# Wiki Log

Append-only. Newest entries at the bottom.
Format: `## [YYYY-MM-DD] <op> | <title>` where `<op>` ∈ `ingest | query | lint | refactor | bootstrap`.
Grep tip: `grep "^## \[" wiki/log.md | tail -10`.

---

## [2026-04-13] bootstrap | Wiki initialised
- Created schema `CLAUDE.md` defining layers, folder layout, page conventions, operations (ingest/query/lint) and rules of engagement.
- Created `wiki/` skeleton: `sources/`, `concepts/`, `entities/`, `frameworks/`, `methods/`, `syntheses/`, `assets/`.
- Created `index.md` with full inventory of uningested sources (~60 files across 7 categories).
- Created `log.md` (this file) and `overview.md` placeholder.
- Ready for first ingest.

## [2026-04-13] refactor | Schema: English wiki, Danish chat; stub sources allowed
- Updated `CLAUDE.md`: wiki content is written in English; chat narration defaults to Danish.
- Added convention: referenced-but-not-ingested sources may be registered as stubs with `status: referenced-not-ingested`.

## [2026-04-13] ingest | Agentic BPM: A Research Manifesto (Calvanese et al. 2026)
- Created source page [[sources/2026-calvanese-agentic-bpm-manifesto]] (35-pp preprint, arXiv:2603.18916v1).
- Created 7 concept pages: [[concepts/agentic-bpm]], [[concepts/framed-autonomy]], [[concepts/process-awareness]], [[concepts/explainability-apm]], [[concepts/conversational-actionability]], [[concepts/self-modification]], [[concepts/perceive-reason-act]].
- Created 18 author entity pages + 1 hub entity ([[entities/wil-van-der-aalst]]).
- Created 3 framework pages ([[frameworks/bdi-agents]], [[frameworks/fipa]], [[frameworks/tropos-i-star]]) and 1 method page ([[methods/agent-oriented-process-mining]]).
- Pre-registered 4 referenced sources as stubs: [[sources/2023-dumas-ai-augmented-bpms]], [[sources/2012-vanderaalst-process-mining-manifesto]], [[sources/2024-kampik-large-process-models]], [[sources/2023-chapela-campa-augmented-process-execution]].
- Updated `index.md`: Sources/Concepts/Entities/Frameworks/Methods populated; source moved out of Uningested.
- Updated `overview.md`: APM adopted as backbone ontology for the wiki; working thesis v1 documented.
- Total pages created: 1 source + 4 source-stubs + 7 concepts + 19 entities + 3 frameworks + 1 method = **35 new pages**.

## [2026-04-13] refactor | Schema: textbook two-step pattern + autonomous batch mode
- Added *Ingest patterns* section to `CLAUDE.md`:
  - **Textbook two-step pattern** — meta-ingest for the whole book, chapter deep-dives on demand.
  - **Autonomous batch mode** — skip per-source confirmation when the user grants scoped autonomy; prefer minimal-sufficient reads; create shared concept pages; report in batches; close with a synthesis.

## [2026-04-13] ingest | Fundamentals of Business Process Management 2nd ed. (Dumas, La Rosa, Mendling, Reijers 2018)
- Meta-ingest per textbook pattern. Read: title, foreword, preface, full TOC, figure list. Did *not* read the 519-page body; chapter deep-dives are on demand.
- Created source page [[sources/2018-dumas-fundamentals-of-bpm]] with chapter index.
- Created 8 concept pages: [[concepts/business-process]], [[concepts/bpm-lifecycle]], [[concepts/process-architecture]], [[concepts/process-discovery]], [[concepts/process-model-quality]], [[concepts/devils-quadrangle]], [[concepts/process-aware-information-system]], [[concepts/bpm-maturity]].
- Created 3 framework pages: [[frameworks/bpmn]], [[frameworks/cmmn]], [[frameworks/dmn]].
- Created 5 method pages: [[methods/process-discovery-methods]], [[methods/flow-analysis]], [[methods/process-simulation]], [[methods/process-redesign-heuristics]], [[methods/process-mining-basics]].
- Created 4 entity pages: [[entities/marcello-la-rosa]], [[entities/jan-mendling]], [[entities/hajo-reijers]], [[entities/michael-rosemann]].
- Updated [[entities/marlon-dumas]] to link Fundamentals as full source (no longer uningested).
- Total new pages this ingest: **21**.

## [2026-04-13] batch-begin | PPM corpus autonomous ingest
- User granted autonomy over `raw/Predictive process monitoring/` (~40 papers).
- Strategy: minimal-sufficient read per paper (abstract + intro + conclusions); shared concept pages for cross-cutting abstractions; one source page per paper; synthesis at close.

## [2026-04-13] ingest | PPM shared scaffolding (concepts)
- Created concept pages for the PPM domain before source pages — for reuse across many papers:
  - [[concepts/predictive-process-monitoring]] (hub), [[concepts/next-activity-prediction]], [[concepts/remaining-time-prediction]], [[concepts/outcome-prediction]], [[concepts/trace-encoding]], [[concepts/lstm-ppm]], [[concepts/transformer-ppm]], [[concepts/inter-case-ppm]].

## [2026-04-13] ingest | PPM corpus (33 unique papers from 38 raw files)
Papers ingested as individual source pages (minimal-sufficient reads — primarily abstract + intro):
- **Pre-deep-learning era (2008–2016):** van Dongen 2008, Lakshmanan 2013, Ceci 2014, Polato 2014, Leontjeva 2015, Metzger 2015, Ferilli 2016, Unuvar 2016, Verenich 2016 (doctoral proposal), Teinemaa 2016 (structured+unstructured), Di Francescomarino clustering-based.
- **Deep learning debut (2017):** Tax 2017, Evermann 2017, Navarin 2017, Senderovich 2017 (inter-case), Di Francescomarino 2017 (a-priori), Verenich 2017 (white-box flow analysis).
- **Neural refinement (2018–2020):** Schönig 2018, Di Francescomarino 2018 (GA-HPO), Verenich 2018 (Apromore), Borkowski 2018 (failure), Hinkka 2019, Nolle 2019 (BINet), Jalayer 2020 (attention), Rama-Maneiro 2020 (DL survey/benchmark), Verenich 2019 (remaining-time survey/benchmark).
- **Transformers & beyond (2021–2023):** Bukhsh 2021 (ProcessTransformer), Dumas 2021 (keynote), Duong 2023 (GNN), Cao 2023 (explainable gated RNN), Wang 2023 (MTLFormer), Berti 2023 (OCEL 2.0).
- **Supporting:** Smith 2018 (DL HPO), Lee 2019 (NSGA-II tutorial).
- 5 duplicate PDFs collapsed into single source pages: tax2017 / "Predictive Business Process Monitoring with LSTM Neural Networks.pdf" (identical); ProcessTransformer (two copies); Apromore paper (Springer version + QUT preprint); Verenich survey (arXiv + ACM TIST).
- Created 11 entity pages for recurring authors: [[entities/ilya-verenich]], [[entities/chiara-di-francescomarino]], [[entities/fabrizio-maggi]], [[entities/irene-teinemaa]], [[entities/arik-senderovich]], [[entities/chiara-ghidini]], [[entities/joerg-evermann]], [[entities/jana-rebecca-rehse]], [[entities/geetika-lakshmanan]], [[entities/merve-unuvar]], [[entities/timo-nolle]].

## [2026-04-13] synthesis | PPM landscape
- Wrote [[syntheses/ppm-landscape]] — cross-cutting synthesis of the 33-paper PPM corpus.
- Organises the literature into four eras (classical probabilistic → deep-learning debut → neural refinement → Transformers & beyond) + survey/benchmark triad + standards/supporting.
- Identifies four design axes (target × architecture × encoding × data perspective).
- Connects PPM to APM capabilities (Recommend role, adaptation triggers, explainability, OCEL 2.0 substrate).
- Flags external sources not yet in `raw/` for future acquisition (Teinemaa 2019 outcome benchmark; Camargo 2019 data-aware LSTM; prescriptive-PPM literature).

## [2026-04-13] refactor | overview.md updated
- Resolved two of the three open questions logged at prior ingest via [[syntheses/ppm-landscape]] and [[sources/2018-dumas-fundamentals-of-bpm]].
- PPM moved out of "Uningested sources" in `index.md`.

## [2026-04-13] batch-close | PPM corpus autonomous ingest complete
- Total new pages this batch (Fundamentals meta-ingest + PPM): **~80**.
  - Fundamentals: 1 source + 8 concepts + 3 frameworks + 5 methods + 4 entities = 21
  - PPM scaffolding: 8 concepts
  - PPM sources: 34 (33 unique papers, including two Apromore raws → one source page)
  - PPM entities: 11
  - Synthesis: 1 ([[syntheses/ppm-landscape]])
- Wiki now has 2 fully-ingested foundation sources (APM Manifesto + Fundamentals textbook), 33 PPM papers, 4 referenced stubs, 23 concepts, 34 entities, 6 frameworks, 6 methods, 1 synthesis.

## [2026-04-13] ingest | ABPMS Manifesto upgraded from stub (Dumas et al. 2023)
- Source acquired via arxiv:2201.12855 + ACM TMIS metadata (PDF not yet in `raw/`).
- Upgraded [[sources/2023-dumas-ai-augmented-bpms]] from stub → full source page.
- Captured: full 14-author list (added Accorsi, Fahland, Gal, Völzer, Weber as not-yet-entitised authors), four core ABPMS characteristics (adaptable / proactive / explainable / context-sensitive), ABPMS↔APM capability mapping table.
- Logged a to-do: acquire PDF + create entity stubs for Fahland/Gal/Völzer/Weber/Accorsi if they recur.
- Adopted convention "ingested via public metadata" for sources known well enough from external metadata to write a substantive page even when no PDF is locally stored.

