Research Log

2026-07-28 — Tier 1 ingest: Li & Thirumalai (2019) + Jones et al. (2019)

Two Tier 1 papers ingested, verified, and integrated. Li & Thirumalai (2019, J. R. Soc. Interface): non-genetic cooperation-based ITH mechanism via unequal public goods allocation. Jones et al. (2019, Haematologica): myeloma clonal evolution under lenalidomide — 66% branching, dissociation between clinical benefit and clonal impact.

Pages created (2): wiki/sources/li2018-ith-mechanism.md, wiki/sources/jones2019-clonal-evolution-myeloma.md

Pages updated (7): intratumor-heterogeneity.md (new §Ecological cooperation-based ITH), branching-evolution.md (quantitative clinical evidence), population-bottleneck.md (Jones 2019 primary data), therapy-resistance.md (new §Clinical Benefit Without Clonal Selection), neutral-evolution.md (min bar), subclonal-architecture.md (min bar), clonal-sweep.md (min bar)

Verification: Li 2019: verified_with_caveats (preprint→published track, all 5 claims verified). Jones 2019: verified (all 5 claims verbatim). Corpus: Sources 68→70, Verified 53→55/70.

2026-07-28 — Verification: Jamal-Hanjani 2017 + Nova 2010

Two previously unverified source summaries verified. Jamal-Hanjani et al. (2017, NEJM): TRACERx 100 — 10 key claims all verified verbatim against extracted text. Validation already at evidence_level III (prospective cohort). Nova et al. (2010, Nano Letters): spider silk mesoscale modeling — 6 key claims all verified. Evidence level revised from VII (expert opinion) to VI (primary computational/theoretical study in peer-reviewed journal).

Pages verified (2): wiki/sources/jamal-hanjani2017-trackingEvolution_nsclc.md (verification_status: verified, evidence_level: III), wiki/sources/nova2010-silk-hierarchical-mechanics.md (verification_status: verified, evidence_level: VI). Corpus: Verified 55→57/70.

2026-07-18 — Zavala et al. 2022 cross-contamination paper ingested

Ingest of Zavala, Aximu-Petri, Richter, Nickel, Vernot, & Meyer (2022) — Molecular Ecology Resources (DOI: 10.1111/1755-0998.13607). This paper from MPI-EVA systematically characterizes index swapping and cross-contamination in multiplex hybridization capture. Key findings integrated: (1) cross-contamination reaches 7–58% per library under unoptimized multiplex capture (PCR to plateau, 30 cycles); (2) the mechanism is jumping PCR — incomplete extension products cross-hybridizing between libraries; (3) optimization (limited PCR cycles avoiding plateau + two-step PCR at 68°C + individual per-well cycle adjustment) reduces contamination to ≤0.0004% (≤1 sequence per event); (4) even singleplex capture shows sporadic 0.0008–0.021% contamination when libraries are reamplified to plateau; (5) a computational detection method using corner library symmetry in index-pair count matrices; (6) chimeric insert sequences form at up to ~15% of reads when sequence complexity is low (second-round capture).

New concept page: wiki/concepts/sequencing-library-artifacts.md — index swapping, cross-contamination (mechanism + magnitude table + corner library detection method + mitigation strategies + implications for subclonal reconstruction + distinction from FFPE chemical artifacts). Created from single-source synthesis (Zavala 2022), confidence: medium pending ingestion of van der Valk et al. (2020) on index hopping.

Pages updated:

• wiki/concepts/ffpe-pre-analytical-variables.md — Added §Distinction from Library Artifacts (Cross-Contamination) section: chemical vs. physical artifact distinction, additive nature of the two artifact classes, combined noise floor for subclonal variant detection. Added Zavala 2022 source + related link.
• wiki/concepts/variant-allele-fraction.md — Minimum bar: added Zavala 2022 to sources frontmatter + related wikilinks.
• wiki/concepts/subclonal-reconstruction.md — Minimum bar: added Zavala 2022 to sources frontmatter + related wikilinks.
• docs/guides/practical-evolutionary-genomics-roadmap.md — Added Milestone B2.5 (Cross-Contamination Detection) between trimming and alignment: index pair extraction + corner library analysis + contamination estimation + fix actions. Updated wiki-data loop mapping table, convergence timeline, wiki resources section, and external resources. Updated Professor’s Audit to cover cross-contamination awareness.

Verification: Lint clean (0 single-source high, 0 frontmatter issues, 0 broken refs, 0 orphans). KG check skipped (infrastructure/methods paper — no biological triples to contradict).

Verification dispatch (2026-07-18): Source verification agent dispatched via wiki bridge. Result: verified — all 6 claims cross-referenced against extracted PDF verified; DOI resolves (10.1111/1755-0998.13607); Semantic Scholar confirms publication (17 citations, 2022-03-09, author list matches). Evidence level V confirmed. No corrections needed.

Corpus state: Sources: 67→68. Concepts: 52→53. Verified: 53/68. KG triples unchanged.

User directive: “Push back if needed” — acknowledged. No pushback warranted; the paper is methodologically sound (4 independent experiments with internal replication), from the world-leading group in degraded-DNA sequencing methodology, published in a well-regarded journal (MER, IF 7.1), with open-source code and deposited data. Evidence level V (methodology validation study). Integration into roadmap was user-requested and executed.

Pages changed: wiki/sources/zavala2022-cross-contamination.md (new), wiki/concepts/sequencing-library-artifacts.md (new), wiki/concepts/ffpe-pre-analytical-variables.md, wiki/concepts/variant-allele-fraction.md, wiki/concepts/subclonal-reconstruction.md, docs/guides/practical-evolutionary-genomics-roadmap.md, wiki/index.md, wiki/log.md

2026-07-16 — Tier 2.5 olog maintenance

Olog relevance audit and cancer olog update. Identified olog-relevant papers from 7 post-Jul-6 ingestions: Gerlinger 2012 (branched evolution, convergent evolution, spatial mutation distribution), Burns et al. 2013 Nature (APOBEC3B enzymatic mutagenesis in breast cancer), Burns Temiz & Harris 2013 NatGen (APOBEC3B pan-cancer mutagenesis, kataegis). Four new objects added to cancer-evolution-olog: APOBECMutagenesis, Kataegis, ConvergentPhenotype, MutationSpatialDistribution. Two new arrows (generatesDiversity, converges) — both classified as proarrows. One new commutativity condition (CC12: Mutagenesis-Diversity). Extended empirical anchors on MutationalSignature, Mutation, IntratumorHeterogeneity. Counts: objects 33→37, arrows 49→51, commutativity 11→12, proarrows 12→14. Cross-domain functor verification: no impact on F, G, H. Pipeline integration: olog maintenance check added to CLAUDE.md ingest workflow Phase 3 (step 11a). Index updated.

Pages changed: wiki/concepts/cancer-evolution-olog.md, CLAUDE.md, wiki/index.md, wiki/log.md

Reason: Prevent olog freeze — the ingest pipeline now includes olog gap audit after concept synthesis. Catch-up for 10 days of missed updates. Tier 3 rebuild remains deferred; Tier 2.5 incremental maintenance is the operational mode.

2026-07-16 (continued) — Buehler 2010 silk mechanics paper ingested

Source summary created for Nova et al. (2010) spider silk hierarchical mechanics (Nano Letters). Relevance assessment: the paper is not directly about clonal evolution but is a foundational methodological source for the olog/cross-domain functor program. Key findings: (1) silk achieves superior properties through a hierarchical two-phase nanostructure (beta-sheet nanocrystals + semiamorphous matrix), (2) nanocrystal size confinement to ~3 nm is the critical design parameter — larger crystals reduce strength by 67% and toughness by 60%, (3) structural design, not chemical complexity, is the enabling principle. Structural isomorphism with cancer’s dual-regime model documented in detail. Potential fourth cross-domain functor (S: SilkMechanics → CancerEvolution) noted as a research direction in the cross-domain functors page. Cancer evolution olog updated with silk mechanics as a methodological source + structural analogue for dual-regime coupling (CC7).

