Management Research Notes

A file-based academic knowledge base for management and business sustainability research.

One Markdown note per peer-reviewed paper. Trusted bibliographic metadata from manifests, analytic distillation from a verbatim-anchored extraction prompt, derived SQLite / CSV / BibTeX indexes that you can rebuild from scratch in seconds.

---

Author

Management Research Notes is authored and maintained by Binqi Tang , a researcher in management and business sustainability. Collaborations are welcome — whether that means using this knowledge base in your own research workflow, adapting it for a new source or subfield, or discussing joint projects. Please open an issue or reach out on GitHub. Citations welcome — see How to cite below.

---

What it is

This project is a Zettelkasten-style knowledge base for academic literature. Each peer-reviewed article is stored as a single Markdown file in notes/ with two parts:

1. YAML frontmatter — structured metadata: trusted bibliographic fields (title, authors, year, journal, DOI), a controlled-vocabulary topic list, the methods, the sample, and three custom analytic fields (unit_of_analysis, level_of_theory, dependent_variable_family) that make cross-paper queries actually useful.
2. Markdown body — a human-readable distillation: verbatim abstract, research question, mechanism / process, theoretical contribution, practical implication, limitations, future research, and a clean APA 7th citation.

A small Python pipeline derives a SQLite index, a flat CSV, and a BibTeX file from the notes — none of which are the source of truth, all of which can be rebuilt with python tools/build_index.py && python tools/export_csv.py && python tools/export_bibtex.py.

Why it exists

Large language models have made it tempting to build a literature review by dropping a pile of PDFs into a chat window and trusting the response. For scholars in management, economics, and adjacent citation-heavy fields — disciplines where every substantive claim is expected to be traceable to a specific prior study — that naive workflow breaks down for three reasons:

• Context windows are finite. Even a 200K-token model cannot hold a few hundred full papers at once. You need distilled notes small enough to load in bulk but faithful enough to support real argument.
• Provenance is non-negotiable. Every analytic claim in a note is traceable to a cited paper. The verbatim abstract is checked as a contiguous substring of the extracted PDF text, every factual claim in a v2 note carries a ≤25-word evidence anchor verified against the PDF by tools/validate_note.py, and a two-layer faithfulness audit (see below) has a fresh-eyes LLM subagent cross-check the prose fields. Hallucinated or paraphrased content does not reach the published library — which is what makes these notes safe to cite in a defensible review.
• Cross-platform and agent-ready. The same note is grep-able from the command line, uploadable as project knowledge to an LLM chat, and citable from a Word / LaTeX / Typst draft via the BibTeX export. Via the SQLite + FTS5 index and the AGENTS.md entry point, any AI agent — Claude Code, Cursor, Windsurf, a custom SDK app — can consume the library to run retrieval-augmented analysis with known faithfulness guarantees. That's the piece that matters for researchers: it lets one scholar (or an agent acting on their behalf) reason across a literature much larger than any single head can hold, without sacrificing the citation trail a peer reviewer, committee, or journal editor will expect.

The corpus shipped here is management and business sustainability, but the pattern — one Markdown file per paper, trusted bibliographic manifests, a verbatim-anchored extraction prompt, a two-layer audit — transfers straightforwardly to any scholarly corpus where paraphrastic drift is costly: economics working papers, organization theory, policy analysis, STS, labor studies. Fork the repo, point it at your own manifest, and the validator and audit layer come along for free.

The current design targets libraries growing from dozens to tens of thousands of papers. Whether the shape holds at that scale is an empirical question rather than a settled assumption: some parts — one file per paper, SQLite-derived indexes, the controlled-vocabulary topic list, even the extraction prompt's field set — may need to evolve as the library grows, new sources appear, or research workflows change. Treat the current architecture as a working hypothesis refined by each release, not a frozen spec.

Faithfulness audit

Every analytic field in every note is checked by a two-layer faithfulness audit before it is accepted into the library:

• Layer 1 — Evidence anchors (mechanical, deterministic). Each factual claim in the frontmatter (sample size, country, industry, time period, theories, methods, keywords) carries a ≤25-word verbatim quote from the PDF. The validator checks each quote is a substring of the extracted PDF text under hyphen-tolerant normalization. A fabricated quote fails deterministically — there is no way to pass Layer 1 with invented evidence.
• Layer 2 — Semantic audit (fresh cold-context subagent). For the prose fields (research question, mechanism, theoretical contribution, practical implication, limitations, future research) a fresh independent auditor reads the PDF, reads the note, and emits a per-field verdict against the rubric at docs/audit-rubric.md: SUPPORTED / PARTIAL / UNSUPPORTED / CONTRADICTED. A note is rejected if any verdict is UNSUPPORTED or CONTRADICTED.

The full library has been swept across releases:

• v0.2.0 sweep (2026-04-17, 90 notes): 88 / 90 initial PASS; two fails (Mahringer 2025, Li 2026) repaired via re-extraction, and the validator was tightened in one place (extended the Not reported in paper escape valve to future_research, mirroring the limitations exemption) based on what the audit found.

• v0.3.0 sweep (2026-04-17, full 105 notes): 105 / 105 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. One PARTIAL verdict (Castelló 2025 citing the EU Digital Markets Act instead of the EU Digital Services Act) was surfaced and fixed before release.

• v0.4.0 sweep (2026-04-29, 113 notes): 113 / 113 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Eight new notes added from AMJ vol. 68 no. 4 (Grégoire editorial + 7 substantive papers).

• v0.5.0 sweep (2026-04-30, 121 notes): 121 / 121 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Eight new notes added from AMJ vol. 68 no. 3 (Bermiss editorial + 7 substantive papers).

• v0.6.0 sweep (2026-04-30, 129 notes): 129 / 129 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Eight new notes added from AMJ vol. 68 no. 2 (Rouse editorial + 7 substantive papers). Cleanest batch yet — only 1 PARTIAL across the 8 new notes.

• v0.7.0 sweep (2026-04-30, 139 notes): 139 / 139 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Ten new notes added from AMJ vol. 68 no. 1 (Gruber + Cronin editorials + 8 substantive papers — first 10-paper issue we've processed).