## [2026-04-13] synthesis | ABPMS → APM evolution
- Filed [[syntheses/abpms-to-apm-evolution]] — side-by-side of the two manifestos, with mapping tables (characteristics→capabilities, frame terminology, author overlap), what changed and what didn't, and a one-sentence summary.
- Linked from [[overview]] (working thesis) and [[index]] (Syntheses section).
- This now becomes the canonical reference inside the wiki for the "ABPMS→APM" narrative axis.

## [2026-04-13] synthesis | APM problem landscape mapped to 12 business themes
- Filed [[syntheses/apm-business-themes]] — each theme grounded in existing theory (GDPR/EU AI Act, Basel/COSO, TAM/UTAUT, Devil's Quadrangle, JTBD/HTA, Fitts/Sheridan LoA, MAPE-K, Porter, RBV, TCE, etc.) before APM's named challenges (F/X/A/M/C-series) are introduced as the proposed response.
- Includes a cross-matrix table mapping 12 business themes to APM challenges and to the classical BPM approach that no longer suffices.
- Primary audience: legal, risk, CFO, CISO, HR, CIO, strategy — translates the research agenda to organisational concerns.
- Linked from [[index]] under Syntheses.

## [2026-04-15] ingest | Dumas Fundamentals of BPM Ch. 5 (Process Discovery)
- Chapter deep-dive per two-step textbook pattern; back-links to 2018 meta source.
- Extracted interview-guide material: Interview→Modeling→Validation cycle, backward/forward strategies, 45/15 structured/free-form balance, sunny-day pitfall, rainy-day questions derived from exception taxonomy.
- Workshop roles (facilitator, modeler, scribe) and session cadence (3–5 sessions, sticky-note session 1, BPMN primer session 2).
- Updated [[methods/process-discovery-methods]] with concrete guide content; previous round-robin/depth-first framing was not in Dumas ch.5 — replaced with the textbook's actual phases.

## [2026-04-15] query | interview structuring for process-model outcomes
- Filed as [[syntheses/interview-structuring-for-process-models]].
- Scanned Dumas ch. 6 (qualitative analysis) and Salmon 2010 HTA paper to complement ch. 5 material.
- Ch. 6 contribution: §6.1 step-level decomposition via observation/interview; §6.3.1 stakeholder categories as interview-roster structure.
- HTA paper (Salmon 2010) judged less directly useful than hoped — framework comparison rather than interview guide; left as referenced-not-ingested.
- Dumas §5.8 referenced further reading on interview technique (Sharp & McDermott, Jeston & Nelis, Berg & Lune, Rosemann's 22 pitfalls) — all stubbed as referenced-not-ingested in the synthesis.

## [2026-04-15] ingest | vom Brocke et al. 2021 — CAMAS Method (Context-Aware BPM)
- New source: [[sources/2021-vombrocke-context-aware-bpm-camas]] (BISE 63(5):533–550). Downloaded from Springer to raw/Process Frameworks & BPM/.
- New concept: [[concepts/context-aware-bpm]] — contingency-theory framing of BPM choices.
- New framework page: [[frameworks/bpm-context-framework]] — vom Brocke 2016 (3 dimensions × 12 factors). Status: source-referenced-not-ingested for the 2016 paper itself.
- New entity: [[entities/jan-vom-brocke]].
- Connects to qualitative-discovery thread: CAMAS gives a formal way to choose between Dumas's evidence/interview/workshop discovery methods given a context. Cross-linked from [[methods/process-discovery-methods]] and [[syntheses/interview-structuring-for-process-models]] *(linkage noted in CAMAS source page)*.
- Headline empirical finding: most existing BPM methods are general-purpose; explorative BPM is under-served — relevant for [[overview]] open questions later.

## [2026-04-15] sync | bulk ingest from Playbook_research/literature
- Added recurring sync rule to CLAUDE.md: scan `/Users/mikeriess/Documents/Playbook_research/literature/` after every iteration; copy + ingest new files in autonomous batch mode.
- 7 new files copied to raw/ (CAMAS was duplicate, skipped).
- 8 sources ingested in parallel via 5 sub-agents:
  - **ABPS bundle** (3): Fournier 2025 (observability), Calvanese 2025 (autonomy/goals/frames), Elyasaf 2025 (self-modifying ABPS) — all AutoBiz/Dagstuhl 25192 outputs operationalising the 2026 APM manifesto.
  - **vom Brocke 2016** — upgraded `frameworks/bpm-context-framework` from referenced-not-ingested to full source. 14-factor morphological box; ternary scale (CAMAS later collapsed to binary). Viewpoint article — first empirical use is CAMAS 2021.
  - **DECLARE manual (Pesic 2008)** — new `frameworks/declare` + `concepts/declarative-process-modelling`. Constraint templates (existence/relation/chain/negation), LTL→FSA semantics, three-valued runtime monitoring.
  - **Ottensooser 2012** — empirical: written use cases > BPMN for untrained readers; UC-then-BPMN order maximises comprehension. Strengthens "translate to NL for validation" rule in interview-structuring synthesis.
  - **Riess 2026** (×2): process-automation-schools poster + paradigm-matrix methods → new `concepts/process-automation-schools`.
- New concepts (9): agent-process-observability, behavioral-variability, causal-process-discovery, normative-frame, abps-autonomy-levels, mape-k-loop, aleatoric-vs-epistemic-uncertainty, declarative-process-modelling, process-automation-schools.
- New entities (6): yuval-david, yves-lesperance, andrea-matta, achiya-elyasaf, estefania-serral-asensio, theresa-schmiedel.
- Open: OCED, "WHY in Business Processes" (Fournier companion), Acitelli/Alman/Maggi/Marrella 2025 planning-based framing — referenced-not-ingested.

## [2026-04-15] query | qualitative discovery method selection matrix
- Filed as [[syntheses/qualitative-discovery-method-selection-matrix]] for use in user's flowchart design.
- Consolidates Dumas §5.2.1–5.2.4 Tables 5.1, 5.2 + §5.2.4 question-box terminators + §5.2.3 workshop logistics + §5.2.2 interview logistics.
- Splits Dumas's "evidence-based" into 3 distinct selection candidates (document analysis, passive observation, active observation) since they have different applicability conditions.
- Adds CAMAS-aware extension mapping context profiles to method dominance.

## [2026-04-16] ingest | Alec Sharp 2014 BPTrends columns (×2)
- Fetched legitimately from BPTrends public archive in response to user query: "do we have concrete interview guide examples?" (answer was no until these were added).
- Files copied to Playbook_research/literature/Process Frameworks & BPM/ and synced to raw/.
- New sources:
  - [[sources/2014-sharp-using-scope-models]] — TRAC (Trigger/Result/Activities/Cases), Process Summary Chart, 90-min facilitated scoping sessions, 3 analysis approaches.
  - [[sources/2014-sharp-whats-wrong-with-this-process]] — 3-question initial assessment (problems/environment/consequences), **concrete stakeholder-specific question packs** (customer/performer/owner), TTT goal format, six-enablers final assessment.
- New entity: [[entities/alec-sharp]].
- New concept: [[concepts/six-enablers-framework]] (Sharp/Burlton).
- Updated [[syntheses/interview-structuring-for-process-models]] §1.2–1.3 with TRAC pre-sketch and Sharp's stakeholder-lens question packs (this was the user's original "do we have concrete interview guides?" ask — now resolved in both source and synthesis).
- Updated [[syntheses/qualitative-discovery-method-selection-matrix]] with new §7 "Sharp's 90-minute facilitated scoping session" as a distinct method-selection candidate alongside Dumas's 3–5 session workshop cadence.
- Updated [[methods/process-discovery-methods]] with practitioner complement section.

## [2026-04-16] ingest | Salmon et al. 2010 — HTA vs CWA
- Source: [[sources/2010-salmon-hta-vs-cwa]] (TIES 11(6)).
- Ingested for its observation-protocol value in BPM discovery: HTA's 8-field task-decomposition template (Kirwan & Ainsworth 1992) is directly reusable as observation field-note structure.
- New concept: [[concepts/hierarchical-task-analysis]] — mapped to BPM discovery (HTA procedure ↔ Dumas's 5-step modelling method; HTA iteration ↔ Interview→Modeling→Validation cycle).
- Reserved for user-request: slide structures for workshop + observation methods (in chat, not filed yet).

## [2026-04-16] query | discovery method session templates (interview / workshop / observation)
- Filed as [[syntheses/discovery-method-session-templates]] — three parallel session templates in slide-ready format, consistent structure.
- Interview template incorporates validation feedback on user's slide: rainy-day block, backward traversal, activity-level (not "sub-process level") terminology.
- Workshop template synthesises Dumas §5.2.3 + Sharp TRAC + Sharp six-enablers enabler-walk.
- Observation template leverages HTA's Kirwan & Ainsworth field-note template + Dumas §5.2.1 passive/active mode distinction.
- Cross-template notes: all three share the iteration discipline, the sunny-day pitfall, and the "never step level" granularity rule.