Pages created: wiki/sources/nova2010-silk-hierarchical-mechanics.md, wiki/sources/jamal-hanjani2017-trackingEvolution_nsclc.md

2026-07-16 (continued) — TRACERx primary cohort ingested

Source summary created for Jamal-Hanjani et al. (2017) — Tracking the Evolution of Non–Small-Cell Lung Cancer (NEJM). The first 100 TRACERx patients, 327 tumor regions. Key findings: (1) CNA heterogeneity predicts survival (HR 4.9, P = 4.4×10⁻⁴) while mutation heterogeneity does not (HR 0.86, P = 0.70); (2) branched evolution and ITH near-universal (median 30% subclonal mutations, 48% subclonal CNAs); (3) temporal ordering of drivers — targetable (EGFR, MET, BRAF) are early/clonal, chromatin/DDR genes are late/subclonal; (4) genome doubling in 76%, early and clonal, enables subsequent CIN; (5) mirrored subclonal allelic imbalance in 62% — direct evidence of ongoing dynamic CIN producing parallel evolution; (6) APOBEC drives late subclonal diversification with 19 tumors carrying subclonal drivers in APOBEC context. Integrated into intratumor-heterogeneity (quantitative ITH benchmarks, CNA vs. mutation prognostic finding, mirroed allelic imbalance), chromosomal-instability (dynamic vs. static CIN distinction), and APOBEC-mutagenesis (subclonal driver induction evidence). Index updated.

Pages changed: wiki/concepts/intratumor-heterogeneity.md, wiki/concepts/chromosomal-instability.md, wiki/concepts/APOBEC-mutagenesis.md, wiki/index.md, wiki/log.md

2026-07-12 — Five missing concept pages created (selection + evolution + FFPE)

Trigger: Filled the four most-referenced missing concept pages plus the FFPE pre-analytical consolidation page. All wikilinks that previously pointed to non-existent pages now resolve.

Pages created (5):

#	Page	Type	Key contribution
1	positive-selection	concept	The evolutionary process by which advantageous mutations (drivers) increase in frequency; s ≈ 0.4% per driver (Bozic 2010); τ_k vs. sweep_time competition; detection via δ-calibrated VAF distribution
2	genetic-drift	concept	Random frequency fluctuations from stochastic sampling; P(survival) ≈ 2s; δ amplifies drift; clonal ≠ selected (clonal passengers fixed by drift at δ ≈ 0.99)
3	negative-selection	concept	Purifying selection; immune-mediated elimination of neoantigen-bearing clones (immunoediting); escape mechanisms (HLA LOH, B2M loss, PD-L1); dN/dS detection
4	linear-evolution	concept	Sequential clonal succession (Nowell 1976); requires τ_k > sweep_time (small N); transitions to branching as N grows; four-mode taxonomy (Turajlic 2019)
5	ffpe-pre-analytical-variables	concept	Five-stage pre-analytical pipeline; three-confounder problem; oxidative damage > deamination; DNA > RNA >> protein hierarchy; ERROR-FFPE-DNA checklist; SMF floor

Cross-page impacts:

• 86 broken wikilinks in blast-radius reduced significantly — [[positive-selection]], [[genetic-drift]], [[negative-selection]], and [[linear-evolution]] were among the most-referenced missing targets
• driver-mutation.md, neutral-evolution.md, branching-evolution.md, clonal-sweep.md, clonal-interference.md all had [[positive-selection]] in their related: frontmatter — now resolved
• ffpe-pre-analytical-variables.md consolidates knowledge previously distributed across variant-allele-fraction.md, APOBEC-mutagenesis.md, and 7 FFPE source summaries

Sources drawn from: 18 unique sources (Nowell 1976, Bozic 2010, Bozic 2016, Greaves & Maley 2012, Turajlic 2019, Gerstung 2020, Graham & Sottoriva 2017, McGranahan & Swanton 2017, PCAWG 2020, Orr 2005, Gerlinger 2012, all 7 FFPE papers)

Verification: All wikilinks resolve; all source files exist; frontmatter complete on all 5 pages; lint output saved. Full L1 lint hung on content hash computation (Phase 5) — manual verification completed instead.

2026-07-12 — FFPE paper series ingest complete (5 papers)

Trigger: User provided 5 papers on FFPE samples and their influence on cancer prognosis/sequencing. Ordered sequential ingest building knowledge layer by layer.

Papers ingested (5):

#	Paper	Core contribution	Analyte
1	Greytak et al. (2015)	Scope: Systematic review — FFPE affects all analytes (DNA, RNA, protein). Concordance 59–99% (genotype), 53–98% (CNA), 33% DEG overlap (RNA). Problem is platform-, gene-, and diagnosis-specific.	All
2	Do & Dobrovic (2015)	DNA mechanism: Cytosine deamination (C→U→T) = 60–80% of FFPE artifacts. UDG removes uracils. Molecular tagging (Safe-SeqS, duplex seq) distinguishes strand-specific artifacts. Template quantity is the critical pre-analytic variable.	DNA
3	Zhu et al. (2019)	Protein counterpoint: FFPE proteomes are highly concordant with FF. FFPE superior to FF for biomarker discovery — formalin cross-linking preserves proteins. Temporal stability 1–15 years. Hierarchy: DNA > RNA >> protein.	Protein
4	Steiert et al. (2023)	Modern synthesis: Complete FFPE-DNA damage spectrum (C→T, C→A, T→A, T→C). AAFs exceed 10%, penetrating standard 5% VAF filters. Three information-loss mechanisms. ERROR-FFPE-DNA checklist.	DNA
5	Pignatta et al. (2025)	Practical RNA-seq: Two-kit comparison. TaKaRa achieves comparable expression with 20-fold less RNA input. Pathologist-assisted microdissection workflow. Operational guidance for FFPE transcriptomics.	RNA

Wiki pages created (5): Source summaries for all 5 papers.

Wiki pages updated (2):

• variant-allele-fraction.md — Major expansion: “FFPE-Induced VAF Artifacts” section with artifact mechanism, quantitative impact table, artifact VAF ceiling, clinical false negatives, SMF floor, analyte-specific bias hierarchy, and cross-references to all 5 papers.
• APOBEC-mutagenesis.md — Added “FFPE Artifact Confound” section: chemical identity of FFPE C→T and APOBEC C→T pathways, CpG vs TpC discriminability, UDG treatment caveat.

Index updated: Sources 59→64. New concept-level synthesis integrated into VAF and APOBEC pages.

Gaps identified:

• No dedicated FFPE pre-analytical variables concept page yet — synthesis currently distributed across VAF page and source summaries
• Missing: Capello2022 and Lismann2025 (not in raw/papers/)
• The TCA/pathology workflow (user’s domain knowledge) has been partially integrated but not formalized as a wiki concept
• Pre-analytical variable taxonomy (cold ischemia time, fixation duration, block storage, section age) not yet systematized

Key cross-paper synthesis:

1. Analyte-specific hierarchy: DNA > RNA >> protein for FFPE impact. DNA suffers artifacts; RNA is intermediate (miRNA more stable); protein is preserved.
2. Artifact VAF ceiling: FFPE artifacts can exceed 10% AAF, penetrating standard 5% bioinformatic filters. Subclonal mutations (5–25% VAF) share the same VAF range as FFPE artifacts.
3. C→T is dominant but not unique: UDG addresses deamination-derived C→T (60–80% of artifacts) but oxidation products, abasic sites, and other base substitutions remain.
4. Template quantity is the master variable: Below ~20% tumor content, amplifiable templates are too few to distinguish true low-VAF mutations from stochastic enrichment of DNA lesions.
5. The SMF measured from FFPE data reflects three confounders with identical low-VAF signatures: true subclonal mutations + CNA-induced miscalibration + FFPE artifacts.

Lint: pending

2026-07-12 — FFPE paper series ingest: Greytak 2015 (1 of 5)

Paper 1: Greytak, Engel, Bass & Moore (2015) — NCI systematic review of FFPE biospecimen accuracy. 34 articles, case-matched FFPE vs. frozen. Foundation for understanding pre-analytical confounders in cancer genomics.

Wiki pages created (1):

• greytak2015-ffpe-biospecimen-accuracy.md — Source summary. Genotype concordance 59–99% depending on platform; FFPE false-positive SNVs 1–15% FDR; CNA discordance 2–47%; diagnosis × preservation interaction (only 33% DEG overlap); key modifiable factors.