• v0.8.0 sweep (2026-05-05, 147 notes): 147 / 147 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Eight new notes added from AMJ vol. 67 no. 6 (Reinecke editorial + 7 substantive papers — first issue from volume 67 we've processed).

• v0.9.0 sweep (2026-05-05, 156 notes): 156 / 156 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Nine new notes added from AMJ vol. 67 no. 5 (Dorobantu editorial + 8 substantive papers). Zero PARTIAL across all 54 prose-field audits — first perfect-SUPPORTED batch.

• v0.10.0 sweep (2026-05-06, 165 notes): 165 / 165 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Nine new notes added from AMJ vol. 67 no. 4 (Bliese editorial + 8 substantive papers). 1 PARTIAL.

• v0.11.0 sweep (2026-05-07, 175 notes): 175 / 175 PASS, 0 UNSUPPORTED, 0 CONTRADICTED. Ten new notes added from AMJ vol. 67 no. 3. Three-batch ≤1-PARTIAL streak.

• v0.11.1 patch (2026-05-07, 175 notes): Bibliographic-integrity correction. A user-flagged check against Google Scholar surfaced that the Bednar 2024 note (and many others) were citing the online-first publication year rather than the issue year per APA 7. Built tools/verify_years.py to cross-check every DOI against CrossRef's published-print field. Found 48 papers (27% of library) carrying the online-first year. All 48 corrected: manifest rows, frontmatter year: fields, and APA citations updated to the issue year. 175/175 now align with CrossRef. The audit prose fields are unchanged (year doesn't affect any audited content claim), so no re-audit was needed.

• v0.11.2 patch (2026-05-09, 175 notes): Bibliographic cross-check extended to all seven CrossRef fields (year, title, journal, volume, issue, pages, authors) via the new tools/verify_metadata.py. The first full sweep on the v0.11.1 library surfaced 21 papers needing manifest backfills (mostly online-first papers in the NBS digest and AMJ vol-69-1 whose issues had since published) plus one wrong-page error (Bamberger's AMR address) and one missing-Oxford-comma title typo in the same paper. After fixes: 175/175 MATCH for year, journal, and authors, and 0 MISMATCH across volume / issue / pages on every paper where CrossRef carries those fields. One remaining flag (Reinecke book review) is documented as a CrossRef-side data corruption, not a note error. Like v0.11.1, no re-audit needed — the audit prose fields are independent of bibliographic metadata.

• v0.12.0 sweep (2026-05-10, 186 notes): Adds AMJ vol. 67 no. 2 (11 substantive papers: Anderson, Arslan, Grégoire, Gruber, Li, Liao, Muethel, Pache, Piazza, Preston, Rostain). 186 / 186 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 0 PARTIAL — the cleanest batch outcome to date. Also introduces Tier 3 of the bibliographic-integrity prevention strategy: the new tools/populate_manifest.py auto-populates each manifest's volume/issue/pages columns from CrossRef BEFORE extraction, and (with --fix-year) auto-corrects any wrong years the manifest may have inherited from the publisher's online-first metadata. This is now Step 0 of /synapse-ingest. The v0.12.0 batch proved the mechanism: Tier 3 caught and fixed 7 wrong-year manifest rows in vol-67-no-2 before extraction started — exactly the v0.11.1 bug class, prevented at the source instead of patched retroactively.

• v0.13.0 sweep (2026-05-12, 197 notes): Adds AMJ vol. 67 no. 1 (10 substantive papers: Barkema, Bettinazzi, Cao, Dorobantu, Han, Jia, two Park papers, Soublière, To, plus Gruber editorial on the new Research Methods Articles portfolio). 197 / 197 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 1 PARTIAL (Cao practical_implication — minor extrapolation from a contributions paragraph). Second consecutive batch ingested under the Tier 3 prevention infrastructure (introduced v0.12.0) and the post-extraction column-merge heuristic (introduced commit 59a9ace). Tier 3 found zero year mismatches in this batch (the manifest had already been captured after print publication), but still backfilled 11 vol/issue/pages columns. Every extraction agent independently used grep on the extracted text to verify anchor contiguity before writing — Layer 1 passed on first try for all 11 papers, the first batch since the column-merge heuristic went into the extraction prompt to do so.

• v0.13.1 patch (2026-05-12, 197 notes): slugify() updated to fold diacritics via NFKD and strip apostrophes, resolving fragmented paper_ids like gr-goire-2024 → gregoire-2024. 5 paper_ids renamed as a deliberate, one-time exception to hard rule 4 (stable IDs) while the library was small and had no known external citations by paper_id.

• v0.13.2 patch (2026-05-12, 197 notes): Adds tools/lint_manifests.py — a per-row structural + CrossRef audit of every manifest. First run surfaced 4 legitimate compound surnames (von Krogh, Lee Cunningham, Ter Wal, van den Oever — added to KNOWN_COMPOUND_SURNAMES allowlist) and 6 latent D'Amico-class bugs in the NBS-2026-02 manifest (full given+family name captured instead of just the family name). 6 more paper_ids renamed; combined with v0.13.1 this closes the v0.13.x cleanup cohort (11 papers total). Hard rule 4's footnote in CLAUDE.md revised to use cohort framing.

• v0.14.0 sweep (2026-05-16, 210 notes): Adds AMJ vol. 66 no. 6 (12 substantive papers + Grimes editorial on generative AI's impact on management scholarship). 210 / 210 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 3 PARTIAL (Couture and Sajjadiani mechanism_process — minor attribution drift; Zhang limitations — added "cross-sectional" label not in paper). First batch ingested under the complete three-gate pipeline: Step 0 populate_manifest.py (Tier 3), Step 0.5 lint_manifests.py (structural), Step 4.5 verify_metadata.py (Tier 2 content). All three gates exit 0. Also backfills v0.11.2-pattern drift on two ASQ papers (Lee Democratic Deviations, Reinecke book review) whose vol-71 issue 2 published after their extraction; their note frontmatter + APA citations now reflect the issued version. Xu 2023 title added to KNOWN_CROSSREF_DATA_ERRORS (CrossRef has malformed Forest<i>and</i>the Trees with no spaces around the italic; note matches the published version correctly).