## [2026-04-20] ingest | Coverage-audit batch: 19 sources across 4 thematic clusters (parallel sub-agents)
- User requested full ingest of all uningested files except `playbooks/` and `mixed/` (CRM + SPC) which are deferred for joint review.
- Dispatched 4 parallel sub-agents by topical cluster.
- **AI Capabilities & Adoption (7 sources):** Korst-Wharton adoption survey, Shen AI-skill-formation paper, Shen Anthropic coding-skills post (webloc → WebFetch), Becker-METR RCT, Xu-TheAgentCompany benchmark, Handa-economic-tasks, Tomlinson-Working-with-AI. New concept hubs: [[concepts/ai-adoption]], [[concepts/ai-skill-formation]], [[concepts/ai-agent-benchmarks]]. 13 new entity pages (first authors + co-leads). Note: first sub-agent attempt failed on image-dimension limit; retried with `pdftotext` extraction strategy.
- **Literature Review Methodology (5 sources):** Kitchenham 2007 (canonical), Qureshi 2023, Agarwal 2024 LitLLMs, Dennstaedt 2024, Scherbakov 2025. New: [[methods/systematic-literature-review]] (hub), [[concepts/llm-assisted-literature-review]]. 5 entities. Back-linked PPM review sources (Verenich 2019, Rama-Maneiro 2020) and `overview.md`.
- **Process Frameworks & BPM (3 sources):** van der Aalst 2011 *Process Mining* meta-ingest (second major textbook, chapter-index + deferred deep-dives); Slack *Operations Management* textbook (adjacent-discipline anchor); Bao & Lee 2007 *Process Control: Passive Systems Approach* (chem-eng — flagged as out-of-scope/terminological-disambiguation). 4 new concepts: [[concepts/conformance-checking]], [[concepts/operational-support]] (PPM foundation), [[concepts/lasagna-spaghetti-processes]], [[concepts/process-mining-spectrum]]. [[entities/wil-van-der-aalst]] upgraded (2011 book promoted from "referenced" to authored-works-in-wiki).
- **Task Analysis & JTBD (4 sources):** Christensen HBR 2016 paper, Ulwick/Strategyn ODI whitepaper, Christensen Institute 3-considerations graphic, Christensen Institute JTBD theory webpage. New: [[frameworks/jtbd]] (hub), [[concepts/outcome-driven-innovation]]. 2 entities (Christensen, Ulwick). Back-links added to `business-process`, `process-discovery-methods`, `qualitative-discovery-method-selection-matrix` §11.
- Totals: **19 source pages + 9 concepts + 1 framework + 1 method + 20 entities = 50 new pages**.
- Updated `index.md`: moved 19 sources out of "Uningested"; added new subsections (adjacent disciplines, Task analysis & JTBD, AI capabilities & adoption, Literature review methodology); removed stale HTA entry from Uningested; remaining Uningested = `mixed/` (CRM, SPC) + `playbooks/` (2 files), deferred per user direction.
- Open lint items flagged by sub-agents: Ulwick whitepaper year 2009 *(unverified)*; Christensen Institute webloc treated as 2025; possibly rename 2011-vanderaalst slug to 2016 (second edition); broken link `sources/2016-teinemaa-outcome-ppm-review` referenced by `concepts/outcome-prediction`; worth considering `syntheses/` page on three-way "process control" disambiguation (chem-eng / SPC / BPM conformance).

## [2026-04-20] ingest | Causation + Riess research line + PrPM stragglers — 3 parallel agents
- User confirmed Philosophy of Science as a new thematic cluster coupled to causation-adjacent BPM topics; Riess publications to be ingested with dedicated hub + research-arc synthesis; 3 PPM stragglers from prior coverage-audit to be cleaned up. Playbooks + `mixed/` deferred for joint review.
- Dispatched 3 parallel sub-agents. No agent wrote to `index.md` / `log.md`; parent consolidated.
- **Philosophy of Science — PHI403 Causation in Science (Anjum & Rocca, NMBU PhD course):** 1 meta-source + 20 per-lecture pages = 21 sources; 11 concepts ([[concepts/causation]] hub, [[concepts/causal-pluralism]], [[concepts/regularity-theory-of-causation]], [[concepts/interventionist-theory-of-causation]], [[concepts/dispositionalism]], [[concepts/mechanisms-in-science]], [[concepts/probabilistic-causation]], [[concepts/rct-limitations]], [[concepts/methodological-pluralism]], [[concepts/evidence-hierarchy]], [[concepts/philosophical-bias]]); 2 entities ([[entities/rani-lill-anjum]], [[entities/elena-rocca]]); NMBU entity updated. Course-code reconciliation: filenames `PHI302` but actual code `PHI403` — wiki content uses PHI403; `raw_path` preserves PHI302 (raw is immutable). Back-links added to [[concepts/outcome-prediction]], [[concepts/predictive-process-monitoring]], [[concepts/aleatoric-vs-epistemic-uncertainty]] (mapped aleatoric/epistemic onto chance/credence), [[concepts/causal-process-discovery]], [[sources/2020-rama-maneiro-deep-learning-ppm-review]] (external-validity caveat via Cartwright & Hardie L11 + L19), [[sources/2021-dumas-process-mining-2-from-insights-to-action]] (philosophical bite — prescriptive monitoring as interventionism).
- **Riess research line:** 5 new sources (Riess 2022 metaheuristics survey in JDIM; Riess 2023 NMBU PhD thesis meta-source — 4 essays, Supervisor Scholderer, Paper III referenced-not-ingested; Riess 2023 NMI temporal-loss paper; Riess 2024 SIMULATION SynBPS; Riess & Jørgensen 2025 BRAGE Norwegian customer-service LLM benchmark). New hub [[entities/mike-riess]] + [[entities/telenor]] + [[entities/ntnu]]. Synthesis page [[syntheses/riess-research-arc]] (~750 words) ties 7-paper arc around three-axis evaluation (accuracy/earliness/temporal consistency), drift-as-lifecycle, simulation-as-control-variable-instrument, operational-framing. New concepts [[concepts/business-process-simulation]] + [[concepts/concept-drift]]. Updates to [[concepts/remaining-time-prediction]], [[concepts/predictive-process-monitoring]], [[concepts/prescriptive-process-monitoring]], [[concepts/lstm-ppm]], [[methods/process-simulation]]. "Cited by" back-links added to 6 prior PPM sources.
- **PPM / PrPM stragglers (3):** [[sources/2005-reijers-limanmansar-best-practices-bpr]] (Omega, 29 BPR heuristics × devil's quadrangle), [[sources/2022-kubrak-prescriptive-ppm-slr]] (PeerJ CS *Quo vadis?*, 37 methods × 6 dimensions), [[sources/2014-groger-prescriptive-analytics-bpo]] (Springer LNBIP, early rBPO recommendations). New concepts [[concepts/prescriptive-process-monitoring]] (canonical PrPM hub), [[concepts/bpr-heuristics]], [[concepts/intervention-policy]]. 7 entities ([[entities/selma-liman-mansar]], [[entities/kateryna-kubrak]], [[entities/fredrik-milani]], [[entities/alexander-nolte]], [[entities/christoph-groger]], [[entities/university-of-tartu]], [[entities/university-of-stuttgart]]). [[entities/hajo-reijers]] + [[entities/marlon-dumas]] updated. [[concepts/devils-quadrangle]] now cites Reijers & Liman Mansar 2005 as origin; [[methods/process-redesign-heuristics]] cites directly; [[syntheses/ppm-landscape]] updated (PrPM open question partially resolved).
- **Totals: 29 source pages + 16 concepts + 1 synthesis + 12 entities = 58 new pages.** Plus ~24 existing pages updated with inbound links / citations.
- Cross-cluster synthesis (for `overview.md`): PrPM ↔ interventionist theory of causation — Kubrak's call for "causality + second-order effects" maps directly to Woodward; RCT limitations (L19) apply to PPM benchmark external validity; dispositionalism (L9) echoes context-aware BPM's "no-one-size-fits-all"; SynBPS provides the control-variable infrastructure that addresses exactly the external-validity gap Cartwright & Hardie diagnose for policy RCTs.
- Open lint items: (a) `[[sources/2016-teinemaa-outcome-ppm-review]]` still orphan-wikilinked from multiple pages but page does not exist; (b) `[[concepts/llm-benchmarking]]` referenced from BRAGE but not yet a page; (c) `[[entities/joachim-scholderer]]` referenced from Riess PhD thesis page *(unverified)*; (d) Selma Liman Mansar post-2005 affiliation (CMU Qatar) marked *(unverified)*.

## [2026-04-20] refactor | Retraction: two 2026 "Riess" sources + derived concept removed
- User identified `sources/2026-riess-paradigm-matrix-methods` and `sources/2026-riess-process-automation-schools-poster` as LLM-generated PDFs incorrectly attributed to Mike Riess — **not his work**. Retraction scope: the two fake sources + the concept page `concepts/process-automation-schools` whose entire content (10-paradigm × 3-tier taxonomy; 8-scenario × paradigm matrix) was derived from them.
- **Files deleted (3):** `sources/2026-riess-paradigm-matrix-methods.md`, `sources/2026-riess-process-automation-schools-poster.md`, `concepts/process-automation-schools.md`.
- **Files substantially rewritten (2):** [[syntheses/riess-research-arc]] — reframed from 7-paper to 5-paper arc (2022 survey → 2023 NMI temporal-loss → 2023 PhD thesis → 2024 SynBPS → 2025 BRAGE); removed the "2026 Paradigm taxonomy" section and the "Gap: no unified theory of automation" paragraph; retitled "Evaluation Rigour as Research Programme"; replaced the paradigms cross-reference with a philosophy-of-science bridge ([[concepts/rct-limitations]], [[concepts/interventionist-theory-of-causation]]). [[entities/mike-riess]] — removed 2 frontmatter source entries, the "Process-automation paradigm taxonomy" research-theme section, and 2 rows from the publications table.
- **Files trimmed (6 pages with inbound links to the deleted concept):**
  - [[index.md]] — removed 2 source entries + 1 concept entry (3 lines).
  - [[concepts/ai-adoption]] — removed trailing related-concepts bullet.
  - [[concepts/agentic-bpm]] — replaced the "Position in the paradigm landscape" bullet with a neutral "APM extends ABPMS" framing grounded in the ABPMS → APM lineage already in the wiki.
  - [[concepts/context-aware-bpm]] — removed cross-reference bullet.
  - [[concepts/lasagna-spaghetti-processes]] — reworded the spaghetti→agentic mapping to reference [[concepts/agentic-bpm]] + [[concepts/framed-autonomy]] directly; removed the related-row cross-ref.
  - [[methods/process-mining-basics]] — replaced the "near-universal entry point" opening sentence with a neutral framing situating process mining within the [[concepts/bpm-lifecycle]] monitoring phase.
  - [[sources/2007-slack-operations-management]] — reworded Lean/JIT open-question bullet to drop the taxonomy reference.
- **Historical log entries preserved** (append-only convention): entries on 2026-04-20 documenting the original ingest of these sources remain as historical record. This retraction entry supersedes them for current-state purposes.
- **Consequences absorbed:** the 10-paradigm × 3-tier automation taxonomy and the 8-scenario matrix are no longer part of the wiki's knowledge base. A legitimate paradigm synthesis — built from real ingested sources (Slack OM, Dumas, van der Aalst, the AI capabilities & adoption cluster) — remains a possible future task per user intent.
- **Sanity check:** zero residual references to the retracted slugs across `wiki/` confirmed post-edit.

