Revisions

First complete build of the 3-essay thèse par articles for Paris Dauphine-PSL — 328 pages. This is the newest shipped artifact; no dissertation prose has changed since.

• newGeneral Introduction and General Conclusion written: three research questions, the three-essay architecture, the connecting thread
• newFront matter to institutional format — English abstract, French résumé (required by Paris Dauphine-PSL), acknowledgements
• theoryEssay 1 — Foundation Models as Digital Platforms (~85 pp): PEMAK functional equivalence; six binary conditions; complementor engagement as the viability threshold
• dataEssay 2 — Configurational Viability (~120 pp): the N=49 csQCA carried over unchanged; two equifinal paths, Death Zone structurally absent, OPEN sufficient 20/20
• theoryEssay 3 — Accordivity (~90 pp): 9-stage institutional-economics genealogy, Contributivity, the Accordivity Stack, propositions P1–P4, GABMS specification
• buildArchived as revisions/rev-v7.0-de/jxu-mom-20260403-v7.0-de.pdf — 328 pages, zero build errors

source: revisions/rev-v7.0-de/CHANGELOG-v7.0-de.md

A 27-page conference-format cut of the configurational analysis. No changelog was written for this build — this entry is read off the archived PDF.

• buildArchived as revisions/rev-v7.0/jxu-mom-20260401-v7.0-ce.pdf — 27 pages, single-column conference template
• fixNo changelog accompanies this edition; the site reports only what the artifact itself shows

source: revisions/rev-v7.0/jxu-mom-20260401-v7.0-ce.pdf (no changelog written)

Cross-format synchronisation: the monograph and paper editions re-cut from the article baseline.

• fixMonograph methodology and data chapters migrated to vendor-level framing (N=49; 44 Y=1, 5 Y=0)
• methodNew §3.3.5 longitudinal validation (Wayback archive) added and wired into monograph/pdf.yaml
• fixChapter 5 normalised away from legacy "atomic" language and stale metadata; Chapters 6–7 harmonised to the Q1 2026 baseline
• fixRegional Platform Incumbent recipe made consistent across discussion and conclusion: INTEGRATED · HIERARCHICAL · ~MONOPOLY
• pePaper edition front matter re-stamped to v7.0-PE, dated 2026-03-30
• buildArchived: jxu-mom-20260330-v7.0.pdf (328 pp, monograph) and jxu-mom-20260330-v7.0-pe.pdf (106 pp, paper edition)

source: revisions/rev-v7.0/CHANGELOG-v7.0.md, CHANGELOG-v7.0-pe.md

Data model clarification: 109 OpenRouter authors are not 49 FM vendors. Additive only — no QCA result changed.

• methodExplicit FM-vendor definition in §3.2 and in the TypeScript interfaces: trains base weights + releases commercially + deploys via API
• dataSchema: vendors.category (FM/FT/IP/RL/ED) and vendor_qca_mapping.v70_verdict (IN/CANDIDATE/EXCLUDE)
• fix8 phantom slug mappings corrected — z.ai, Moonshotai, Reka, StepFun, Kwai, IBM, ByteDance, Mistral AI; all 49 vendors now join to the models table
• methodOpenRouter's dual role documented: a case in the study (Y=1, T02) and the census instrument (109 authors via the Wayback archive)
• newInfrastructure: audit_vendors.py (HuggingFace verification) and scripts/verify_v70_consistency.sh, which now gates CI
• dataObservation window deliberately unchanged at Q1 2026 — v7.0 clarifies the data model, it does not re-date the census
• buildArchived as revisions/rev-v7.0/jxu-mom-20260330-v7.0-ae.pdf — 27 pages

source: revisions/rev-v7.0/CHANGELOG-v7.0-ae.md

Census finalized at N=49; z.ai consolidation; 12 new vendors; T14/T15 added; Collective/Proprietary Path labels fixed.

• dataCensus finalized: N=49 (44 Y=1 + 5 Y=0) — up from N=40
• newz.ai consolidation: Zhipu AI/GLM + Z-AI merged → z.ai (T06, API brand of Zhipu/THUDM)
• new12 new vendors added: Tencent, Meituan, IBM, AI21 Labs, Black Forest Labs, Kwai, Upstage, Reka, Writer, Stability AI, Xiaomi (+ Aleph Alpha as Y=0)
• newT14 new config: MONOPOLY·INTEGRATED·WIDE·PAID·HIERARCHICAL·OPEN → IBM (1 case)
• newT15 new config: ~MONOPOLY·~INTEGRATED·WIDE·PAID·HIERARCHICAL·OPEN → AI21/BFL/Kyutai/Xiaomi (4 cases)
• newAleph Alpha added as Y=0 (T04 deviant) — strategic pivot away from standalone FM API
• fix"Unbundled specialist" / "Platform aggregator" → "Collective Path" / "Proprietary Path" throughout web app and paper
• data~INTEGRATED: cons=0.919, cov=0.773, n=34; MONOPOLY·WIDE: cons=0.929, cov=0.295, n=13
• dataOPEN=1 → Y=1: 20/20 vendors (was 13/13); overall solution consistency 0.917 preserved

Complementor Engagement framework; N=40; Collective/Proprietary Path naming.