Wiki pages updated (1):

• variant-allele-fraction.md — Added “FFPE-Induced VAF Artifacts” section: C>T deamination mechanism, quantitative impact table, clinical false negatives, SMF floor (1–15% phantom subclonal mutations from fixation).

Index updated: Sources 59→60.

2026-07-12 — Q&A synthesis: CNA/SNV discarding pipeline + entity pages

Trigger: Q&A on why subclone detection is blinded in CNA regions and where SNVs get discarded.

Wiki pages updated (3):

• copy-number-alteration.md — Added §5.4 “The Pipeline: Where SNVs Get Discarded” — two-step algorithmic exclusion (clustering pre-filter + multiplicity determination drop), tool-specific behavior table (PyClone, PhyloWGS, ABSOLUTE, DeCiFer), net effect in high-aneuploidy tumors.
• cancer-cell-fraction.md — Expanded Calculation section with N_T vs. m distinction table (five-row variable glossary). N_T = regional property; m = mutation-specific property.
• subclonal-reconstruction.md — Added “SNV Discarding in CNA Regions” section with where discarding happens, DeCiFer/DCF rescue, and consequences table.

Wiki pages created (3):

• bladder-cancer.md — Entity page; APOBEC3B-signature cancer type
• cervical-cancer.md — Entity page; HPV-driven, extreme C/G mutation bias
• head-and-neck-cancer.md — Entity page; kataegis-characterized, strong APOBEC3B upregulation

Index updated: Entities 18→27. New entries added to entity list.

Lint: 3 frontmatter issues (new entities missing confidence — fixed). Missing wikilinks: 153→150.

2026-07-12 — Burns 2013 Nature Genetics ingest

Operations:

• Ingested Burns, Temiz, & Harris (2013) Nature Genetics — companion to Burns 2013 Nature. Pan-cancer evidence for APOBEC3B mutagenesis across 6 cancer types (4,800+ tumors, >1M mutations).
• Single-paper ingest; no parallel subagent dispatch needed.

Wiki pages created (1):

• burns2013-apobec3b-multiple-cancers.md — Source summary. TCGA-based analysis across 19 cancer types; 6 converge on APOBEC3B (bladder, cervical, lung adeno, lung squamous, head/neck, breast); dose-response correlations; kataegis linkage; hierarchical clustering.

Wiki pages updated (3):

• APOBEC-mutagenesis.md — Added “Pan-Cancer Evidence” section (Burns 2013 NG multi-cancer findings, convergence argument, mechanistic model linking deamination to kataegis, caveat on enzyme attribution revision by Petljak 2022). Added to sources and related.
• apobec3b.md — Added Burns 2013 NG to sources and Wiki sources section.
• mutator-phenotype.md — Updated APOBEC3B enzymatic mutators paragraph to reference both Burns 2013 papers (Nature + Nature Genetics). Fixed broken source ref (.md → .pdf).

Index updated: Sources 58→59. New entry added to Sources list.

Lint: 3 new missing wikilinks (bladder-cancer, cervical-cancer, head-and-neck-cancer — legitimate entity pages not yet created). 0 tag issues. 0 new frontmatter issues.

2026-07-11 — CNA/SNV Q&A + hitchhiking/aneuploidy-discard synthesis + 8 missing concept pages

Trigger: Two Q&A rounds: (1) CNA/SNV relationship and SNV scaling, (2) hitchhiking principle and how high aneuploidy discards SNVs for evolutionary analysis.

Wiki pages created (8):

• copy-number-alteration.md — CNA types, measurement (logR/BAF), CIN vs. catastrophic mechanisms, CNA/SNV interdependence via CCF
• aneuploidy.md — Clonal vs. subclonal aneuploidy, fitness paradox, just-right CIN
• neo-antigen.md — Tumor-specific antigens; CNA-mediated depletion, immunoediting
• immune-evasion.md — HLA LOH, neoantigen loss, elimination-equilibrium-escape
• phylogenetic-tree.md — Trunk/branch/private mutations, CCF/crossing constraints
• crossing-rule.md — Multi-region phylogenetic constraint (CCF crossing → siblings)
• mutator-phenotype.md — Loeb’s 1991 hypothesis; genetic/enzymatic/chromosomal mutators; constitutive vs. gradient-responsive; epimutator parallel
• clonal-interference.md — τ_k vs. sweep_time transition; Greaves & Maley’s dominant dynamic; branching architecture; cross-domain functor breakdowns

Wiki pages updated (1):

• chromosomal-instability.md — Immune-visibility at high CIN, CIN/ITH inverted outcome curves

Index updated: Concepts 39→47. New entries across Foundational, Genomic Alterations, Selection & Fitness, Mutational Processes, Clinical, Methods.

Key synthesis:

• Hitchhiking: neutral passengers ride driver-driven expansions to fixation when δ ≈ 1 (Bozic 2016)
• High-aneuploidy discard chain: CNA confounds VAF → m unknown → CCF underdetermined → SNVs in CNA regions discarded → molecular clock decimated, diploid bias, blind to CNA-region subclones
• CIN (process rate, ∩) and ITH (state, ∪) produce inverted outcome curves
• Mutator phenotype is constitutive in cancer — cannot “turn off” exploration on a fitness peak (cross-domain functor failure G-CC3)

Lint: 0 new issues. All 8 new pages use valid tags.

2026-07-03 — Buehler 2011 ingestion: Hierarchical Patterns and Category Theory

Operations:

• Ingested Giesa, Spivak, & Buehler (2011): “Reoccurring Patterns in Hierarchical Protein Materials and Music: The Power of Analogies” — a short (~9 page) methods paper introducing category-theoretic ontology logs (ologs) as a rigorous framework for cross-domain analogy construction and validation.
• Ran 3-reviewer panel (EIC, methodology, clonal evolution perspective). Scores: EIC 68/100, Methodology 42/100, Perspective/evolution 84/100. The paper’s single worked example (silk ↔ music isomorphism) is trivial; the category theory is ornamental; no empirical validation or predictions. But the FUNCTORIAL COMMUTATIVITY CONDITION — an analogy is valid iff the functor preserves compositional structure — is exactly the formal discipline the wiki’s cross-domain synthesis was missing.
• Wiki pages created (1):
  • buehler2011-reoccurring-patterns.md — Source summary (evidence level VII, confidence: medium).
• Wiki pages updated (2):
  • compression-progress-evolution.md — Added “Category-Theoretic Validation” section: functor G: ClonalCat → CompressionCat, commutativity conditions identifying where the analogy holds (driver mutations under strong selection) and breaks (passengers, drift, genome as self-referential compressor/endofunctor). Buehler added as 5th source.
  • dual-regime-evolution.md — Added “Category-Theoretic Analysis of the Dual-Regime Coupling” section: DarwCat (genetic, partial order) and NonDarwCat (epigenetic, groupoid), failure of a strict functor between them, coupling mechanisms as coherence conditions, the mapping as a profunctor rather than a functor. Buehler added as 5th source.
• Index updated: sources 34→35.
• Lint: pending.

Key decisions:

• Buehler is a METHOD paper, not a CONCEPT paper. Its primary value is providing the formal toolkit for validating the cross-domain analogies already built into the wiki (compression-progress-evolution, dual-regime-evolution). We created a source summary and updated existing concept pages with category-theoretic sections, but did NOT create a standalone concept page for hierarchical ologs. The olog framework is a tool for knowledge representation, not a concept in clonal evolution.
• The formal discipline is the contribution. The functorial commutativity condition forces us to specify exactly where each cross-domain analogy holds and where it breaks. This elevates the wiki’s synthesis from heuristic mappings to mathematically precise knowledge claims with explicit validity boundaries.
• The genome as endofunctor. The category-theoretic analysis revealed a fundamental limitation of the compression-evolution analogy: the genome IS both compressor and compressed data (self-referential), requiring an endofunctor — a richer categorical structure than Schmidhuber’s agent-environment separation. This is not a refutation of the analogy but a precise specification of where it needs extension.
• The profunctor insight. The dual-regime coupling (genetic ↔ epigenetic) is not a functor but a profunctor — a relation mapping mutations to distributions of possible epigenetic outcomes conditioned on genetic context. This formalizes why the coupling mechanisms (IDH → hypermethylation, MGMT → mutation rate) are necessary: they are the coherence conditions that make the profunctor well-behaved in specific contexts.
• The methodology reviewer’s 42/100 is accurate for the paper qua paper — it’s a thin proof-of-concept with ornamental math. But the perspective reviewer’s 84/100 reflects what the paper PROVIDES for our purpose: a formal criterion for valid analogies that transforms the wiki’s cross-domain work from metaphor to mathematics.