## [2026-04-20] query | Three study sketches at the APM × Riess-research-line intersection
- User asked for concrete study skeletons at the intersection of the APM manifesto agenda and the Riess research line. Produced three `syntheses/study-sketch-*.md` pages, each with motivation + gap, RQs, hypotheses (testable), method (design, data, instruments, analysis), validity threats, deliverables and venue candidates. All claims forward-cited to existing ingested sources — no new external research introduced.
- [[syntheses/study-sketch-synbps-apm]] — extends SynBPS with LLM-agent hooks + frame DSL + trajectory logger; factorial evaluation of autonomy × process-variability × drift. Primary venue: BPM 2027 main track. Philosophical grounding: Cartwright-Hardie external validity (PHI403 L11).
- [[syntheses/study-sketch-temporal-consistency-agents]] — generalises Riess 2023's three-axis evaluation (accuracy/earliness/TC) from remaining-time to full PrPM recommendation space; compares LLM-agents to LSTM/Transformer baselines on the same BPIC logs. Primary venue: Nordic Machine Intelligence (continuity) or ICPM. Philosophical grounding: Popper risky-predictions (PHI403 L18).
- [[syntheses/study-sketch-agent-trajectory-drift]] — defines agent-induced drift taxonomy beyond Riess 2022 process-oriented types; conformance-checking + BINet-style detection; MAPE-K adaptation policies. Two-pronged data: SynBPS-APM controlled + Telenor observational. Primary venue: BPM 2027 industry track or DSS. Philosophical grounding: same-cause-same-effect (PHI403 L5).
- Index updated; no source-level changes.

## [2026-04-21] query | LLM+BPM reading list synthesis
- User asked for a curated list of LLM+BPM sources (ingested + referenced) for a literature-review reading list.
- Produced [[syntheses/llm-bpm-reading-list]] organising the corpus into five layers: A) core ingested LLM+BPM (APM manifesto, ABPMS, autonomy, self-modifying ABPS, Fournier observability, TheAgentCompany, BRAGE, SynBPS), B) referenced-not-ingested priorities (Kampik 2024 Large Process Models, Acitelli 2025 frame-synthesis, Kampik & Okulmus SIGNAL, Dong 2024 AgentOps, Chopra 2018 normative-MAS, Clio, PM Manifesto, PM 2.0), C) adjacent LLM-SLR cluster flagged separately (Qureshi, Agarwal, Dennstädt, Scherbakov), D) AI-adoption context (Handa, Tomlinson, Korst, METR, Shen), E) entry points + ten corpus-acknowledged gaps + suggested reading order.
- Index updated with synthesis entry.
- No new source pages created; no stubs added beyond those already in the wiki. Kampik 2024 remains the highest-priority referenced-not-ingested.

## [2026-04-21] query | APM manifesto core-messages synthesis
- User asked for a thorough overview of the core messages in the APM manifesto, going deeper than the existing source-page summary.
- Sub-agent read the full 35-page PDF ([[sources/2026-calvanese-agentic-bpm-manifesto]]) and produced a structured analytical report covering central thesis, paradigm shift, two-level architecture (macro/micro), four-capability ordering with rationale, 24 challenges thematically grouped (F/X/A/M + C), distinctions the paper insists on, positioning against prior work, explicit disclaimers, and a critical read of strengths and thin spots.
- Filed as [[syntheses/apm-manifesto-core-messages]]. Complements [[syntheses/abpms-to-apm-evolution]] (which covers the 2023→2026 paradigm shift) by analysing APM on its own terms.
- Index updated.
- No new source pages or concepts created; all wikilinks resolve to existing content. Critical read flags practitioner-facing gaps (frame formalism, LLM-reasoning integration, migration strategy, liability, killer use cases) as candidates for future ingest targets.

## [2026-04-21] sync | Position-paper folder → raw/ + autonomous batch ingest (9 PDFs)
- User authorised scan of `/Users/mikeriess/Documents/Process-automation-position-paper/literature/Literature/` and ingest of PDFs not yet in `raw/`. Compared by filename+size per schema: 16 duplicates skipped, 9 new PDFs copied to matching subfolders.
- **Copied (9):** ABPS×5 — `AI-Augmented Business Process Management Systems.pdf`, `Agentic Business Process Management Systems.pdf`, `Practitioner Perspectives on Agent Governance in Business Processes.pdf`, `Towards eXplainable Autonomous Business Processes.pdf`, `Why We Need First-Class Abstractions for Goals and Normative Frames.pdf`. Process Frameworks & BPM×3 — `Large Process Models.pdf`, `From process mining to augmented process execution.pdf`, `Process Mining Manifesto.pdf`. Task Analysis & JTBD×1 — `Jobs to Be Done Theory - Christensen Institute.pdf`.
- **Disambiguated 2 title-overlap cases by reading title pages:**
  - `Agentic Business Process Management Systems.pdf` = **NEW paper** (Dumas, Milani, Chapela-Campa 2026, arXiv:2601.18833v1, 3-author keynote from 2025 Workshop on AI for BPM). **NOT** the same as the 18-author APM Manifesto. Introduces the Agentic BPM Pyramid.
  - `Why We Need First-Class Abstractions for Goals and Normative Frames.pdf` = **alternate PDF version** of existing [[sources/2025-calvanese-autonomy-business-process-execution]] (740 kB camera-ready vs 1.16 MB preprint; same authors, same PMAI'25 paper). Added as `raw_path_alt` on the existing page.
- **Stub upgrades (3):** [[sources/2024-kampik-large-process-models]], [[sources/2023-chapela-campa-augmented-process-execution]], [[sources/2012-vanderaalst-process-mining-manifesto]] — full source pages with frontmatter, 200–500-word summaries, key_claims, connections. `status: referenced-not-ingested` removed. The wiki's "Referenced (not yet ingested)" section is now empty.
- **Metadata-only source enriched:** [[sources/2023-dumas-ai-augmented-bpms]] — `raw_path` now points to the local PDF; to-do section replaced with an open-follow-ups note (5 un-entitised co-authors; deep re-read of named research challenges still open).
- **New source pages (3):** [[sources/2026-dumas-agentic-bpms-pyramid]], [[sources/2025-fettke-explainable-autonomous-business-processes]], [[sources/2025-vu-practitioner-perspectives-agent-governance]].
- **New concept page (1):** [[concepts/agentic-bpm-pyramid]] — 4-layer Descriptive → Predictive → Prescriptive → Agentic BPM classification; orthogonal to the APM manifesto's capability stack.
- **New entity pages (4):** [[entities/david-chapela-campa]], [[entities/henrik-leopold]], [[entities/hoang-vu]], [[entities/nataliia-klievtsova]].
- **Trivial raw_path addition:** [[sources/2025-christensen-institute-jtbd-theory]] — PDF now in raw/ (previously webloc-only).
- **Totals: 9 PDFs copied + 3 stub upgrades + 1 metadata enrichment + 3 new sources + 1 concept + 4 entities + 2 raw_path updates = 23 page-level operations.**
- Index updated: all 3 former stubs moved into `Ingested — BPM foundations`; Dumas 2023 ABPMS relocated there from `Ingested via public metadata`; 3 new 2025-2026 sources + 4 entities + 1 concept added; `Referenced (not yet ingested)` section now empty.
- **Batch run conditions:** two parallel `general-purpose` subagents executed the ingest; both returned stream-timeout errors mid-reporting but all file writes completed successfully before timeout (verified via tail-check on each new/updated source page — no truncation).
- **Residual items worth logging:**
  - Dumas 2023 ABPMS still needs a deep re-read to enumerate the specific research challenges by name (abstract only mentions them).
  - Five ABPMS co-authors (Rafael Accorsi, Dirk Fahland, Avigdor Gal, Hagen Völzer, Ingo Weber) remain un-entitised — to be created if they recur.
  - Position-paper folder contained 3× `main.pdf` + 1× `Process_automation_position_paper.pdf` — user's own draft/builds, intentionally skipped.