• v0.15.0 sweep (2026-05-20, 222 notes): Adds AMJ vol. 66 no. 5 (11 substantive papers + Dencker editorial on positioning research on novel phenomena). 222 / 222 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 0 PARTIAL — a perfectly clean batch: 12/12 Layer 1 first-try PASS and 12/12 Layer 2 PASS with no PARTIAL verdicts at all. Ingested under the complete three-gate pipeline (populate_manifest → lint_manifests → verify_metadata), all exit 0 with no drift findings or new false positives. Notable papers include the Lander/Voronov pair on Scottish and Canadian whisky authenticity, and the Carnabuci/Rua-Gomez network-formation studies.

• v0.16.0 sweep (2026-05-20, 233 notes): Adds AMJ vol. 66 no. 4 (10 substantive papers + Wellman editorial on publishing multimethod research). 233 / 233 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 1 PARTIAL (Kundro future_research — mild overgeneralization of the paper's actual future directions). Ingested under the complete three-gate pipeline; all gates exit 0 with no drift findings or new false positives. 11/11 Layer 1 first-try PASS (fifth consecutive batch at 100%). The Banerjee audit is a nice illustration of the Layer 2 value: H4 in that paper predicted one direction but the data showed the opposite (significant) — the note correctly reports the empirical result rather than the failed hypothesis, which the auditor verified as faithful rather than CONTRADICTED.

• v0.17.0 sweep (2026-05-20, 244 notes): Adds AMJ vol. 66 no. 3 (10 substantive papers + Langley editorial on opening up AMJ's research methods repertoire). 244 / 244 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 1 PARTIAL (Toivonen practical_implication — practitioner prescriptions synthesized from a paper with no dedicated implications section). Ingested under the complete three-gate pipeline; all gates exit 0 with no drift findings or new false positives. 11/11 Layer 1 first-try PASS (sixth consecutive batch at 100%). This sweep completes the AMJ vol. 66 + vol. 67 coverage from no. 3 onward — the pilot now spans 13 consecutive AMJ issues plus vol. 68 and vol. 69.

• v0.18.0 sweep (2026-05-23, 257 notes): Adds AMJ vol. 66 no. 2 (12 substantive papers + Krogh editorial on AI research opportunities). 257 / 257 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 1 PARTIAL (Zhang mechanism_process — minor mediator-vs- DV imprecision in one study). 13/13 Layer 1 first-try PASS (seventh consecutive batch at 100%). Three of this issue's papers (Krogh→von Krogh, Cunningham→Lee Cunningham, Wal→Ter Wal) are the compound surnames already in the KNOWN_COMPOUND_SURNAMES allowlist; Step 0.5 suppressed all three cleanly, and verify_metadata's note-level author check passed without new false positives (notes carry the full byline forms that match CrossRef; the short form lives only in the paper_id slug). This release also promotes the extraction_model-verbatim instruction into docs/extraction-prompt.md — a validated systemic fix (13/13 agents kept the field uniform this batch vs. up to 7 outliers needing manual cleanup in prior batches), eliminating a recurring per-batch provenance-field normalization.

• v0.19.0 sweep (2026-05-23, 270 notes): Adds AMJ vol. 66 no. 1 (12 substantive papers + Gruber editorial on the 23rd editorial term's strategic priorities). 270 / 270 PASS, 0 UNSUPPORTED, 0 CONTRADICTED, 0 PARTIAL — the third perfectly-clean batch (after v0.12.0 and v0.15.0). This release also served as a standalone validation of two infrastructure fixes from prior releases:

  • extraction_model verbatim (codified in v0.18.0, commit cc16c71): extraction agents were deliberately dispatched WITHOUT the inline reminder; 13/13 still kept the field uniform purely from reading docs/extraction-prompt.md. The codified location is independently load-bearing.
  • Editorial-abstract escape (added in commit 5b4ae81, pre-v0.19.0): Gruber editorial — the first editorial extracted after the fix — wrote "Not reported in paper" for Abstract instead of the sentence-fragment workaround used by all five earlier editorials. 13/13 Layer 1 first-try PASS (eighth consecutive batch at 100%). All three gates exit 0; zero post-extraction remediation needed.

• v0.19.1 patch (2026-05-27, 270 notes): Provenance correction — extraction_model field. No content changes; no new notes. The v0.18.0 fix had standardized this field to claude-opus-4-6 (the bundle's stale default), but the actual extraction model for batches from 2026-04-18 onward was Claude Opus 4.7 (the parent session's model, with 57 notes carrying direct agent self-reports as evidence). Updated 108 notes that predated my v0.18.0 fix-to-verbatim with the accurate claude-opus-4-7 value; left 105 pre-cutover notes at claude-opus-4-6 (no direct evidence of model identity for those earlier batches). Constant in tools/prepare_paper.py updated to claude-opus-4-7 so future bundles record the accurate model. Library is now provenance-correct for everything where we have direct evidence; pre-cutover notes retain their original metadata as historical record. All three gates still exit 0 after the sweep (provenance field is not validated or audited).

• v0.20.0 sweep (2026-06-11, 352 notes): Adds the NBS December 2025 monthly digest — 82 peer-reviewed papers across Ecological Economics (including the biodiversity-and- finance special issue), AMJ, Business & Society, Review of Finance, Strategic Management Journal, and others. 82 / 82 PASS (Layer 1 + Layer 2). Introduced masthead-based SSRN→published detection (read the page masthead, not the footer DOI, to catch working papers that reached their journal version — 9 upgraded to the published DOI/metadata) and a code-orchestrated parallel-dispatch extraction workflow whose pipeline() fan-out guarantees the subagent count, eliminating hand-dispatch drift. Four incomplete-PDF stubs (Soboleva, Thomas, Willis, Zhu) were re-acquired and ingested. A separate commit cleaned CrossRef metadata drift in 7 prior-release NBS-2026-02 notes (vol/issue/pages now indexed). All three gates exit 0; 6 Review of Finance DOIs are valid but CrossRef-pending and will verify once indexed.