• theoryNew §2.4: Complementor Engagement (CE) theoretical framework — Active Orchestration vs. Ecosystem Generativity
• newCollective Path / Proprietary Path pathway naming throughout
• newAdept AI (T04) added as Y=0 — Amazon acqui-hire June 2024
• newZhipu AI/GLM + 01.AI/Yi added as Y=1 (Chinese open-weight vendors, T06)
• dataN=40 (was N=37); 37 Y=1 + 3 Y=0; overall consistency 0.917
• exclMosaicML excluded — vendor metamorphosis (not org death); MPT ecosystem continues under Databricks
• newChanson-Rocchi structural analogue: CE = API adoption threshold preceding FM viability
• newNet-new bibliography: iansiti2004, nambisan2017

N=34 data alignment, truth table corrections, Adept AI investigation.

• fixN=34 data alignment and truth table corrections across all paper files
• newAdept AI investigation initiated (Amazon acqui-hire evidence review)
• dataN=34; overall consistency 0.939

IOE theoretical grounding; §2.4 propositions; Death Zone formalized.

• theoryIOE theoretical grounding: Williamson (TCE), West/Spolsky (commoditize-your-complement)
• theory§2.4 Propositions P1a/P1b (Proprietary equifinality), P2a/P2b (open heterogeneity), P3 (structurally absent)
• newDeath Zone analysis formalized: ~MONOPOLY · INTEGRATED = empty set, n=0
• newIOE references: hart1995, west2006, spolsky2002, williamson1985, bottcher2022, weking2022

First paper edition: condensed from monograph for journal submission.

• peFirst paper edition (PE) derived from v6.5 monograph
• dataN=33 census (31 Y=1 + 2 Y=0)
• newCondensed to journal-length format; methodology chapter preserved

Research questions aligned to QCA findings; IOE governance pivot completed.

• fixResearch questions (§1.3) revised to map directly to QCA results and six tenets
• newNew question: "Missing Configuration" — why ~MONOPOLY·INTEGRATED is structurally absent
• methodqca_logic_mapping.md: master reference table linking all 6 logical steps across monograph

Overfull resolution; typography cleanup; introduction polished.

• fixZero overfull hbox warnings after systematic table and margin fixes
• fixIntroduction §1.1–§1.4 polished; "Institutional Gap" framing refined

Governance Pivot — IOE as second theoretical pillar.

• theoryNew §2.1.3: IOE lens — Transaction Cost Economics, Property Rights Theory, Agency Theory
• theoryChapter 6 reframed through Institutional Path Dependence and "Ideological Firewall" (US–China)
• newNew §6.5: archetypes as institutional governance forms with policy focus per configuration
• newChapter 7 expanded: Structural Separation, Tiered Licensing, Experimentalist Governance

Reference bugfix and formatting release.

• fixAppendix bibliography corrections; citation type mismatches resolved
• fixHeading level compliance: all L4 headings flattened per style guide

Appendix refactoring; QCA calibration evidence integrated.

• methodAppendix restructured: calibration evidence matrix with three-source triangulation
• newQCA calibration protocol for each of 6 conditions documented with secondary citations

Major methodology revision: constant condition resolved; 25 new citations.

• methodConstant condition problem resolved: HIERARCHICAL condition redefined with proper variance
• new~25 new peer-reviewed citations added across Chapters 2–4
• fixReference type corrections: 7 misclassified entries fixed (Evans2006, Radin2013, etc.)

Paris Dauphine-PSL institutional formatting; narrative polish.

• newTitle page updated to Université Paris Dauphine-PSL institutional format
• fixPEMAK framework positioning refined; economic dimension arguments sharpened
• fix"Titans vs. Champions" and market order narrative polished in Chapter 6

Major Chapter 4–5 restructuring; real data collection; population expanded.

• dataPrimary data collection phase: vendor census expanded with real market data
• methodChapter 4–5 restructured to separate data chapter from analysis chapter
• newPopulation of FM vendors expanded; inclusion/exclusion criteria formalized

Comprehensive methodology documentation; QCA literature review added.

• method§2.3 QCA Methodology Literature Review: Ragin (2008), Fiss (2011), Rihoux & Ragin (2009)
• methodComplete calibration protocol with secondary research citations for all conditions
• newTerminology refinement: condition names standardized to MONOPOLY/INTEGRATED/WIDE/PAID/HIERARCHICAL/OPEN

Atomic csQCA shift — from fuzzy-set to crisp-set methodology.

• methodFROM fsQCA (continuous 0.0–1.0 scores) TO csQCA (binary 0/1 membership)
• methodRationale: increased analytical transparency, reduced calibration subjectivity, simpler replication
• methodThree-source triangulation calibration protocol established

Initial monograph — first complete draft.

• newFirst complete monograph draft submitted
• newSix archetypes identified via QCA; PEMAK framework applied to FM vendors
• dataInitial census N≈30; fuzzy-set QCA analysis