## [2026-04-22] query | Synthesis: BPM phases + BPR legacy (RQ1 + RQ2.1 + RQ2.2)
- User asked three research questions: (RQ1) phases of BPM and where it is heading; (RQ2.1) how Re-engineering (Hammer 1990/1993) lives on today and what has changed; (RQ2.2) what has not changed and what we can learn from 90s BPR experiences.
- Evidence-gathering sub-agent surveyed 19 ingested sources spanning Slack OM 2007 → Dumas 2026 A-BPMS; produced a structured dossier with direct quotations, tensions between sources, and gaps.
- Filed as [[syntheses/bpm-phases-and-bpr-legacy]] (~3000 words). Structure: seven-phase reconstruction (OM roots → BPR → lifecycle → Process Mining → Prescriptive turn → AI-augmentation → Agentic), followed by continuity-and-change analysis pivoted on [[sources/2005-reijers-limanmansar-best-practices-bpr]] as the hinge between Hammer-era BPR and modern BPM. Three synthesis observations: phases inherit rather than replace; anti-universalism is the dominant correction; culture remains the largest failure-mode surface.
- Key framing: no direct Hammer 1990/1993 ingest — BPR is reconstructed via Reijers 2005, Sharp 2014, Dumas 2018. Flagged as gap with specific ingest candidates (Hammer directly, Davenport & Stoddard 1994 critique, Willcocks & Smith).
- No new source / concept / entity pages created; cross-links into 19 existing sources and 9 existing concepts. Index + log updated.

## [2026-04-22] refactor + query | BPR stubs + Kubrak deep re-read + synthesis interlude
- User requested Option B from ingest plan: register BPR primary/critic stubs + close the Kubrak 2022 gap + produce download list.
- **5 Tier-1 stubs registered** under `Referenced (not yet ingested)` in index: [[sources/1990-hammer-reengineering-work]], [[sources/1993-hammer-champy-reengineering-the-corporation]], [[sources/1993-davenport-process-innovation]], [[sources/1993-hall-rosenthal-wade-how-to-make-reengineering-really-work]], [[sources/1994-davenport-stoddard-reengineering-mythic-proportions]]. All carry `status: referenced-not-ingested`, `raw_path: ""`, DOI/URL in frontmatter, and a "Cited from" note pointing to [[syntheses/bpm-phases-and-bpr-legacy]].
- **Kubrak 2022 PrPM-SLR deep re-read** completed by sub-agent. [[sources/2022-kubrak-prescriptive-ppm-slr]] updated with six-dimension classification (performance objective, metric, intervention, modelling technique, data input, intervention policy), expanded key_claims, and a full Connections section.
- **Headline finding from the deep read:** Kubrak 2022 contains **zero direct references to Hammer, BPR, reengineering, Reijers & Liman Mansar, or the devil's quadrangle**. PrPM is historiographically blind to its design-time predecessor. Guiding-by-similarity (nearest-neighbour over past traces) is the modal method, not ML/causal inference. Only one real-world deployment attempt exists in the corpus (Dees et al. 2019), and it failed.
- **[[syntheses/bpm-phases-and-bpr-legacy]] extended** with a new "Interlude — Does PrPM repeat or learn from BPR?" section between the three-observations synthesis and the gaps section. Verdict: *"technically post-BPR, historiographically pre-BPR."* Gaps section updated to retire the Kubrak-as-gap bullet.
- Sub-agent flagged 7 candidates for future creation (not created this pass): `sources/2018-teinemaa-alarm-based-prescriptive`, `sources/2022-fahrenkrog-petersen-fire-now-fire-later`, `sources/2021-bozorgi-prescriptive-cost-aware-cycle-time`, `sources/2021-shoush-dumas-prescriptive-under-resource-constraints`; `concepts/alarm-based-prescription`, `concepts/intervention-discovery`, and possible extension of `concepts/evidence-hierarchy` to cover an in-vivo-validation gap.
- Delivered a tiered download list (Tier 1 primary + Tier 2 BPR→BPM bridges + Tier 3 empirical/culture + Tier 4 agentic-era) to user in chat — not filed as a wiki page.

## [2026-04-27] ingest | BPR Tier-1 batch (3 stub upgrades + 2 new sources) + raw/-rename policy
- User dropped 5 PDFs into `raw/Process Frameworks & BPM/` and `raw/AI Capabilities & Adoption/`. Each ingest delegated to its own background sub-agent (5 parallel agents).
- **3 stubs upgraded to `status: ingested`:**
  - [[sources/1990-hammer-reengineering-work]] — full re-read of HBR Reprint 90406. 7 reengineering principles, Ford 75% AP cut, MBL 30→1 case-manager. Flag: IBM Credit case is NOT in the 1990 article (it's in the 1993 book) — agent did not hallucinate it.
  - [[sources/1994-davenport-stoddard-reengineering-mythic-proportions]] — corrected stub claim from "five myths" to **seven myths** (verbatim: Reengineering's Novelty / Clean Slate / IS Leadership / Reengineering vs. Quality / Top-Down Design / Reengineering as Transformation / Reengineering's Permanence). Empirical base: 200+ company interviews + 35 reengineering initiatives via Jarvenpaa & Stoddard 1993.
  - [[sources/1993-hammer-champy-reengineering-the-corporation]] — 2001 Updated & Revised edition (HarperBusiness). Headline finding from the 2001 prologue + FAQ pp. 239-240: **the authors retrospectively reframe their own thesis — "process" is now the most important of the four key words, not "radical"**. Direct conceptual bridge between BPR and modern continuous BPM.
- **2 new source pages created:**
  - [[sources/1999-attaran-wood-how-to-succeed-at-reengineering]] — mid-decade BPR retrospective with 5 success factors and 5 implementation barriers; Hallmark / IBM Credit / Ford / Liberty Mutual / Wal-Mart cases. Unique value-add: explicit "BPR cannot succeed without a continuous-improvement substrate" claim (proto-BPM-lifecycle move).
  - [[sources/2026-laban-schnabel-neville-llms-corrupt-documents-delegate]] — DELEGATE-52 benchmark (Microsoft Research, arXiv:2604.15597v1). Frontier models corrupt 25% of document content over 20 delegated edits; agentic tool use does NOT improve performance (+6% degradation); Python is the only "ready" domain (RS@20 ≥ 98%) of 52. Direct empirical bound on agentic-BPM autonomy claims.
- **`raw/`-immutability rule relaxed.** User requested that PDFs in `raw/` be renamed to readable kebab-case slugs for discoverability ("Formålet er at jeg selv let skal kunne finde artiklen og læse den"). 5 PDFs renamed (`hammer_reengineering.pdf` → `1990-hammer-reengineering-work-dont-automate-obliterate.pdf`; `249760.pdf` → `1994-davenport-stoddard-reengineering-mythic-proportions.pdf`; `Howtosucceedatreengineering05172016.pdf` → `1999-attaran-wood-how-to-succeed-at-reengineering.pdf`; `Michael%20Hammer...HarperBusiness%20%282001%29.pdf` → `1993-hammer-champy-reengineering-the-corporation-2001-ed.pdf`; `2604.15597v1.pdf` → `2026-laban-schnabel-neville-llms-corrupt-documents-delegate.pdf`). All 5 source pages' `raw_path` updated. `CLAUDE.md` schema updated: file *contents* remain immutable, but filenames may be renamed for readability; ingestion workflow now includes "propose readable slug if filename non-meaningful".
- Index updated: 3 stubs moved from `Referenced (not yet ingested)` → `Ingested — BPM foundations` (year-ordered); Attaran & Wood 1999 added to BPM foundations; Laban et al. 2026 added to AI capabilities & adoption (between Tomlinson 2025 and Shen 2026). Two BPR stubs remain referenced-not-ingested (Davenport 1993 *Process Innovation* book, Hall/Rosenthal/Wade 1993 HBR — PDF for the latter not yet obtained, was confused with Attaran & Wood).