2026-07-03 — Gabora 2017 ingestion: Honing Theory

Operations:

• Ingested Gabora (2017): “Honing theory: A complex systems framework for creativity” — a 48-page synthetic theoretical paper proposing creativity as entropy-reducing restructuring of worldviews through non-Darwinian “communal exchange.” The paper explicitly rejects Darwinian (BVSR) theories of creativity.
• Ran full 5-reviewer panel. Editorial Decision: Major Revision. Scores: EIC 72, Methodology 52, Domain 55/100, Perspective/evolution 85, DA 3 CRITICAL. All reviewers flagged the anti-Darwinian straw man, the circular psychological entropy construct, and the metaphorical (not demonstrated) SOC claim. But HT provides the conceptual vocabulary the wiki was missing: the non-Darwinian evolutionary regime.
• Wiki pages created (2):
  • gabora2017-honing-theory.md — Source summary (evidence level VII, confidence: medium). Includes full reviewer panel synthesis.
  • dual-regime-evolution.md — New concept page: cancer evolution as a hybrid of Darwinian (genetic) and non-Darwinian (epigenetic) evolutionary regimes, coupled through feedback loops. Psychological entropy → genomic entropy mapping; DDR as entropy detection; SOC in cancer genomes; dual-regime therapeutic implications.
• Wiki pages updated (1):
  • compression-progress-evolution.md — Cross-referenced to dual-regime evolution and Gabora.
• Index updated: sources 33→34, concepts 32→33.
• Lint: pending.

Key decisions:

• The Gabora-Schmidhuber synthesis is the session’s major intellectual product. Schmidhuber provides the formal computational why (compression progress drives creativity); Gabora provides the biological how (SOC, neural synchrony, context-driven restructuring restructure the worldview). In cancer: genetic level = Schmidhuberian (Darwinian compression), epigenetic level = Gaborian (non-Darwinian restructuring). The two frameworks are complementary, not contradictory, and neither paper cites the other.
• The dual-regime model is the wiki’s synthesis, not Gabora’s claim. She developed HT for cultural evolution, not cancer. The mapping onto the genetic/epigenetic distinction — which is compelling and internally coherent — is our cross-domain contribution. Confidence: medium.
• Did NOT create separate pages for SOC-in-cancer or entropy-sensing-restructuring. The dual-regime concept page covers both adequately. These may be split into dedicated pages if additional sources are ingested.
• HT’s anti-Darwinian framing, while rhetorically overwrought and attacking a straw man, contains a genuine insight that applies to cancer: the presence or absence of a self-assembly code determines which evolutionary regime operates. Cancer has both. This is a cleaner demonstration of Gabora’s principle than anything in cultural evolution.

2026-07-03 — Schmidhuber 2009 ingestion: Compression Progress

Operations:

• Ingested Schmidhuber (2009) preprint: “Driven by Compression Progress” — a theoretical manifesto formalizing curiosity as reward for compression progress (first derivative of compressibility). The paper itself never mentions biology, but the compression-evolution isomorphism is deep and precise.
• Ran full 5-reviewer academic-paper-reviewer panel. Editorial Decision: Major Revision. Scores: EIC 55, Methodology 45, Domain 28/55, Perspective/evolution 85, DA 4 CRITICAL. The paper overreaches on art/consciousness claims and is isolated from the cognitive science literature, but the core formal framework (curiosity = d(compressibility)/dt) is genuinely valuable.
• Wiki pages created (2):
  • compression-progress-evolution.md — New concept page: the full compression-evolution isomorphism. Natural selection as compression of environmental regularities into genomes, fitness as compression quality, cancer as decompression, clonal sweeps as compression breakthroughs, adaptive therapy as curiosity-driven exploration, punctuated equilibria as compression plateaus punctuated by breakthroughs.
  • schmidhuber2009-compression-progress.md — Source summary (evidence level VII, confidence: medium). Includes the full reviewer panel synthesis.
• Wiki pages updated (3):
  • clonal-evolution.md — Cross-reference to compression-progress-evolution in the Non-Malignant Somatic Evolution section. Schmidhuber added as 7th source.
  • clonal-sweep.md — New “Sweeps as Compression Breakthroughs” section: sweeps as discrete learning events for the tumor, compression gradients as determinants of sweep likelihood, therapy as forced decompression.
  • punctuated-evolution.md — Added compression breakthrough framing: punctuations as compression breakthroughs, stasis as compression plateaus.
• Index updated: sources 32→33, concepts 31→32.
• Lint: pending.

Key decisions:

• Confidence: medium for both the source summary and the concept page. The compression-evolution isomorphism is conceptually powerful and internally coherent, but it is a wiki synthesis, not Schmidhuber’s own claim, and has not been empirically tested.
• The source summary is an evidence-level-VII theoretical piece. Its primary value is the conceptual framework it provides, not empirical findings. The review panel was unusually consistent in identifying the same strengths (formal tractability, beauty/interestingness distinction, rejection of naive Shannon surprise) and weaknesses (unfalsifiability, literature isolation, consciousness overreach).
• The concept page deliberately frames the evolution connection as the wiki’s synthesis. Schmidhuber never made this connection himself. The page carries a confidence note explaining this.
• Did NOT create entity pages for Schmidhuber or CXCR4 at this time. The compression framework is captured in the concept page; Schmidhuber is referenced via the source summary.

Research query addressed: User asked for a full review of Schmidhuber 2009 with emphasis on what insights it offers for clonal evolution. The perspective reviewer (evolutionary biology lens) scored the connection at 85/100 and provided the detailed mapping that became the concept page.

2026-07-03 — McDermott 2015 ingestion: Chromothriptic cure of WHIM syndrome

Operations:

• Ingested McDermott, Gao & Murphy (2015) Addendum in Rare Diseases: “Chromothriptic cure of WHIM syndrome: Implications for bone marrow transplantation” — a single case study (evidence level VI) of spontaneous somatic reversion via chromothripsis, where deletion of the mutant CXCR4 allele in a single HSC cured WHIM syndrome through conditioning-free hematopoietic repopulation.
• Ran full 5-reviewer academic-paper-reviewer panel on the addendum (EIC, methodology, domain, perspective, devil’s advocate). Editorial decision: Major Revision. Key findings: the case is biologically remarkable but the conditioning-free engraftment claim overreaches the mouse data (irradiated recipients), the 164-gene deletion confounds attribution to CXCR4 alone, and the revertant mosaicism literature is not engaged.
• Wiki pages updated (5):
  • chromothripsis.md — Added “Beneficial Chromothripsis” subsection: chromothripsis as value-neutral mechanism, WHIM-09 as definitive example of therapeutic chromothripsis outside cancer.
  • hopeful-monster.md — Added “Hopeful Monsters That Benefit the Organism” section: extends concept beyond cancer, establishes value-neutrality of somatic evolution.
  • mcgranahan2017-clonal-heterogeneity-tumor-evolution.md — Added sixth key finding: chromothripsis can occasionally have a positive impact on patient outcome (McDermott 2015 WHIM case), originally present in the PDF but omitted from the initial source summary.
  • clonal-evolution.md — Added “Non-Malignant Somatic Evolution” section: WHIM-09 as definitive demonstration that Darwinian somatic evolution operates identically in non-malignant tissues, with estimated s ≈ 0.01–0.1.
  • punctuated-evolution.md — Added “Non-Malignant Punctuated Evolution” section: WHIM-09 as punctuated evolutionary dynamics in a non-malignant HSC.
• Source summary written: mcdermott2015-chromothriptic-cure-whim.md
• Index updated: sources count 31 → 32.
• Lint: 0 new issues (163 pre-existing missing wikilinks unchanged).