• v0.21.0 sweep (2026-06-23, 465 notes): Adds the NBS January 2026 monthly digest — 113 peer-reviewed papers, led by the Journal of Business Ethics (40) with Research Policy (14), the Journal of Environmental Economics and Management, Organization Science, Human Relations, The Journal of Finance, MIS Quarterly, and others. 113 / 113 PASS (Layer 1 + Layer 2). Each batch ran as two ~5–6-paper waves through the pipeline() extraction workflow to stay under a transient server rate-limit while keeping the subagent count code-guaranteed. Surfaced more SSRN working papers that had reached their published journal version (upgraded to the journal DOI/metadata — e.g. Wang → Journal of Business Ethics, Duguay → Journal of Accounting and Economics) and several filename-lossy title restorations (dropped colons / question marks repaired against CrossRef). Seven incomplete-PDF stubs were found; six were re-acquired and ingested and one (Myers, a special-issue editorial) was skipped. A separate commit cleaned CrossRef metadata drift in 3 prior-release NBS-2026-02 notes. All three gates exit 0 (validate, verify_metadata, lint_manifests).

• v0.22.0 sweep (2026-06-24, 542 notes): Adds Academy of Management Journal volume 65, issues 1-6 — 77 peer-reviewed papers with issue counts 13, 13, 13, 12, 13, 13. This completes AMJ pilot coverage from vol. 65 no. 1 through vol. 69 no. 1. 77 / 77 PASS the faithfulness audit for the AMJ volume-65 batch; the full library now has 542 / 542 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-65 manifests lint cleanly, all notes validate, and CrossRef metadata verification exits 0 with no mismatches. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 542 records.

• v0.23.0 sweep (2026-06-27, 617 notes): Adds Academy of Management Journal volume 64, issues 1-6 — 75 peer-reviewed papers with issue counts 12, 12, 13, 12, 13, 13. This completes AMJ pilot coverage from vol. 64 no. 1 through vol. 69 no. 1. 75 / 75 PASS the faithfulness audit for the AMJ volume-64 batch; the full library now has 617 / 617 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-64 manifests lint cleanly, all notes validate, and CrossRef metadata verification exits 0 with no mismatches; remaining notices are documented CrossRef-side false positives or lookup warnings. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 617 records.

• v0.23.1 patch (2026-06-27, 617 notes): Provenance correction for the AMJ volume-64 Codex batch. The 75 AMJ volume-64 notes now record extraction_model: "gpt-5.5" instead of the stale bundle value claude-opus-4-8. Five audit-identified wording issues were narrowed: Abdurakhmonov et al. (2021) removed an overbroad country-generalization phrase, Jacobs et al. (2021) softened a direct practical prescription to a case-grounded association, Lifshitz-Assaf et al. (2021) removed an unsupported stage-gate reference, Bain et al. (2021) distinguished experimental boundary tests from the Study 3 intervention target, and Ji et al. (2021) removed an unstated national/historical-setting limitation. Abstracts, evidence anchors, bibliographic metadata, and citations are unchanged. The patch also makes GPT-5.5 the current Codex extraction/audit default in the workflow docs and sets issue-level parallelism to a fixed 6-agent cap with fallback to 5, 3, or serial execution if stability degrades.

• v0.24.0 sweep (2026-06-28, 691 notes): Adds Academy of Management Journal volume 63, issues 1-6 — 74 peer-reviewed papers with issue counts 13, 12, 12, 12, 12, 13. This extends AMJ pilot coverage backward to vol. 63 no. 1, so the AMJ pilot now spans vol. 63 no. 1 through vol. 69 no. 1. 74 / 74 PASS the faithfulness audit for the AMJ volume-63 batch; the full library now has 691 / 691 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-63 manifests lint cleanly, all notes validate, and CrossRef metadata verification exits 0 with no mismatches; remaining notices are documented CrossRef-side false positives or lookup warnings. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 691 records.

• v0.25.0 sweep (2026-06-30, 767 notes): Adds Academy of Management Journal volume 62, issues 1-6 — 76 peer-reviewed papers with issue counts 12, 12, 13, 13, 12, 14. This extends AMJ pilot coverage backward to vol. 62 no. 1, so the AMJ pilot now spans vol. 62 no. 1 through vol. 69 no. 1. 76 / 76 PASS the faithfulness audit for the AMJ volume-62 batch; the full library now has 767 / 767 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-62 manifests lint cleanly, all notes validate, and scoped CrossRef metadata verification for all 76 AMJ volume-62 notes exits 0 with no mismatches. The full-library CrossRef sweep was attempted during release preparation but interrupted by a network-level HTTPS handshake stall; existing pre-AMJ62 notes were unchanged from the v0.24.0 metadata-verified snapshot. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 767 records.

• v0.26.0 sweep (2026-07-02, 860 notes): Adds Academy of Management Journal volume 61, issues 1-6 — 93 peer-reviewed papers with issue counts 16, 16, 15, 15, 16, 15. This extends AMJ pilot coverage backward to vol. 61 no. 1, so the AMJ pilot now spans vol. 61 no. 1 through vol. 69 no. 1. 93 / 93 PASS the faithfulness audit for the AMJ volume-61 batch; the full library now has 860 / 860 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-61 manifests lint cleanly, all notes validate, and scoped CrossRef metadata verification for all 93 AMJ volume-61 notes exits 0 with no mismatches. The audit sweep has one non-blocking PARTIAL verdict (Aristidou 2018 limitations), retained as harmless compression around scope-boundary wording rather than a faithfulness failure. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 860 records.