## [2026-04-27] ingest | Legal-cluster Tier-1 batch (6 sources + 10 concepts + 4 entities + 1 synthesis)
- User authorised full ingestion of the curated Legal-cluster after a credibility-driven prune: 9 raw PDFs reduced to 6 (3 vendor-marketing pieces moved to ~/.Trash, see preceding chat exchange). Remaining cluster: 1 high-credibility peer-reviewed paper, 1 primary-source legal text, 2 mellem-høj sources (IAPP opinion, BTLJ blog), 2 lav-mellem (C-Risk vendor overview, BCC26 status snapshot).
- All 6 PDFs read in full (47 pages total: Lazcoz/de Hert 20p, C-Risk 11p, Navaie 7p, Okan 5p, BCC26 3p, GDPR Art 22 1p). No subagents — single-pass read in parallel via Read tool.
- **6 source pages created** under `Ingested — Legal & Regulatory (AI process automation)` in index:
  - [[sources/2023-lazcoz-dehert-humans-in-gdpr-and-aia-governance]] — anchor: doctrinal analysis of human intervention under GDPR Art. 22 + AIA. Reframes Art. 22 as procedural right tied to Art. 5(2) accountability. Two intervention mechanisms (22(1) human-in-the-loop = essential component; 22(2) human-out-of-the-loop on request = safeguard). DPIA as central governance tool.
  - [[sources/gdpr-article-22-text]] — primary source legal text. Citation note: gdpr-info.eu is a private publication; for formal citation use EUR-Lex.
  - [[sources/2025-navaie-iapp-engineering-gdpr-compliance-agentic-ai]] — engineering complement. Four runtime primitives: purpose locks, execution traces, tiered memory governance, live controller/processor mapping. Key insight: execution traces simultaneously serve GDPR Art. 15 + DPIA + AI Act Art. 12/72 — one artefact, multiple regimes.
  - [[sources/2025-okan-btlj-blog-ccpa-vs-gdpr-automated-decision-making]] — US/EU comparison. Three structural divergences (consent opt-in/opt-out, appeal conditional/unconditional, scope narrow/broad). SCHUFA (CJEU C-634/21) as the load-bearing EU enforcement precedent.
  - [[sources/2025-crisk-dora-nis2-overview]] — DORA × NIS2 mapping. Five DORA pillars; NIS2 essential vs important entities; regulation-vs-directive distinction. Vendor content but factually accurate.
  - [[sources/2026-bcc26-eu-digital-regulation-interpretation-to-implementation]] — Jan 2026 status snapshot. AI Act August-2026 milestone, NIS2 17/27-state transposition lag, DORA first-year review, Digital Omnibus Nov-2025 softening of AI literacy.
- **10 new concept pages**: [[concepts/gdpr-article-22]] (anchor) · [[concepts/automated-decision-making]] (hub) · [[concepts/human-oversight]] · [[concepts/dpia]] · [[concepts/eu-ai-act]] · [[concepts/dora]] · [[concepts/nis2]] · [[concepts/automation-bias]] · [[concepts/profiling]] · [[concepts/gdpr-accountability-principle]].
- **4 new entity pages**: [[entities/paul-de-hert]], [[entities/guillermo-lazcoz]], [[entities/keivan-navaie]], [[entities/ilke-okan]].
- **1 new synthesis page**: [[syntheses/legal-risk-mapping-ai-process-automation]] (~2400 words). Structure: 5-layer regulatory stack (GDPR + AI Act + DORA + NIS2 + CCPA) → risk-surface table mapping 7 failure modes to 5 regimes → 4 convergence points (executive accountability, third-party risk, demonstrable evidence, human oversight) → operational fix via Navaie's four primitives → cross-jurisdiction CCPA thread → connections to BPM/agentic-BPM cluster (APM C4 challenge, normative frame, agent-process-observability) → 5-step practical risk-mapping protocol → gaps + open questions.
- **Schema move**: introduced new index section `Ingested — Legal & Regulatory (AI process automation)` between "AI capabilities & adoption" and "Literature review methodology". Contains 6 sources, year-ordered. Eligible for further population as the cluster grows (next Tier-1 candidates: SCHUFA judgment, Italian Foodinho/Deliveroo DPA decisions, ENISA NIS2 transposition tracker, final adopted CCPA ADMT rules).
- Cross-cluster bridges established: legal cluster ↔ APM cluster (synthesis explicitly grounds APM Manifesto's challenge C4); legal cluster ↔ observability cluster (Navaie's traces = Fournier's observability traces); legal cluster ↔ Telenor entity ([[entities/telenor]] Responsible AI principles flagged as adjacent — now cross-referenceable when that source is ingested from raw/Responsible AI/).
- **Open follow-ups for future ingest passes**:
  - Dedicated source page on CJEU SCHUFA (C-634/21, Dec 2023) — load-bearing across multiple sources in this cluster.
  - AIA Recital + Article numbering shifts (2021 proposal → 2024 final Reg.) flagged in Lazcoz/de Hert source page.
  - AI Act × DORA interaction — gap noted in C-Risk source page; not addressed in any current cluster source.
  - ENISA NIS2 transposition tracker as authoritative source for per-MS picture.
- **Totals**: 6 source pages + 10 concept pages + 4 entity pages + 1 synthesis + 4 index sections updated + 1 log entry = 26 page-level operations.

## [2026-05-04] ingest | Process-modelling quality & verification cluster (3 sources + 5 concepts + 3 entities + 1 synthesis)
- User dropped 3 PDFs into `raw/` to support Path-B (rigorous) build of an exhaustive evaluation guide for BPMN models. All three peer-reviewed primary sources, validated against title pages.
- **Filename rename + relocation** per `raw/`-rename policy: moved from flat `raw/` into `raw/Process Frameworks & BPM/` with readable slugs:
  - `7PMG - Mendling et al (2010).pdf` → `2010-mendling-reijers-vanderaalst-7pmg.pdf`
  - `Verification of Workflow Nets - Aalst.pdf` → `1998-vanderaalst-verification-of-workflow-nets.pdf`
  - `Process_models_representing_knowledge_for_action_a.pdf` → `2006-krogstie-sindre-jorgensen-revised-sequal-framework.pdf`
- All three PDFs read end-to-end (~50 pages total; per Read-tool extraction). No subagents — single-pass parallel read.
- **3 source pages created** under new index section `Ingested — Process modelling quality & verification`:
  - [[sources/1998-vanderaalst-verification-of-workflow-nets]] — formal anchor: WF-net definition (i, o, strongly-connected when extended), three-clause soundness (option to complete · proper completion · no dead transitions), Theorem 11 (sound iff extended net live + bounded), Theorem 12 (free-choice ⇒ polynomial-time decidable), 8 transformation rules T1–T4 + composites T5 preserving soundness, almost-free-choice extension.
  - [[sources/2010-mendling-reijers-vanderaalst-7pmg]] — operational pragmatic-quality toolkit. G1 size, G2 routing degree, G3 single start/end, G4 structuredness, G5 avoid OR, G6 verb-object labels, G7 decompose >50. Empirical foundation: 600 SAP EPCs + 2000 industrial EPCs + n=73 understanding study + n=29 label-style experiment. Industry-workshop priority elicitation (21 modellers) but theoretical priority scheme out of scope.
  - [[sources/2006-krogstie-sindre-jorgensen-revised-sequal-framework]] — semiotic theoretical scaffold. 8 quality types from Stamper's ladder (physical/empirical/syntactic/semantic/perceived-semantic/pragmatic/social/organisational); revised version adds D^O (optimal domain) and K^N (knowledge-need), splits semantic quality into descriptive (M vs D, as-is) + prescriptive (M vs D^O, to-be) + perceived (K vs M); active-model framing (articulation D→M + activation M→D); pragmatic quality redefined as enabling learning + action.
- **5 new concept pages**:
  - [[concepts/soundness]] (anchor — formal three-clause definition + variants)
  - [[concepts/workflow-net]] (formal substrate)
  - [[concepts/7pmg]] (operational checklist)
  - [[concepts/sequal-framework]] (theoretical scaffold)
  - [[concepts/process-model-complexity-metrics]] (bottom-up metrics: |N|, |A|, density, CFC, mismatch, depth, cross-connectivity)
- **3 new entity pages**: [[entities/john-krogstie]], [[entities/guttorm-sindre]], [[entities/havard-jorgensen]] (NTNU/SINTEF axis — SEQUAL lineage).
- **1 new synthesis page**: [[syntheses/process-model-quality-and-soundness-evaluation-guide]] (~5500 words). Structure:
  1. Scope/audience (process analyst with as-is BPMN deliverable)
  2. Five-layer quality stack (physical/empirical · notation-syntactic · behavioural-syntactic = soundness · semantic · pragmatic · organisational/social)
  3-8. Per-layer evaluation criteria with primary-source citations
  9. Trade-off table (6 inter-layer tensions)
  10. **LLM-assisted review rubric** (full prompt scaffold + JSON output schema + calibration notes)
  11. Top-12 anti-patterns
  12. Recommended evaluation workflow (soundness before semantic walkthrough)
  13. Acknowledged gaps (8 next-ingest candidates: Lindland 1994 original SEQUAL, Mendling 2008 metrics monograph, Verbeek WOFLAN diagnostics paper, Trcka 2009 data anti-patterns, Reijers-Mendling 2008 modularity, Rosemann 22 pitfalls, Moody 2009 Physics of Notations, Cardoso 2005 CFC)
  14. Quick-reference card (1-page compressed)
- **Updated**: [[concepts/process-model-quality]] — added SEQUAL link as theoretical parent, added soundness section, added complexity-metrics link, added forward link to new synthesis.
- **Index updated**: new sources section + 5 new concept entries + 3 new entity entries + 1 new synthesis entry.
- **Cross-cluster bridges**: quality-cluster ↔ APM cluster (active models prefigure agentic execution); quality-cluster ↔ observability cluster (articulation/activation = read/write loop); quality-cluster ↔ interview-structuring synthesis (semantic-validation methodology shared with Ottensooser-grounded interview guide).
- **Open follow-ups**: 8 next-ingest candidates listed in synthesis §13. Most load-bearing: Lindland 1994 IEEE Software (original SEQUAL); Mendling 2008 metrics; Rosemann 22 pitfalls.
- **Totals**: 3 source pages + 5 concept pages + 3 entity pages + 1 synthesis + 1 concept-page update + 4 index-section updates + 1 log entry = 18 page-level operations.

## [2026-05-04] synthesis | BPMN modelling practical builder's guide
- User followed the evaluation guide with a request for the constructive counterpart: a very-detailed practical guide with requirements and criteria for *building* a BPMN model (not just evaluating one).
- **1 new synthesis page**: [[syntheses/bpmn-modelling-practical-guide]] (~9000 words, 19 sections). Structure:
  1. Inputs required before modelling (TRAC scope, stakeholder roster, source material, modelling-goal taxonomy with quality emphasis per goal, tool selection)
  2. BPMN element catalogue — flow objects, typed events, activity types, connecting objects, swimlanes, artifacts, sub-process variants
  3. Mandatory structural requirements (process boundaries, sequence flow rules, gateway rules, pool/lane rules, soundness preconditions)
  4. Five-step modelling method (Dumas §5.3 distilled into operational order: boundaries → activities → resources → control flow → additional)
  5. Naming and labelling standards — verb-object for activities, state for events, question for XOR, with detailed examples + summary table per element type
  6. Pattern library — 10 patterns: sequence, exclusive choice, parallel split/join, structured loop, multi-instance, event-based choice, boundary events (interrupting/non-interrupting), compensation, message exchange, DMN integration. Each with diagram, rules, anti-patterns.
  7. Quality criteria — concrete thresholds operationalising 7PMG (size, connector degree, structuredness, density, label criteria, decomposition, layout)
  8. Decomposition strategy (when, what — SESE regions, embedded vs reusable, naming hierarchies, anti-patterns)
  9. Data perspective — when and how (data objects vs data stores, associations, state-based naming)
  10. Resource/organisational perspective — pool design, lane design, handoffs, black-box pools, anti-patterns
  11. Exception handling — exception taxonomy, modelling patterns, rules, "what about exceptions" reflex
  12. Multi-perspective integration — BPMN+DMN, BPMN+CMMN, BPMN+DECLARE, when BPMN is wrong tool
  13. Validation workflow — five gates from self-review through process-owner sign-off to audience-fit check
  14. Governance and lifecycle — versioning, repository structure, annotations as metadata, runtime linking, maintenance triggers
  15. Top-20 anti-patterns
  16. Modelling-goal × pattern matrix (decision aid)
  17. Quick-reference checklist (1-page condensed)
  18. Acknowledged gaps (8 next-ingest candidates: OMG BPMN 2.0 spec, Dumas Ch3/Ch4 deep-dives, Mendling 2008 metrics monograph, Reijers-Mendling 2008 modularity, Silver 2011 Method and Style, OMG sample collection, Allweyer 2010, Rosemann 22 pitfalls)
  19. Relationship to sister documents (interview-structuring → builder's guide → evaluation guide chain)
- **Index updated**: synthesis added in alphabetical position (between bpm-phases and discovery-method-session-templates).
- **Cross-document chain established**: this guide is the missing middle of [[syntheses/interview-structuring-for-process-models]] (upstream/elicitation) → builder's guide (this) → [[syntheses/process-model-quality-and-soundness-evaluation-guide]] (downstream/assessment). All three cross-link.
- **Totals**: 1 new synthesis + 1 index update + 1 log entry = 3 page-level operations.