Key decisions:

• Ingested the Rare Diseases Addendum, not the primary Cell paper (Cell 160, 686-699), which is not currently in the corpus. The addendum serves as an accessible summary but the Cell paper should be obtained for complete methodological detail.
• Confidence: medium for the source summary (single case study with 164-gene confound and conditioning mismatch).
• Tagged as low-evidence (case study), genomic-alterations (chromothripsis mechanism), clinical (therapeutic implications).
• Did NOT create a new “somatic rescue” concept page — the material is adequately covered within chromothripsis and hopeful-monster pages. A dedicated page may be warranted if additional revertant mosaicism papers are ingested.

Research query addressed: User asked about the McGranahan2017 claim that chromothripsis cured a WHIM patient. Synthesized from wiki-first analysis across 3 parallel subagents. Findings: the case IS clonal evolution in its purest form — mutation, selection, clonal sweep — just in a non-malignant direction. Somatic evolution is value-neutral.

2026-06-16 — Vault initialization

• Initialized wiki/index.md and wiki/log.md
• Seeded with 7 papers in raw/papers/

2026-06-16 — Source-summary ingestion

• Ingested 7 source-summaries in parallel: Nowell (1976), Greaves & Maley (2012), Nik-Zainal et al. (2012), McGranahan & Swanton (2017), Turajlic et al. (2019), Gerstung et al. (2020), Tarabichi et al. (2021)
• Spot-checked McGranahan & Swanton (2017), Nowell (1976), and Greaves & Maley (2012) summaries against PDFs — all accurate

2026-06-16 — Concept vocabulary extraction

• Extracted 47 concepts from 7 source-summaries
• Organized into 7 thematic clusters: Foundational, Selection & Fitness, Genomic Alterations, Mutational Processes, Subclonal Inference Methods, Clinical Implications, Entities

2026-06-16 — Concept pages (batch 1): Selection and Evolution

• Created: clonal-evolution, clonal-expansion, clonal-sweep, driver-mutation, passenger-mutation
• 5 pages, all anchored to 3-4 sources each, confidence: high

2026-06-16 — Concept pages (batch 2): Heterogeneity and Progression

• Created: intratumor-heterogeneity, subclonal-architecture, branching-evolution, punctuated-evolution, neutral-evolution
• 5 pages, all anchored to 3-4 sources each, confidence: high

2026-06-16 — Concept pages (batch 3+): Methods, Genomics, and Clinical

• Created: chromosomal-instability, genetic-instability, whole-genome-duplication, mutational-signature, kataegis, subclonal-reconstruction, variant-allele-fraction, cancer-cell-fraction, therapy-resistance, metastasis
• Created entity pages: TRACERx, PCAWG
• 12 pages total; all multi-sourced except kataegis (discovery paper, single source)

2026-06-16 — Index, lint, and finalization

• Updated wiki/index.md with all 31 pages
• Ran L1 lint: 87 unique wikilinks, 31 resolved, 56 planned-but-uncreated (47 concepts/entities, 5 alias mismatches, 4 entity pages)
• 2 single-source high-confidence flags (kataegis.md, PCAWG.md) — acceptable exceptions
• 0 orphan pages, 0 frontmatter failures
• Lint report saved to outputs/lint-2026-06-16.md

Wiki stats

• Total pages: 31 (7 source-summaries, 20 concepts, 2 entities, index, log)
• Total commits: 8
• Papers covered: Nowell 1976 through Tarabichi 2021 (45-year span)

2026-06-17 — Source-summary ingestion batch 2: Tumor growth kinetics

• User queried “what is Gompertzian growth?” → identified gap: no source in vault covered tumor growth kinetics
• User added 6 new papers to raw/papers/ (7 files, 1 duplicate: Castorina 2009 PMC vs. published version)
• Extracted text from all PDFs via pdftotext; 5/6 extracted cleanly; sottoriva2015 had syntax warnings but yielded 1939 readable lines
• Ingested 6 source-summaries: Castorina et al. (2009), Traina et al. (2010), Sarapata (2013), Sottoriva et al. (2015), Graham & Sottoriva (2017), Hassan & Al-Saedi (2024)
• Subagents hit permission wall on file writes; wrote all 6 source-summaries directly

2026-06-17 — Review gate (batch 2)

• Spot-checked Castorina et al. (2009), Sottoriva et al. (2015), and Graham & Sottoriva (2017) against extracted texts
• All three passed: bibliographic details, core claims, and key values verified

2026-06-17 — Concept synthesis: Gompertzian growth

• Created: gompertzian-growth — anchored to 5 sources, confidence: high
  • Cross-referenced with: clonal-expansion, neutral-evolution, cancer-cell-fraction, subclonal-architecture, driver-mutation
  • Covers: equation, key properties, fit to empirical data, clone detectability constraint, therapeutic scheduling (Norton-Simon), open question about 1/f null model correction
• Updated: neutral-evolution — added Graham & Sottoriva (2017) 1/f test + ~30% finding; added Sottoriva et al. (2015) Big Bang model; added growth model caveat (Gompertzian correction to null distribution)
• Updated: clonal-expansion — added Growth Model Constraints section explaining how Gompertzian kinetics affect clone detectability

2026-06-17 — Index and log

• Updated wiki/index.md: sources 7→13, new Growth Kinetics category with gompertzian-growth
• New concept: gompertzian-growth (1)
• Updated concepts: neutral-evolution, clonal-expansion (2)

2026-06-20 — Source-summary ingestion: PCAWG Consortium (2020)

• User added PCAWG Consortium flagship paper to raw/papers/PCAWG2020.pdf (23 MB, Nature 2020)
• Extracted full text via pdftotext (5,248 lines)
• Created source-summary: pcawg2020-pan-cancer-analysis
• Review gate: spot-checked 10 key claims against PDF text — all verified (91%/4.6 drivers, 22.3% chromothripsis, 3.6% driver overlap, 4 telomere clusters, Strombolian/Plinian, BRCA1 templated insertions, MBD4 CpG mutagenesis)
• Created concept: chromothripsis (1)
• Updated concepts: driver-mutation (pan-cancer driver landscape, biallelic inactivation, non-coding drivers, rank-and-cut), passenger-mutation (rank-and-cut method), clonal-evolution (PCAWG three-precondition Darwinian framing) (3)
• Updated wiki/index.md: sources 13→14, concepts 21→22
• L1 lint: passed — 2 new planned-but-uncreated wikilinks (retrotransposition, telomere-maintenance), 0 frontmatter failures, chromothripsis has 3 sources (above single-source threshold)

2026-06-20 — Source-summary ingestion: Al Bakir et al. (2023) TRACERx metastasis

• User added TRACERx NSCLC metastasis paper to raw/papers/Bakir2023_NsccTRACERx_Lung.pdf (16 MB, Nature 2023)
• Extracted full text via pdftotext (2,974 lines)
• Created source-summary: bakir2023-tracerx-metastasis
• Review gate: spot-checked 12 key claims against PDF text — all verified (75% late, 25% early, 83% misclassification, 32% polyclonal, <20% LN gateway, dN/dS values, <8mm tumour size, 33% metastasis-unique drivers, 68.6% shared drivers, seeding CCF P=6.4×10^−5, 81.8% platinum signature, maintained drivers)
• Updated concepts: metastasis (major — timing, dissemination, LN role, selection, two-category model, platinum), clonal-sweep (metastatic timing relative to last sweep), driver-mutation (two-category metastasis model)
• Updated entity: TRACERx (421 cohort findings)
• Updated wiki/index.md: sources 14→15
• L1 lint: passed — 0 new broken wikilinks (all flagged are pre-existing planned-but-uncreated), 0 frontmatter failures

2026-06-20 — Source-summary ingestion: Bozic et al. (2010) driver-passenger model

