite.org/dois/10.5281/zenodo.19300260).
- **AEGIS (Tannenbaum):** 19545991 is a version of 19542056. Keep this project distinct from Calboreanu's AEGIS/LATTICE and Mazzocchetti's Aegis papers. A shared name does not establish a common implementation. [Version relations](https://api.datacite.org/dois/10.5281/zenodo.19545991).
- **Recursive Meta-Governance:** Edward Kipkalya, 2026-02-27, deposited version label Type 1.0. The description makes formal claims and describes lightweight simulations and conceptual cases; this audit did not validate its proofs. [Primary metadata](https://api.datacite.org/dois/10.5281/zenodo.18803430).
- **The Metabolic State Roadmap:** Mark Anthony Brewer, 2026-04-11. The deposited description contains a roadmap, constitutional mechanisms and strong operational assertions. Reading those assertions does not verify deployment, and this limited check cannot justify declaring the work irrelevant or false. [Primary metadata](https://api.datacite.org/dois/10.5281/zenodo.19504961).
- **EAF:** four pinned public files were inspected: [README](https://github.com/tombudd/eudaimonic-alignment/blob/cbb9556ea4d47b8ee8d9473e353cf6d705132147/README.md), [public boundary](https://github.com/tombudd/eudaimonic-alignment/blob/cbb9556ea4d47b8ee8d9473e353cf6d705132147/PUBLIC_BOUNDARY.md), [constitutional spec](https://github.com/tombudd/eudaimonic-alignment/blob/cbb9556ea4d47b8ee8d9473e353cf6d705132147/formal/constitutional-spec.md), and [audit receipts](https://github.com/tombudd/eudaimonic-alignment/blob/cbb9556ea4d47b8ee8d9473e353cf6d705132147/architecture/audit-receipts.md). The longer paper and companion benchmark repository were not audited.

## Limits and next review requirements

This is a source-resolution audit of a supplied inventory, not an exhaustive search of prior work. The current retrieval date is later than the historical reports; unversioned pages were not reconstructed as they appeared in September. Version dates are retained where available. The 35 classifications indicate inventory coverage only, not completeness of the literature or correctness of its claims.

Future novelty comparisons should record inclusion criteria and dated queries, resolve compound series into individual works, and use a shared extraction form: objective, governance level, decision boundary, enforcement assumptions, public artifact, evaluation population and untested claims. Record a missing or inaccessible field as unknown. Compare the proposed HFA mechanism at that level before asserting a difference, and evaluate beneficial human outcomes separately from compliance with a software rule.