## [2026-05-04] ingest | Dumas Ch3 deep-dive + Ch5 §5.4 expansion + four behavioural anomalies
- User flagged that we hadn't properly mined Dumas's BPMN modelling chapters — particularly §5.4 with its detailed behavioural-anomaly criteria (deadlock as one of several examples). Verified by reading §5.4 in full + Ch3 §3.1–3.2 from `raw/Process Frameworks & BPM/Fundamentals_of_BPM 2018.pdf`.
- **Ch5 source page substantially expanded**: [[sources/2018-dumas-fundamentals-of-bpm-ch5-discovery]] §5.4 section grew from 6-line summary to full deep-dive covering:
  - 5.4.1 Syntactic Quality and Verification — full element-level structural rules table (activities, events, gateways, flows, data associations) + model-level rule + four behavioural anomalies + block structure + Dumas's soundness formulation (compared with Aalst 1997)
  - 5.4.2 Semantic Quality and Validation — validity vs completeness with worked example, validation methods, process-owner approval as normative-character endorsement
  - 5.4.3 Pragmatic Quality and Certification — three usability aspects (understandability, maintainability, learning), two checks (visual-logical consistency + meaningful labels), block-structuring as design-in approach, label-quality common problems
- **6 new concept pages**:
  - [[concepts/deadlock]] — Token stuck; XOR-split → AND-join most common cause; violates soundness (1)
  - [[concepts/livelock]] — Token cycles in loop forever; loop with always-true exit; violates soundness (1)
  - [[concepts/lack-of-synchronization]] — Multiple tokens on same flow; AND-split → XOR-join most common cause; violates soundness (2)
  - [[concepts/dead-activity]] — Activity never executable; provably-false condition or disconnected; violates soundness (3); distinct from operationally-dead (log-based) detection
  - [[concepts/block-structure]] — SESE fragment with matched split-join; sound by construction; corresponds to Aalst transformation rules T1–T4; trade-off: limited expressiveness vs free-choice WF-nets
  - [[concepts/token-semantics]] — BPMN execution semantics via tokens: gateway token rules (XOR routes 1, AND spawns N, OR spawns k≤N), implicit termination, multi-instance reasoning, mapping to Petri-net theory
- **1 new source page (Ch3 deep-dive)**: [[sources/2018-dumas-fundamentals-of-bpm-ch3-essential-process-modeling]] — covers §3.1 First Steps with BPMN (events/activities/arcs, token semantics, naming conventions in detail, Mapping/Abstraction/Purpose framework, conceptual-vs-executable distinction), §3.2 Branching and Merging (XOR/AND/OR with full token semantics), §3.2.3 OR-join active-branch concept + join-type heuristic (XOR → AND → OR fallback), tooling spectrum.
- **Updated existing concept pages**:
  - [[concepts/soundness]] — added behavioural-anomalies cross-reference table; Dumas-vs-Aalst formulation comparison table; updated related-concepts links
  - [[concepts/process-model-quality]] — added full structural-rules section (element-level + model-level) per Dumas §5.4.1; added behavioural-anomaly table; cross-link to block-structure
- **Updated synthesis pages**:
  - [[syntheses/bpmn-modelling-practical-guide]] — §1.4 expanded with conceptual-vs-executable model distinction; §3.5 expanded into full behavioural-correctness section with the four anomalies + block-structure trade-off; §5.1 naming-conventions strengthened with Ch3 specifics (max 5 words, generic-verb avoidance, part-of-speech consistency); §6 pattern library intro adds token-semantics framing + join-type selection heuristic
  - [[syntheses/process-model-quality-and-soundness-evaluation-guide]] §5.3 expanded to detail all four anomalies with concept-page links + appearance patterns + block-structure prevention
- **Index updated**: 1 new source + 6 new concept entries + updated header.
- **Cross-cluster bridges**: BPMN modelling concept-cluster now properly anchored — token-semantics ↔ workflow-net (Petri-net mapping), block-structure ↔ 7PMG G4 (structuredness empirical evidence), four anomalies ↔ soundness (one per clause + one for proper completion).
- **Open follow-ups**: Ch4 deep-dive (Advanced Process Modeling — typed events, exceptions, boundary events, compensation, event sub-processes); Ch3 §3.3 (business objects/resources) + §3.4 (process decomposition) deep-dive.
- **Totals**: 1 new source page + 6 new concept pages + 1 source-page expansion + 2 concept-page updates + 2 synthesis updates + 1 index update + 1 log entry = 14 page-level operations.

## [2026-05-06] ingest | LLM-BPMN modelling cluster (BPMN Assistant + Kampik LPM correction + Varsani neuro-symbolic ERP)
- User dropped 5 PDFs into `raw/ABPS/`: BPMN Assistant, Kampik LPM (full PDF), Kampik LPM correction, Neuro-Symbolic AI in SAP ERP, and a duplicate of Dumas 2023 ABPMS Manifesto (arxiv 2309.00900v3).
- Granted batch autonomy. Deleted duplicate. Renamed 4 raw files to readable slugs (per the raw/-rename policy).
- **3 new source pages**:
  - [[sources/2026-licardo-bpmn-assistant]] — first ingested LLM→BPMN modelling source. JSON IR + 5 atomic editing functions; beats raw XML on all evaluated LLMs (GPT-5.1, Claude 4.5 Sonnet, DeepSeek V3, Llama 3.3 70B, Qwen 2.5 72B, Gemini 2.0 Flash); 43% latency cut, >75% output-token cut. Conformance F1 0.72 vs 0.69. Open source at github.com/jtlicardo/bpmn-assistant.
  - [[sources/2024-kampik-large-process-models-correction]] — Springer erratum to Fig. 1; restores missing labels in the LPM architecture diagram.
  - [[sources/2025-varsani-neuro-symbolic-ai-sap-erp]] — concrete SAP ERP instantiation of LPM-style architecture. LLM (GPT-4 + LLaMA 2) + middleware (RAG + RLHF + schema embeddings) + SAP HANA + ABAP rule engine. Reports 89.6% query-translation accuracy vs 74.8% LLM-only / 71.2% symbolic-only. Year/venue *unverified* — author at Cognex/SAP, IEEE-style preprint without explicit publication date.
- **Existing source page upgraded**: [[sources/2024-kampik-large-process-models]] — added §5 (three benefits: knowledge-task reduction, observability, design/execution/analysis convergence), §6 three-step feasibility roadmap (modelling/analysis augmentation → fusing data → continuous improvement), §7 positioning vs BloombergGPT / ProcessGPT / Vidgof / Berti-Qafari / Klievtsova-Grohs. Updated `raw_path` to renamed file under `ABPS/`. Linked to Correction and to BPMN Assistant as Step-1 instantiation.
- **2 new concept hub pages**:
  - [[concepts/llm-assisted-process-modelling]] — three architectural patterns (direct XML / structured IR / atomic editing functions); validation as first-class layer; quality dimensions per BPMN Assistant comparison.
  - [[concepts/neuro-symbolic-bpm]] — five-layer architectural pattern (data → atom layer → middleware → fine-tuned LLM → classical BPM tooling); two core motifs (grounding + symbolic decision authority); self-correction loops; distinctions vs CoT/RAG/tool-use; LPM vs APM positioning.
- **Updated existing pages**:
  - [[frameworks/bpmn]] — added "LLM-assisted BPMN authoring" section linking BPMN Assistant + concept hubs.
  - [[concepts/agentic-bpm]] — added "Adjacent paradigm: LPM" section linking Kampik, BPMN Assistant, Varsani.
  - [[syntheses/llm-bpm-reading-list]] — promoted Kampik out of stub (now in §A.5 with Correction + Varsani as instantiations); §A.4 dedicated to LLM-driven process modelling (BPMN Assistant); added §B sub-list of L1-L8 LLM-BPMN priorities cited from Licardo et al.; updated §E.2.1 gap to *partially filled*; reading order updated (8 steps, includes neuro-symbolic alternative paradigm + LLM→BPMN modelling).
  - [[wiki/index.md]] — added 3 new source entries + 2 new concept entries; updated header.
- **Gap closure**: §E.2.1 (LLM-based process discovery & redesign) of the LLM+BPM reading list moves from *uncovered* to *partially filled* — BPMN Assistant covers the editing/refinement side; LLM-assisted extraction *from event logs* remains uncovered.
- **Totals**: 3 new source pages + 2 new concept pages + 1 source-page expansion + 3 existing-page updates (1 framework, 1 concept, 1 synthesis) + 1 index update + 1 log entry = 11 page-level operations.