• v0.27.0 sweep (2026-07-03, 955 notes): Adds Academy of Management Journal volume 60, issues 1-6 — 95 peer-reviewed papers with issue counts 16, 16, 16, 17, 15, 15. This extends AMJ pilot coverage backward to vol. 60 no. 1, so the AMJ pilot now spans vol. 60 no. 1 through vol. 69 no. 1. 95 / 95 PASS the faithfulness audit for the AMJ volume-60 batch; the full library now has 955 / 955 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-60 manifests lint cleanly, all notes validate, and scoped CrossRef metadata verification for all 95 AMJ volume-60 notes exits 0 with no mismatches; two AMJ volume-60 author-name notices are documented as CrossRef-side family-name parsing false positives. Two audit-identified future-research fields (Fan 2017 and Katila 2017) were narrowed before release; the final AMJ volume-60 audit scan has 570 / 570 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 955 records.

• v0.28.0 sweep (2026-07-05, 1,050 notes): Adds Academy of Management Journal volume 59, issues 1-6 — 95 peer-reviewed papers with issue counts 16, 15, 16, 16, 16, 16. This extends AMJ pilot coverage backward to vol. 59 no. 1, so the AMJ pilot now spans vol. 59 no. 1 through vol. 69 no. 1. 95 / 95 PASS the faithfulness audit for the AMJ volume-59 batch; the full library now has 1,050 / 1,050 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-59 manifests lint cleanly, all notes validate, and scoped CrossRef metadata verification for all 95 AMJ volume-59 notes exits 0 with two documented CrossRef-side title false positives in issue 1 (Durand and Zavyalova malformed inline HTML). The final AMJ volume-59 audit scan has 570 / 570 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 1,050 records.

• v0.29.0 sweep (2026-07-07, 1,128 notes): Adds Academy of Management Journal volume 58, issues 1-6 — 78 peer-reviewed papers with issue counts 13, 13, 13, 13, 13, 13. This extends AMJ pilot coverage backward to vol. 58 no. 1, so the AMJ pilot now spans vol. 58 no. 1 through vol. 69 no. 1. 78 / 78 PASS the faithfulness audit for the AMJ volume-58 batch; the full library now has 1,128 / 1,128 PASS, 0 UNSUPPORTED, and 0 CONTRADICTED verdicts. All six AMJ volume-58 manifests lint cleanly, all notes validate, and scoped CrossRef metadata verification for all 78 AMJ volume-58 notes exits 0 with three documented CrossRef-side false positives: Byron title inline-HTML spacing, Little / Smith Major author parsing, and Joshi title inline-HTML spacing. Six audit-identified prose fields were narrowed before release (Eggers, Gabriel, Gurses, Kish-Gephart, Lioukas, and Wo); the final AMJ volume-58 audit scan has 468 / 468 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. The public indexes were rebuilt from notes: SQLite, CSV, and BibTeX all contain 1,128 records.

• v0.30.0 tooling (2026-07-09, 1,128 notes): Schema-and-tooling release — no new notes. Introduces extraction v3, which adds three empirical body sections to every future note — Hypotheses / Propositions, Data & Measures, and Key Findings — plus three matching Layer 1 evidence anchors (hypotheses_source, measures_overview, findings_overview) and three new Layer 2 audited prose fields under audit rubric v2 (nine fields for v3 notes; a reversed finding direction is CONTRADICTED, per the sign-reversal rule). Applied going-forward only: the validator, auditor, and indexer are version-gated on extraction_version, all 1,128 existing v1/v2 notes revalidate unchanged (1,128 / 1,128 OK), and the rebuilt SQLite/CSV indexes carry the three new columns (still 1,128 records; key_findings and hypotheses are FTS5-searchable). Also adds docs/pipeline-runbook.md — a vendor-neutral ingest/audit/publish runbook referenced from AGENTS.md — hardens the shared CrossRef client (retry-with-backoff, negative 404 caching) used by all three metadata gates, and fixes stale documentation references (ingest_batch.py, the tools list, the audit invocation examples).

• v0.31.0 sweep (2026-07-10, 1,141 notes): Adds Academy of Management Journal volume 57, issue 1 — 13 peer-reviewed papers — the first live extraction-v3 issue: every empirical note carries Hypotheses / Propositions, Data & Measures, and Key Findings with three additional verbatim evidence anchors, audited under rubric v2 (nine prose fields per note). All 13 extractions validated on the first attempt; the final audit scan has 117 / 117 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. Two audit-loop repairs before release: one note narrowing (Marr: per-study status-loss manipulation attribution) and one tooling fix — the audit prompt's references-stripper false-matched a line-initial prose sentence ("References to relationships…", Koerner) and cut the Discussion from the audit input; the stripper gained a prose-guard plus regression tests, and Koerner re-audited cleanly against the full text. The release also raises the audit text budget to 240K chars with anchor-aware splicing (every evidence anchor's context is guaranteed visible to the auditor), migrates both CrossRef false-positive registries to JSON data files (tools/known_crossref_issues.json, tools/known_compound_surnames.json), adds a --model provenance flag to the ingest tools (this issue's notes record claude-fable-5), and adds anti-overreach scope guidance to the extraction prompt. Manifest lint and scoped CrossRef pass 13/13 with zero new false-positive entries; SQLite, CSV, and BibTeX all contain 1,141 records.

• v0.40.0 v3 backfill batch 07 (2026-07-14, 1,167 notes): Upgrades AMJ volume 67, issue 1 and volume 66, issue 6, 24 notes total to the v3 schema — all v2 augmentations. The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 934 v2, 172 v3. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 216 / 216 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, 0 CONTRADICTED (all 24 notes overall pass), after three evidence-based repairs cleared in a single round — all three on legacy prose fields written before the v3 standard, each an "applied extension" scope-drift narrowed back to the paper's own scope: a practical implication that had added an unstated "platform designers / category curators" audience (Soublière), a future- research field carrying two unframed added directions (Park, coevolutionary lock-in), and a limitation asserting a single-firm generalizability caveat that the paper's own Generalizability section argues against (Jia). A fresh independent re-audit of the three repaired notes returned every field SUPPORTED. library.bib regenerated byte-identical (the all-augmentation signature — only SQLite and CSV changed).