• User added Bozic et al. (2010) to raw/papers/Bozic2010_AccumulationDriverPassenger_Mutation.pdf (PNAS 2010)
• Extracted full text via pdftotext (1,041 lines)
• Created source-summary: bozic2010-driver-passenger-model
• Review gate: spot-checked 12 claims against PDF — all verified (s=0.004±0.0004, GBM/pancreatic consistency, 34,000 driver positions, u=3.4×10^−5, τ_k formula, 8.3y/4.5y waiting times, FAP validation, stochastic variation)
• Updated concepts: driver-mutation (expanded selective advantage — branching process model, s=0.4% derivation, FAP validation, inter-driver waiting times, stochastic variation), passenger-mutation (linear accumulation model, driver-passenger formula)
• Updated wiki/index.md: sources 15→16
• L1 lint: passed — 0 broken wikilinks, 0 frontmatter failures

2026-06-20 — Source-summary ingestion: Bozic et al. (2013) combination therapy model

• User added Bozic et al. (2013) to raw/papers/Bozic2013_EvolutionaryDynamics_Cancer.pdf (eLife 2013)
• Extracted full text via pdftotext (1,434 lines)
• Cross-checked claims against existing wiki — no contradictions found; wiki was conceptually aligned but lacked quantitative framework
• Created source-summary: bozic2013-combination-therapy
• Review gate: spot-checked 10 claims against PDF — all verified
• Updated concept: therapy-resistance (major — cross-resistance framework n1/n2/n12, X ≈ M n12 μ formula, simultaneous vs sequential proof, multi-lesion burden analysis, cancer stem cell fraction, vemurafenib clinical data, fitness cost bounds)
• Updated wiki/index.md: sources 16→17
• L1 lint: passed — 0 broken wikilinks, 0 frontmatter failures

2026-06-20 — Source-summary ingestion: Bozic et al. (2016) clonal/subclonal passenger model

• User added Bozic et al. (2016) to raw/papers/Bozic2016_QuantifyingClonalSubclonalPassenger.pdf (PLOS Computational Biology 2016)
• Extracted full text via pdftotext (1,500 lines)
• Contradiction check: cross-checked against neutral-evolution and passenger-mutation — no contradictions. 1/f distribution is the δ = 0 special case; Bozic 2016 provides the generalization for δ > 0.
• Created source-summary: bozic2016-clonal-subclonal-passenger
• Review gate: spot-checked 12 claims against PDF — all verified
• Updated concepts: neutral-evolution (major — death-birth ratio δ framework, fixation probability, frequency spectrum, clonal ≠ truncal, tree shape, TCGA δ ≈ 0.997), passenger-mutation (m_s and m_c formulas, δ-dependence)
• Updated wiki/index.md: sources 17→18
• L1 lint: passed — 0 new broken wikilinks (all flagged are pre-existing planned-but-uncreated), 0 frontmatter failures

2026-06-20 — Long-form mathematical synthesis: branching process scaffolding

• Dispatched 4 parallel subagents with long-form math analysis of the Bozic-Nowak trilogy (2010, 2013, 2016)
• Created concept: branching-process-model — architectural scaffolding unifying all three papers (branching process definition, discrete vs continuous time, u-scale reconciliation table, δ evolution, stagnation formulation, compound μ, model boundaries)
• Updated concepts: driver-mutation (P(survival) ≈ 2ks ≈ 0.8%, log(k) structure), therapy-resistance (Mathematical Architecture — μ derivation, cell/lineage distinction, extinction filter, p_erad independence), neutral-evolution (1/f as joint test of neutrality + pure birth, δ calibration)
• Fixed formula error: bozic2016 source-summary CDF corrected from [1 + log(…)/u]^(-k) to [u/(u − log(…))]
• Updated wiki/index.md: concepts 22→23
• L1 lint: passed — 0 broken wikilinks in new concept page, 0 frontmatter failures

2026-06-20 — Source-summary ingestion: Dananberg et al. (2024) APOBEC review

• User added Dananberg et al. (2024) to raw/papers/Dananberg2024_ApobecMutagenesis_Cancer.pdf (Cancers 2024)
• Extracted full text via pdftotext (1,465 lines)
• Created source-summary: dananberg2024-apobec-cancer-review
• Updated concept: APOBEC-mutagenesis — added tissue specificity and timing section (normal-tissue prevalence table, early vs late APOBEC activity by tissue), in vivo evidence section (APOBEC3A as most potent carcinogenic APOBEC3, APOBEC3G bladder cancer role, APOBEC3B variable effects), expanded germline modulation (A3AB ethnic variation, rs1014971 SNP)
• Updated wiki/index.md: sources 20→21

2026-06-20 — Source-summary ingestion: Petljak et al. (2022) APOBEC3 mechanisms

• User added Petljak et al. (2022) to raw/papers/Petljak2022_ApobecMutagenesisMechanism.pdf (Nature 2022)
• Extracted full text via pdftotext (11,415 lines)
• Contradiction detected: Paper directly contradicts the existing APOBEC-mutagenesis page (APOBEC3B as primary source → APOBEC3A is main driver; APOBEC3B can restrain APOBEC3A). Page was corrected.
• Created source-summary: petljak2022-apobec3-mechanisms
• Updated concept: APOBEC-mutagenesis — corrected APOBEC3A/B roles, added causal evidence section, UNG/REV1 downstream processing, episodic mutagenesis, updated germline modulation mechanism
• Updated wiki/index.md: sources 19→20

2026-06-20 — Source-summary ingestion: Conticello (2008) AID/APOBEC family

• User added Conticello (2008) to raw/papers/Conticello2008_AidApobecFamily.pdf (Genome Biology 2008)
• Extracted full text via pdftotext (1,387 lines)
• Created source-summary: conticello2008-aid-apobec-family
• Created long-planned concept: APOBEC-mutagenesis — resolves 10+ pre-existing wikilinks across mutational-signature, kataegis, clonal-evolution, passenger-mutation, intratumor-heterogeneity, and source-summaries
• Review gate: 9 claims spot-checked against PDF — all verified
• Updated wiki/index.md: sources 18→19, concepts 23→24

2026-06-23 — Intermediate clones + punctuated equilibrium synthesis

• Research query: punctuated equilibrium vs. gradualism, with focus on intermediate clones and clonal ≠ truncal distinction
• Created concept: intermediate-clones — synthesizing Bozic et al. (2016) δ framework, Turajlic et al. (2019) punctuated equilibrium, and Tarabichi et al. (2021) detection limits
• Updated: punctuated-evolution — added “The Illusion of Punctured Gradualism” section and revision history
• Updated: passenger-mutation — linked clonal ≠ truncal to intermediate clone bottlenecks
• Updated: neutral-evolution — linked δ-driven fixation to intermediate clone extinction
• Updated: wiki/index.md — concepts 24→25
• L2 audit: 3 pages sampled, 14 claims verified, 0 fabricated, 0 drift, 3 pre-existing missing wikilinks flagged
• Audit output: outputs/audit-2026-06-23.md

2026-06-26 — Imprinting literature ingestion

• Ingested 6 sources in parallel on genomic imprinting and kinship theory:
  • Paper: Haig (2004) — Genomic imprinting and kinship. Annual Review of Genetics.
  • Paper: Falls et al. (1999) — Genomic imprinting: implications for human disease. The American Journal of Pathology.
  • Paper: Monk et al. (2019) — Genomic imprinting disorders. Nature Reviews Genetics.
  • Paper: Patten et al. (2014) — The evolution of genomic imprinting. Heredity.
  • Paper: Wilkins & Haig (2003) — What good is genomic imprinting? Nature Reviews Genetics.
  • Article: Extended Brain (2026) — The genetic text: how David Haig reimagines evolution as interpretation. Substack.
• Pipeline upgrade: replaced pdftotext with pymupdf4llm for PDF pre-extraction (preserves structure, reading order, italics, hyperlinks). Added rule: math-heavy papers require subagent to Read PDF directly for equations.
• Visual information policy: added Mermaid.js diagram support for concept pages (Obsidian-native); three-tier system (Mermaid > extracted figures > structured captions) documented in CLAUDE.md.
• Review gate: spot-checked Falls (1999) and Haig (2004) summaries against original PDFs — all key claims verified (IGF2 LOI 70% Wilms, three-mechanism model, weak vs strong kinship theory, four-cluster approach).
• Created concept pages: genomic-imprinting, loss-of-imprinting (both with Mermaid diagrams)
• Updated wiki/index.md — sources 21→27, concepts 25→27
• Updated wiki/index.md with new “Imprinting and Epigenetics” concept section
• Updated CLAUDE.md: pymupdf4llm extraction rule, math-heavy paper protocol, Mermaid diagram policy
• L2 audit completed: 3 pages, 18 claims, 0 fabricated, 0 drift, 1 gap (genetic-instability missing epigenetic instability), 3 missing wikilinks. Output: outputs/audit-2026-06-26.md