## [2026-05-06] query | BPMN's role in automation projects + BPMN vs flowcharts (filed as synthesis)
- User asked for a synthesis on BPMN's role in automation projects, with a flowchart comparison if sources support it.
- Mapped relevant ingested sources: [[sources/2018-dumas-fundamentals-of-bpm-ch3-essential-process-modeling]] (conceptual↔executable distinction), [[sources/2012-ottensooser-graphical-vs-textual]] (empirical readability — BPMN vs *textual* use cases, not flowcharts), [[sources/2026-licardo-bpmn-assistant]] (LLM-driven executable BPMN), [[sources/2024-kampik-large-process-models]] (BPMN as classical-tooling substrate in LPM), [[sources/2026-calvanese-agentic-bpm-manifesto]] (BPMN as operational specification in APM), [[sources/2026-dumas-agentic-bpms-pyramid]] (automation/autonomy at apex), [[sources/2010-mendling-reijers-vanderaalst-7pmg]] + [[sources/1998-vanderaalst-verification-of-workflow-nets]] (formal semantics).
- **Honest gap stated up front in the synthesis**: wiki has no direct BPMN-vs-flowchart empirical source. Argument hangs on (a) BPMN's documented formal semantics + executable extensions, (b) flowcharts' acknowledged absence of these properties — flagged inline where claims go beyond ingested-source coverage.
- **Filed as new synthesis**: [[syntheses/bpmn-in-automation-projects]]. 8 sections: notation matters → BPMN distinguishing features → automation-stack positioning (BPMS/LPM/APM) → BPMN-related pitfalls → BPMN-vs-flowchart decision matrix → hybrid usage pattern → LLM-era cost shift → wiki gaps.
- **Cross-links added**: [[frameworks/bpmn]] gets a "When to choose BPMN over flowcharts" section pointing to the synthesis; [[wiki/index.md]] gets the new entry.
- **Wiki gaps surfaced** (as candidates for future ingest): direct BPMN-vs-flowchart comparison study; Dumas Ch10 (executable BPMN) deep-dive; BPMS implementation patterns (Camunda/Bizagi/Signavio); BPMN-RPA integration source.
- Totals: 1 new synthesis page + 2 existing-page updates (frameworks/bpmn, index) + 1 log entry = 4 page-level operations.

## [2026-05-11] query | Redesign heuristics — list of the 29 + side-by-side mapping to PrPM
- User asked (Danish) whether the wiki has sources on redesign heuristics, then requested both (a) a wiki-form short overview of the 29 heuristics and (b) a side-by-side synthesis vs. PrPM.
- Extracted Table 1 of [[sources/2005-reijers-limanmansar-best-practices-bpr|Reijers & Liman Mansar (2005)]] (pp. 285–292) and added the **named list of 29 heuristics** to [[concepts/bpr-heuristics]]: 3 Customers · 0 Products · 5 Operation · 4 Behaviour · 7 Org-structure · 4 Org-population · 1 Information · 2 Technology · 3 External Environment = 29. Each entry annotated with framework element, dimensions touched, limits, and the paper's cited application example (Pacific Bell, Ford, IBM Credit, Disney, Bell Atlantic, Hallmark, Duke Power, Taco Bell, Loews, Microsoft, Hammer & Champy cases).
- **Filed new synthesis**: [[syntheses/redesign-heuristics-vs-prpm]]. Maps each of the 29 heuristics onto Kubrak (2022) PrPM intervention classes. Result: 4 direct runtime analogues (triage, resequencing, flexible assignment, specialist–generalist), 6 partial (order types, task elimination, knock-out, exception, order assignment, extra resources, control addition), 19 with no PrPM analogue (all structural / external / informational heuristics). Inverse view: PrPM adds two primitives 2005 could not express — alarm-based prescription and CATE/uplift causal policies; and *loses* the devil's-quadrangle scaffold (no PrPM method in Kubrak's corpus evaluates against the four-way trade-off).
- **Cross-links**: synthesis links to [[syntheses/bpm-phases-and-bpr-legacy]] §Interlude (narrative companion) and [[syntheses/ppm-landscape]]; pulls in [[sources/2014-groger-prescriptive-analytics-bpo]] and [[sources/2026-dumas-agentic-bpms-pyramid]] for the agentic-extension agenda.
- Totals: 1 concept-page expansion (bpr-heuristics with full 29-item list) + 1 new synthesis page + 1 index update + 1 log entry = 4 page-level operations.

## [2026-05-13] refactor | Redo Part D of ai-agent-heuristics-devils-quadrangle.html under stricter constraints
- User asked to redo the Part D paper-angle brainstorm with two filters: (a) strong anchor in existing Riess work, (b) little or no new empiricism — and if neither possible, suggest horizon-expanding alternatives.
- Diagnosed the original six angles against the new constraint: only the devil's-hexagon angle (Angle 2 of v1) survived; the other five (factorial SynBPS, trajectory mining, MOO PrPM, TC-of-frame-respect, Telenor field study) each required significant new measurement work.
- **Replaced Part D entirely** with v2: six new angles + new framing + suggested-ordering + "what if none of these fit" closer. Added an **Empirics** pill to each card to make the budget visible (None / Literature only / One worked example).
- **Anchored-and-low-empiricism (4):** (1) Devil's hexagon as standalone-conceptual paper (anchor: Riess 2023 TC axiom); (2) Agentic-redesign-heuristics SLR (anchor: Riess 2022 drift-survey method, Kitchenham); (3) Formal taxonomy of design-time × runtime levers (anchor: existing [[syntheses/redesign-heuristics-vs-prpm]]); (4) Frame-first redesign methodology paper (anchor: this synthesis + Calvanese frames, one worked example only).
- **Horizon-expanding (2):** (5) Deontic-logic semantics for normative frames (formal-methods venue — Fundamenta Informaticae / LMCS); (6) Causal re-foundation of the devil's quadrangle (anchor: PHI403 Anjum & Rocca causation + Bozorgi causal PrPM). Both tagged HORIZON in the HTML.
- **Fix:** the new HORIZON badges initially used a non-existent CSS variable `--accent-2`; switched to `--tag-I` (burnt orange) which already exists in the file's design tokens.
- **Synthesis MD** updated: `updated:` bumped to 2026-05-13; "Practical artefacts" §HTML overview map gained a sentence describing the v2 Part D brainstorm.
- Totals: 1 HTML asset rewrite (Part D, ~195 lines replaced) + 1 synthesis MD touch + 1 log entry = 3 page-level operations.

## [2026-05-12] query | Agentic-era reinterpretation of the 29 BPR heuristics + pedagogical HTML map
- User (Danish) asked for a synthesis that reinterprets the Reijers & Liman Mansar (2005) heuristics for the AI-agent era, mapped against the devil's quadrangle, plus a pedagogical HTML overview.
- Anchor sources: [[sources/2005-reijers-limanmansar-best-practices-bpr]] (catalogue), [[sources/2026-calvanese-agentic-bpm-manifesto]] (APM capabilities), [[sources/2026-dumas-agentic-bpms-pyramid]] (explicit call for "agentic extension of Reijers & Liman-Mansar", p. 45), [[sources/2025-calvanese-autonomy-business-process-execution]] (normative frames), [[sources/2025-elyasaf-self-modifying-abps]] (adaptation/evolution split), [[sources/2025-fournier-agentic-ai-process-observability]] (trajectory observability), [[sources/2022-kubrak-prescriptive-ppm-slr]] (PrPM precedent — [[syntheses/redesign-heuristics-vs-prpm]] establishes the ~10/29 runtime baseline this synthesis widens).
- **Filed new synthesis**: [[syntheses/ai-agent-redesign-heuristics-2026]]. Structure: (A) re-reads all 29 heuristics with a four-tag scheme (R re-anchored / A amplified / I inverted / N newly run-time) and revised devil's-quadrangle profiles per heuristic; (B) adds 10 brand-new heuristics (B1 Frame-first design, B2 Verification-by-judge, B3 Tool-mediated action, B4 Trajectory observability, B5 Adaptation/evolution separation, B6 Token/inference budget, B7 Confidence-thresholded knock-out, B8 Conversational handoff, B9 Frame-respecting outsourcing, B10 Determinism-by-request); (C) counts the reinterpretations (≈14 A, ≈11 R, ≈5 I, ≈13 N) and argues the binding axis shifts from time/cost to quality-as-trust; (D) three using-principles. Argues the devil's quadrangle survives but its quality axis silently absorbs new sub-dimensions (hallucination, frame-violation, alignment, auditability).
- **Filed companion HTML map**: [[assets/ai-agent-heuristics-devils-quadrangle.html]]. Self-contained pedagogical HTML — inline CSS + SVG devil's quadrangle showing the four agent-era axis refinements + 30 + 10 heuristic cards in a filterable grid (filter by tag and by framework element). Each card shows id, name, element, tag, four-axis impact symbols (↑↑/↑/—/↓/↓↓/~), one-sentence agent-era reading, and the dominant new failure mode. No external dependencies. Color-coded by reinterpretation tag.
- **Cross-links added**: synthesis appears in [[wiki/index.md]] Syntheses section; back-links queued for [[concepts/bpr-heuristics]], [[concepts/devils-quadrangle]], [[concepts/agentic-bpm]], [[concepts/framed-autonomy]]; sibling-synthesis links to [[syntheses/redesign-heuristics-vs-prpm]] and [[syntheses/bpm-phases-and-bpr-legacy]].
- Scope honesty: socio-cultural integration and deployment evidence remain thin (Vu 2025); several "amplified" claims rest on architectural plausibility. Caveat is stated inside the synthesis.
- Totals: 1 new synthesis page + 1 new HTML asset + 1 index update + 4 concept-page back-link updates + 1 log entry = 8 page-level operations.