• v0.39.0 v3 backfill batch 06 (2026-07-12, 1,167 notes): Upgrades AMJ volume 67, issue 3 and issue 2, 21 notes total to the v3 schema — all v2 augmentations. The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 958 v2, 148 v3. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 189 / 189 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, 0 CONTRADICTED (all 21 notes overall pass), after two evidence-based repairs cleared in a single round — one new-field precision fix to Data & Measures (Preston relabeled Study 1's 135/51 White/Black split from "predecessors" to the appointed coaches, matching the paper's own count) and one legacy-limitations fix (Compagni's invented "Steering Committee meetings", a specific absent from the paper text). A fresh re-audit of the two repaired notes returned every field SUPPORTED. library.bib regenerated byte-identical (the all-augmentation signature — only SQLite and CSV changed).

• v0.38.0 v3 backfill batch 05 (2026-07-12, 1,167 notes): Upgrades AMJ volume 67, issue 5 and issue 4, 18 notes total to the v3 schema — all v2 augmentations. The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 979 v2, 127 v3. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 162 / 162 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, 0 CONTRADICTED (all 18 notes overall pass), after four evidence-based repairs cleared in a single round — two new-field precision fixes to Data & Measures (the Hallila fixed-effect-set list and the Marti 260-statement attribution) and two legacy future-research scope trims (Bliese's invented method-pairing examples and Madsen's measurement call, which the paper frames as a limitation, not a stated direction). A fresh re-audit of the four repaired notes returned every field SUPPORTED. library.bib regenerated byte-identical (the all-augmentation signature — only SQLite and CSV changed).

• v0.37.0 v3 backfill batch 04 (2026-07-12, 1,167 notes): Upgrades AMJ volume 68, issue 1 and volume 67, issue 6, 18 notes total to the v3 schema — all v2 augmentations. The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 997 v2, 109 v3. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 160 / 162 prose-field verdicts SUPPORTED, 0 UNSUPPORTED, 0 CONTRADICTED (all 18 notes overall pass). These two issues' older v2 notes carried an unusual amount of scope drift in their practical-implication and future-research fields, so eighteen evidence-based repairs across two rounds narrowed added audiences, prescriptions, and examples back to the papers' own scope (and fixed three factual slips). Two residual PARTIALs were accepted and documented rather than edited: the Hagtvedt future-research note lightly extends a stated limitation, and the Trzebiatowski data-measures note carries a source-ambiguous turnover lag that is internally consistent with the paper's methods — editing either to appease the auditor would violate the faithful-note rule.

• v0.36.0 v3 backfill batch 03 (2026-07-11, 1,167 notes): Upgrades AMJ volume 68, issue 3 and issue 2, 16 notes total to the v3 schema — all v2 augmentations (the AMJ v1 re-extraction tier was cleared in batch 02, so batches 03+ are pure augmentation). The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 1,015 v2, 91 v3. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 144 / 144 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, 0 CONTRADICTED, after two evidence-based scope repairs (the Fitzsimons and Knight practical implications were narrowed to the paper's own stated audience and claims).

• v0.35.0 v3 backfill batch 02 (2026-07-11, 1,167 notes): Upgrades AMJ volume 68, issue 5 and issue 4, 19 notes total to the v3 schema (11 legacy v1 notes fully re-extracted, 8 v2 notes augmented in place). The record total is unchanged at 1,167; the version-tier census shifts to 61 v1, 1,031 v2, 75 v3, and one re-extracted note (Lazar) was reclassified empirical-mixed → empirical-quantitative. Each touched note passed a fresh full independent 9-field rubric-v2 audit: 170 / 171 prose-field verdicts SUPPORTED, 0 UNSUPPORTED, 0 CONTRADICTED, after five evidence-based repairs (per-study support-pattern and scope corrections on Chung, Dutta, and Preston). One residual PARTIAL remains, on the Li note's Limitations: its "30% and 7%" mediation figures are verbatim-faithful to the paper, but two-column typesetting splices that sentence into the reference list, so the independent auditor could not re-verify the 7% — the faithful sentence was left unchanged rather than edited to appease the extraction artifact.

• v0.34.0 v3 backfill batch 01 (2026-07-11, 1,167 notes): Upgrades the two most recent AMJ issues to the v3 schema — AMJ volume 69, issue 1 and volume 68, issue 6, 17 notes total (16 legacy v1 notes fully re-extracted, 1 v2 note augmented in place). This adds Hypotheses / Propositions, Data & Measures, and Key Findings to every note in both issues; the record total is unchanged at 1,167 (a backfill adds no papers). Each touched note passed a fresh full independent 9-field rubric-v2 audit: 153 / 153 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, 0 CONTRADICTED, after three evidence-based repairs (Peng and Lee key-findings over-generalizations, and the Mahringer note's practical implication narrowed to the paper's own scope). The version-tier census shifts accordingly: 72 v1, 1,039 v2, 56 v3.

• v0.33.0 sweep (2026-07-11, 1,167 notes): Adds Academy of Management Journal volume 57, issue 3 — 13 peer-reviewed papers produced under extraction v3 and independently audited under rubric v2. One audit-identified field was repaired from PDF evidence before release: Crossland's Theoretical Contribution had listed CEO tenure among the constructs the paper's discriminant-validity analysis distinguished career variety from (the tests covered openness to experience, risk propensity, age, and education, not tenure). A fresh re-audit cleared the repair; the final issue scan has 117 / 117 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. All 13 notes validate, the manifest lints cleanly, and scoped CrossRef verification matches all seven checked bibliographic fields for every paper. SQLite, CSV, and BibTeX were rebuilt sequentially and each contains 1,167 records.

• v0.32.0 sweep (2026-07-10, 1,154 notes): Adds Academy of Management Journal volume 57, issue 2 — 13 peer-reviewed papers produced under extraction v3 and independently audited under rubric v2. Three audit-identified fields were repaired from PDF evidence before release: Beckman's board-measure timing, Rogan's agency-level robustness qualification, and Wang's patent-count-mediated indirect pathway. Fresh re-audits cleared each repair; the final issue scan has 117 / 117 prose-field verdicts SUPPORTED, 0 PARTIAL, 0 UNSUPPORTED, and 0 CONTRADICTED. All 13 notes validate, the manifest lints cleanly, and scoped CrossRef verification matches all seven checked bibliographic fields for every paper. SQLite, CSV, and BibTeX were rebuilt sequentially and each contains 1,154 records.