2026-06-26 — Bozic trilogy equation audit + Mermaid diagrams

• Verified Bozic 2010, 2013, and 2016 key equations against original PDFs — all equations in existing source summaries confirmed accurate
• Added Mermaid diagrams to 4 concept pages:
  • branching-process-model — decision tree (cell fate at division) + δ-driven phylogeny shapes (star vs linear)
  • clonal-sweep — sweep condition dynamics (τ_k > sweep time, early vs late regimes, therapy shortcut)
  • genetic-instability — four instability types (CIN, MSI, mutator, epigenetic) with optimal-level tension
  • subclonal-architecture — three evolutionary patterns (punctuated, gradual, neutral) with phylogenetic signatures
• Updated genetic-instability with epigenetic instability (MLIDs from Monk 2019) — addressing L2 audit gap
• Added revision history entries to all 4 updated concept pages

Wiki stats

• Total pages: 50 (21 source-summaries, 25 concepts, 2 entities, index, log)
• Papers covered: Nowell 1976 through Monk et al. (2019) (43-year span)
• New theme: genomic imprinting as epigenetic substrate for clonal evolution

2026-06-27 — Molecular clock concept page via synthesis_agent bridge

• Dispatched academic-research-skills:synthesis_agent in wiki-bridge mode to synthesize wiki/concepts/molecular-clock.md from 6 source summaries (Greaves & Maley 2012, Graham & Sottoriva 2017, Nik-Zainal et al. 2012, Gerstung et al. 2020, McGranahan & Swanton 2017, Tarabichi et al. 2021) and 4 concept pages
• Created bridge spec: docs/superpowers/specs/wiki-synthesis-agent-bridge.md
• Updated CLAUDE.md workflows (Ingest Phase 3, Query→Update→Audit) to reference bridge for 3+ source synthesis
• Updated wiki/index.md: concepts 27→28, new molecular-clock entry under Mutational Processes
• Cross-paper tension inventory resolved: CP-001 (rate constancy vs. spectrum shifts), CP-002 (subclonal selection prevalence)
• Resolves a long-planned wikilink target from the initial concept vocabulary (June 2026)

2026-06-27 — Bridge re-synthesis of 5 core concept pages

• Dispatched academic-research-skills:synthesis_agent in wiki-bridge mode to re-synthesize 5 high-source-count concept pages with formal cross-paper tension inventories:
  • clonal-evolution (5 sources): Added evolutionary cycle Mermaid diagram, quantitative anchors (s≈0.4%, 4.6 drivers, 22.3% chromothripsis), Evolutionary Modes section with process-to-mode Mermaid diagram, expanded Clinical Significance to 6 subsections. Tension inventory: 6 candidate pairs, 3 conditional differences resolved (Nowell vs Greaves on sweep exclusivity, Nowell vs Turajlic on selection continuity, Nowell vs PCAWG on variation sources).
  • neutral-evolution (7 sources): Added two Mermaid diagrams (neutral vs selection process flow, δ-shifted frequency spectrum across 4 δ values). Formally resolved 1/f vs δ tension as nested models (1/f is δ=0 special case of Bozic 2016 spectrum). Strengthened Big Bang empirical evidence (349 glands, 15 tumors, 6 spatial categories). Added neutral evolution–molecular clock connection. Tension inventory: 3 candidate pairs.
  • APOBEC-mutagenesis (7 sources): Added APOBEC pathway Mermaid diagram (deamination → UNG fork → SBS2/SBS13). Formalized Burns→Petljak historical consensus shift as structured tension table. Added “APOBEC as Clock-Disrupting Process” section (episodic rate violation, tissue-specific asynchrony, 3 implications for neutral inference, practical mitigation).
  • subclonal-architecture (4 sources): Added molecular clock connection (architecture as clock substrate). Crossing-rule with numeric two-region example. Formalized detection limit convergence (Tarabichi CCF floor + Turajlic doublings blind zone).
  • passenger-mutation (6 sources): Added passenger accumulation Mermaid diagram (5 temporal strata). Explicit clock mechanism section. Formal Nik-Zainal neutral-recorder vs Bozic drift-fixation tension.
• All 5 pages preserved existing content, revision histories, and frontmatter. Cross-paper tension inventories accumulated across all dispatches.
• Total bridge dispatches in this session: 6 (1 new concept + 5 re-syntheses)

2026-06-28 — LLM context-anchoring architecture: design, research, and P0 implementation

• Deep-research investigation on LLM context degradation and state externalization for scientific workflows (12 web sources across LLM context mechanics, RAG/knowledge retrieval, state externalization, Claude Code patterns, and scientific rigor)
• Wrote design spec: docs/superpowers/specs/2026-06-28-llm-context-anchoring-design.md — full architecture adapted from user’s software-engineering context-anchoring template to research knowledge management
• Created P0 skills:
  • .claude/skills/context-anchor/SKILL.md — discipline skill: checkpoint writes, recovery protocol, mandatory triggers for multi-step wiki tasks, red flags
  • .claude/skills/wiki-index-first/SKILL.md — technique skill: index-first navigation, decision flow, fallback conditions
• Created .claude/context-anchor.template.md — anchor schema template
• Created .claude/rules/README.md — rules directory placeholder for P1 path-scoped rules
• Updated CLAUDE.md: Session Start rule (ignition circuit), skill index entries, design spec reference
• Architecture draws from: HEMA dual memory (Ahn & Song 2025), PEEK context maps (Gu et al. 2026), ActiveContext reasoning anchors (Li et al. 2026), Martin Fowler’s two-layer context anchoring (Garg 2026), LLM Wiki pattern (Rezvani 2025), and Claude Code community layered memory patterns

2026-06-28 — Query → Update: kataegis mechanistic expansion

• Research Q&A on how kataegis results in hypermutation. Traced each claim to its source.
• Updated kataegis: expanded mechanistic inference with Conticello (2008) enzyme chemistry (zinc-dependent deamination, H[AV]E-x[24-36]-PCxxC catalytic core), Petljak et al. (2022) causal evidence for APOBEC3A as primary driver, explicit two-pathway model for mutational outcomes (C>T via replication across uracil, C>G via UNG excision + REV1 TLS), and molecular clock disruption implication (temporally clustered mutations violate gradual-accumulation assumption).
• Added Conticello (2008) and Petljak (2022) to kataegis sources (was 1, now 3 — single-source high confidence flag resolved).
• Added molecular-clock to kataegis related links.
• Lint: 0 new issues. Single-source high confidence flag resolved (kataegis: 1 source → 3).

2026-06-28 — Query → Update: asexual evolution, tetraploidy, and Haig’s framework

• Research Q&A on McGranahan & Swanton (2017) asexual evolution analogy + tetraploid adaptation, and connection to David Haig’s evolutionary concepts.
• Updated kataegis: added “Asexual Evolution and Linked Mutations” section. Kataegic focus as permanently linked haplotype — the linked-genome principle at micro-scale. Phylogenetic methods cannot treat kataegic mutations as independent markers. Added McGranahan & Swanton (2017) to sources.
• Updated genomic-imprinting: added two items to Relevance to Clonal Evolution — (6) asexual evolution and exposure of recessive mutations (functional haploidy → single-hit kinetics where branching-process-model assumes two-hit), (7) tetraploid-imprinting intersection (whole-genome-duplication buffers, imprinting exposes — opposing forces on the same parameter, resolved by context-dependent interpretation of the genetic text). Added McGranahan & Swanton (2017) to sources; added whole-genome-duplication and branching-process-model to related links.
• Lint: 0 new issues.