Run the audit on a single note. Layer 1 (the mechanical anchor check) runs standalone; the full two-layer audit reads an independent auditor's verdict:

# Layer 1 only (mechanical evidence-anchor check):
python tools/audit_note.py notes/<paper_id>.md --skip-layer-2

# Full two-layer audit (Layer 2 verdict supplied by an independent auditor):
python tools/audit_note.py notes/<paper_id>.md \
  --layer-2-json incoming/_audits/<paper_id>.layer2.json

Or, from inside a Claude Code session: /audit-note <paper_id>.

What's in this snapshot

This main-branch snapshot contains 1,167 curated notes:

• NBS 2026-02 — 77 notes distilled from the Network for Business Sustainability (NBS) February 2026 monthly research digest. (62 notes shipped in v0.2.0; 15 previously-missing papers were recovered and added in v0.3.0.)
• NBS 2025-12 — 82 notes from the NBS December 2025 monthly digest, spanning the Ecological Economics biodiversity-and-finance special issue, AMJ, Business & Society, Review of Finance, Strategic Management Journal, and other journals (added in v0.20.0).
• NBS 2026-01 — 113 notes from the NBS January 2026 monthly digest, led by the Journal of Business Ethics (40) with Research Policy (14), the Journal of Environmental Economics and Management, Organization Science, Human Relations, The Journal of Finance, MIS Quarterly, and other journals (added in v0.21.0).
• AMJ pilot — 895 notes across 70 recent issues of the Academy of Management Journal (vol. 57 no. 1 through vol. 57 no. 3, vol. 58 no. 1 through vol. 58 no. 6, vol. 59 no. 1 through vol. 59 no. 6, vol. 60 no. 1 through vol. 60 no. 6, vol. 61 no. 1 through vol. 61 no. 6, vol. 62 no. 1 through vol. 62 no. 6, vol. 63 no. 1 through vol. 63 no. 6, vol. 64 no. 1 through vol. 64 no. 6, vol. 65 no. 1 through vol. 65 no. 6, vol. 66 no. 1 through vol. 66 no. 6, vol. 67 no. 1 through vol. 67 no. 6, vol. 68 no. 1 through vol. 68 no. 6, and vol. 69 no. 1). v0.33.0 added vol. 57 no. 3 (13 notes); v0.32.0 added vol. 57 no. 2 (13 notes); v0.31.0 added vol. 57 no. 1 (13 notes, the first extraction-v3 issue); v0.29.0 added vol. 58 no. 1-6 (78 notes); v0.28.0 added vol. 59 no. 1-6 (95 notes); v0.27.0 added vol. 60 no. 1-6 (95 notes); v0.26.0 added vol. 61 no. 1-6 (93 notes); v0.25.0 added vol. 62 no. 1-6 (76 notes); v0.24.0 added vol. 63 no. 1-6 (74 notes); v0.23.0 added vol. 64 no. 1-6 (75 notes); v0.22.0 added vol. 65 no. 1-6 (77 notes); v0.19.0 added vol. 66 no. 1 (13 notes); v0.18.0 added vol. 66 no. 2 (13 notes); v0.17.0 added vol. 66 no. 3 (11 notes); v0.16.0 added vol. 66 no. 4 (11 notes); v0.15.0 added vol. 66 no. 5 (12 notes); v0.14.0 added vol. 66 no. 6 (13 notes); v0.13.0 added vol. 67 no. 1 (11 notes); v0.12.0 added vol. 67 no. 2 (11 notes); v0.11.0 added vol. 67 no. 3 (10 notes); v0.10.0 added vol. 67 no. 4 (9 notes); v0.9.0 added vol. 67 no. 5 (9 notes); v0.8.0 added vol. 67 no. 6 (8 notes); v0.7.0 added vol. 68 no. 1 (10 notes); v0.6.0 added vol. 68 no. 2 (8 notes); v0.5.0 added vol. 68 no. 3 (8 notes); v0.4.0 added vol. 68 no. 4 (8 notes); the original pilot covered the remaining three issues (28 notes).

Paper type	Count
empirical-quantitative	668
empirical-qualitative	247
empirical-mixed	111
editorial	72
conceptual	54
review	9
book-review	6
Total	1,167

All notes have passed the semantic audit. The corpus contains 61 legacy v1 notes, 934 v2 notes, and 172 v3 notes; v2/v3 notes carry an evidence: anchor block checked by Layer 1, and v3 notes add Hypotheses / Propositions, Data & Measures, and Key Findings. See Faithfulness audit above.

Repository layout

management-research-notes/
├── README.md                          ← you are here
├── AGENTS.md                          ← tool-agnostic entry point for AI agents
├── LICENSE                            ← MIT
├── CITATION.cff                       ← cite-this-repo metadata
├── CLAUDE.md                          ← rules for any Claude Code session in this folder
├── docs/
│   ├── extraction-prompt.md           ← the canonical extraction prompt (v3)
│   └── audit-rubric.md                ← rubric the Layer 2 auditor uses
├── notes/                             ← 1,167 curated paper notes (the source of truth)
│   └── nbs-2026-02-spoor-2026.md
├── index/                             ← derived views, all rebuildable
│   ├── synapse.db                     ← SQLite + FTS5 (~19 MB)
│   ├── papers.csv                     ← flat tabular export
│   ├── library.bib                    ← BibTeX, one @article per note
│   └── topics.json                    ← 14-domain controlled vocabulary
├── tools/                             ← the Python pipeline
│   ├── pdf_to_text.py                 ← PDF → plain text (pdftotext + pdfplumber)
│   ├── populate_manifest.py           ← Tier 3 gate: backfill vol/issue/pages + fix year (CrossRef)
│   ├── lint_manifests.py              ← structural manifest lint (catches name-capture bugs)
│   ├── prepare_paper.py               ← bundle a paper for extraction
│   ├── ingest_batch.py                ← walk a folder of PDFs
│   ├── validate_note.py               ← verbatim-abstract + bib + taxonomy + anchors
│   ├── audit_note.py                  ← two-layer faithfulness audit (Layer 1 + Layer 2)
│   ├── verify_metadata.py             ← Tier 2 gate: cross-check bib fields vs CrossRef (verify_years.py = year-only alias)
│   ├── build_index.py                 ← rebuild SQLite from notes/
│   ├── export_csv.py                  ← rebuild papers.csv
│   └── export_bibtex.py               ← rebuild library.bib
├── library/
│   ├── NBS/2026-02/
│   │   ├── manifest.tsv               ← trusted bibliographic source for the batch
│   │   └── missing.tsv                ← papers NBS listed but PDFs unavailable
│   └── AMJ/vol-58-no-1 ... vol-69-no-1/
│       └── manifest.tsv               ← per-issue manifests for the AMJ pilot
│       (pdfs/ and text/ are intentionally NOT published — see Copyright below)
└── .synapse/
    ├── config.yaml                    ← validator policy + custom-field schema
    └── schema.sql                     ← SQLite schema

Quick start

git clone https://github.com/binqi20/management-research-notes.git
cd management-research-notes

# Optional: rebuild the SQLite index from the notes
python tools/build_index.py

# Query the library — answers come back in milliseconds
sqlite3 index/synapse.db "SELECT COUNT(*) FROM papers;"
sqlite3 index/synapse.db \
  "SELECT paper_type, COUNT(*) FROM papers GROUP BY 1 ORDER BY 2 DESC;"
sqlite3 index/synapse.db \
  "SELECT p.id, p.title FROM papers p JOIN topics t ON t.paper_id = p.id
   WHERE t.topic = 'circular-economy' ORDER BY p.year DESC;"

# Full-text search across abstracts and analytic fields
sqlite3 index/synapse.db \
  "SELECT id FROM papers_fts WHERE papers_fts MATCH 'stakeholder AND legitimacy';"

# Re-export BibTeX or CSV
python tools/export_bibtex.py
python tools/export_csv.py

Optional dependencies (for ingesting new PDFs locally)

brew install poppler            # gives you pdftotext
pip install pyyaml pdfplumber   # pyyaml: validator; pdfplumber: extraction fallback

The validator (tools/validate_note.py) reads the corresponding extracted text file under library/<source>/<issue>/text/ to perform its verbatim-abstract substring check. That folder is intentionally absent from the public repo because the text is derived from copyrighted PDFs — see below. For a public clone, the SQLite, CSV, and BibTeX queries above all work without it.

For AI agents

If you are an AI agent (or writing one) and want to consume or contribute to this knowledge base, start with AGENTS.md — it is the tool-agnostic entry point covering:

• the data formats you can consume without running any tooling (Markdown notes, SQLite + FTS5, CSV, BibTeX),
• the 7 hard rules any agent must follow when creating or editing notes,
• the faithfulness guarantees this library carries (every abstract verbatim-anchored, every prose field audited, zero contradictions),
• and how to cite the knowledge base if your agent surfaces a note.

For Claude Code-specific operational conventions (slash commands, subagent dispatch, tool names), see CLAUDE.md as well.

Copyright and licensing

The code, the curated notes, the topic taxonomy, and the derived indexes in this repository are released under the MIT license and are © 2026 Binqi Tang.

The original journal articles that the notes summarize are NOT included. They remain the copyright of their respective publishers and authors. To reproduce a note end-to-end, obtain the source PDF through legitimate channels (your institution, the publisher, or the author's preprint), drop it into incoming/, and run the pipeline locally.

Each note contains a short verbatim abstract (treated as fair-use quotation/commentary) and an original analytic distillation written by the author of this repository. Notes cite the original work via DOI in every file's frontmatter and APA citation block.

How to cite

This repository ships with a CITATION.cff, which GitHub uses to render a "Cite this repository" button in the right sidebar. It will give you both APA and BibTeX automatically. Or, manually:

@software{tang_mgmt_research_notes_2026,
  author       = {Tang, Binqi},
  title        = {Management Research Notes: A File-Based Academic Knowledge
                  Base for Management and Business Sustainability Research},
  year         = {2026},
  version      = {0.40.0},
  doi          = {10.5281/zenodo.19564336},
  url          = {https://doi.org/10.5281/zenodo.19564336},
  license      = {MIT}
}

If you use a specific note, please also cite the original paper using the DOI in that note's APA citation block — the analytic distillation is a commentary, not a substitute.

Roadmap

This project is intended to be a long-running research-infrastructure project, not a one-shot data drop. The near-term roadmap:

• More monthly batches — extend NBS coverage with each new digest and continue the AMJ backfill into earlier volumes, keeping paper IDs stable across updates.
• Additional journal sources — add Web of Science exports and journal RSS feeds as parallel library/{source}/{issue}/ trees, reusing the same pipeline.
• Audit layer hardening — cross-model auditing (run Layer 2 with a second model vendor), automated re-audit on prompt changes, and a lightweight public audit-summary CSV so consumers can see which prose fields carry PARTIAL verdicts without needing to regenerate.
• Vector search — only after the library passes ~5,000 notes; the SQLite FTS5 index is plenty for now.
• Community contributions — issues and pull requests welcome from collaborators who want to share extraction prompts, topic taxonomies, or curated subsets.

Acknowledgements

• The Network for Business Sustainability (NBS) for curating the monthly research digest that seeds this library.
• Anthropic Claude Code for the extraction-and-validation workflow that drives the ingestion pipeline.
• OpenAI Codex for the issue-level extraction and independent-audit orchestration used in the AMJ volume 64 backfill.
• The authors of every paper in notes/ — without their original scholarship, there is nothing to distill.

---

Maintained by Binqi Tang. MIT licensed. Issues and PRs welcome.