2026-06-28 — Infrastructure: Quartz static site for mobile reading

• Installed Quartz 5 in quartz/ with content symlink to wiki/.
• Configured: baseUrl bjorn99.github.io/clonal-evolution, ignorePatterns (log, private, templates, .obsidian), 44 plugins.
• Build succeeds: 61 markdown files → 189 emitted files.
• Deployment: Cloudflare Pages (free tier, supports private repos, unlimited bandwidth). GitHub Pages blocked on private repos. Setup via Cloudflare dashboard: connect repo, build command cd quartz && npm ci && npx quartz plugin install --from-config && npx quartz build, output dir quartz/public.
• Updated .gitignore and system architecture doc.

2026-07-02 — Article ingestion: “The Flaw Is Source Code”

• User provided article: Extended Brain (2026) “The Flaw Is Source Code” — conceptual synthesis of Buehler’s materials science (“the flaw is source code”) with Boden’s creativity taxonomy, proposing horizontal/vertical paths to radical creativity via productive error and phase transitions.
• Deep-dive dissection: identified structural transfer to clonal evolution — mutational signatures as crack patterns, productive error in somatic evolution, horizontal/vertical paths as clonal innovation modes, phase transition preconditions mapping to clonal sweep conditions. Critical assessment: noted individual-genius bias, fire-to-fusion overstatement, RLHF-as-creativity-killer strawman, and the absence of a criterion for distinguishing productive from unproductive failures ex ante. The transfer from creativity/philosophy to cancer biology is the author’s synthesis — confidence on the concept page set to medium.
• Created source-summary: extended-brain-2026-flaw-is-source-code — evidence level VII (opinion/expert commentary), confidence: low
• Created concept: productive-error (1) — the principle that replication errors fuel evolutionary innovation; cancer as productive error escaping organismal suppression. Mermaid diagram showing error fates. Covers horizontal/vertical paths to clonal innovation, phase transition preconditions in clonal architecture, clinical significance (therapy as phase transition trigger, signatures as crack patterns, error rate as therapeutic target, pre-existing resistance as pre-nucleated error). Confidence: medium.
• Updated concepts: mutational-signature (added “The Flaw Is Source Code” subsection — signatures as crack patterns encoding diagnostic information about mutational processes), genetic-instability (added productive-error framing — the tension between error suppression and error tolerance) (2)
• Updated wiki/index.md: sources 27→28, concepts 27→28
• L1 lint: passed — 0 frontmatter issues, 0 broken refs, 0 new missing wikilinks (170 missing are all pre-existing planned-but-uncreated from initial vocabulary extraction)
• Note: the remaining articles in raw/articles/ are queued for subsequent ingestion per user instruction (one-by-one deep-dive approach)

2026-07-02 — Article ingestion: “Life Is Interpretation”

• User provided article: Extended Brain (2026) “Life Is Interpretation” — layered argument that life is interpretation (not metaphorically but structurally), building from BMC biophysics through Pattee’s epistemic cut to the claim that interpretation is a qualitatively distinct form of matter.
• Deep-dive dissection: four-layer evidence chain (empirical BMC properties → mechanistic write-read-rewrite loop → philosophical epistemic cut → metaphysical “form of matter” claim). Critical assessment: “interpretation as a form of matter” is unfalsifiable metaphysics; “the first productive error was the first sign” is poetic but underspecified (sign and reader must co-emerge); bacterium-to-philosopher continuum stretches “interpretation” to near-breaking; Newman & Sarkar (2025) is a preprint. Transferred only what maps structurally: the write-read-rewrite loop as imprinting mechanism, the epistemic cut running through a single structure, cancer as rogue interpretation.
• Created source-summary: extended-brain-2026-life-is-interpretation — evidence level VII, confidence: low
• Updated concepts: productive-error (added “The First Sign: Productive Error as the Origin of Interpretation” section — first error as origin of interpretation, sign-reader co-emergence, somatic mutations as signs, cancer as rogue interpretation), genomic-imprinting (added item 8: write-read-rewrite loop as mechanistic substrate for imprinting maintenance and LOI failure; cancer as rogue interpretation) (2)
• Updated wiki/index.md: sources 28→29
• L1 lint: passed — 0 issues across all 8 checks
• Content hash added for new source-summary

2026-07-02 — Article ingestion: “The Monster Is Back, and It’s Hopeful”

• User provided article: Judson (2008) “The Monster Is Back, and It’s Hopeful” — NYT Opinionator blog post on Goldschmidt’s hopeful monster concept and its modern revival through developmental genetics.
• Created source-summary: judson-2008-hopeful-monster — evidence level VII, confidence: low
• Created concept: hopeful-monster (1) — resolves long-planned wikilink target from initial concept vocabulary. Covers: Goldschmidt’s original formulation (1933, 1940), Fisher’s microscope-analogy counter-argument, modern genetic evidence (Ultrabithorax, Sex combs reduced, naked-neck chickens), Turajlic et al. (2019) cancer application (CIN produces hopeful monsters routinely), relationship to punctuated evolution, clinical significance. Mermaid diagram: CIN → catastrophic events → hopeful monster → adaptive expansion vs. lethality. Confidence: high (empirical claims from Turajlic 2019, PCAWG 2020).
• Updated wiki/index.md: sources 29→30, concepts 28→29
• L1 lint: passed — 0 issues across all 8 checks. Missing wikilinks dropped 170→163 (hopeful-monster wikilink resolved)

2026-07-03 — Tooling review + article ingestion: Orr (2005)

• User requested review of markdown_hero (Python package v0.1.0). Assessment: not worth adopting. Features (strip, chunk, append, break, Word export, lint) are either covered by existing tooling (lint-wiki.sh for domain-specific validation) or have no use case in this wiki’s workflow (Word export, chunking). Adds a dependency with no benefit.
• Article ingestion: Orr (2005) “The genetic theory of adaptation: a brief history.” Nature Reviews Genetics.
• Pre-extraction with pymupdf4llm to /tmp/orr2005.md (1357 lines) — following the newly codified PDF pre-extraction rule.
• Created source-summary: orr2005-genetic-theory-of-adaptation — evidence level V (narrative review), confidence: high.
• Paper provides foundational population-genetic theory underlying cancer evolutionary models: Fisher’s geometric model → Bozic-Nowak branching process (why driver fitness effects are small and exponentially distributed), Gillespie’s SSWM → somatic evolution (mutation rates, selection coefficients, adaptive walks of 2-5 steps match ~4.6 drivers/tumor), EVT → universality of driver fitness distributions, diminishing returns → clonal sweep dynamics, parallel evolution → recurrent driver mutations, cost of complexity → cancer genome simplification, Fisher vs Goldschmidt → theoretical resolution of hopeful monster debate.
• Concept page update opportunities flagged (but deferred): driver-mutation.md (theoretical grounding in EVT + diminishing returns), clonal-evolution.md (Fisher’s geometric model + Gillespie’s mutational landscape as theoretical foundation).
• Updated wiki/index.md: sources 30→31
• L1 lint: passed — 0 issues across all 8 checks

2026-07-05 — Source-summary ingestion: Miething (2019)

• User added editorial to raw/articles/Miething2019_clonalEvolutionMyeloma.pdf
• Pre-extraction with pymupdf4llm to /tmp/miething2019.md
• Created source-summary: miething2019-clonal-evolution-myeloma — evidence level VII (editorial/commentary), confidence: medium.
• Key content: response depth determines clonal architecture at relapse (CR/vgPR → bottleneck → branching; PR → linear/stable); maintenance lenalidomide does not alter clonal patterns, implying immune-mediated mechanism; direct evidence for compression-entrenchment dynamics in clinical myeloma.
• Updated wiki/index.md: sources 35→36

2026-07-05 — Source-summary ingestion: Geng et al. (2016)

• User added research article to raw/papers/Geng2016_ImportanceofGeneticDiversity.pdf
• Pre-extraction with pymupdf4llm to /tmp/geng2016.md (1,385 lines)
• Created source-summary: geng2016-genetic-diversity-phenotypic-plasticity — evidence level IV (case-control/common-garden), confidence: medium.
• Key content: Invasive clonal plant (Alternanthera philoxeroides) with near-complete genetic uniformity in introduced Chinese populations (94% identical) achieves full bioclimatic niche occupancy through phenotypic plasticity, not genetic diversity. First ecological source in the wiki corpus. Cross-domain functorial mapping (ecology → cancer) provided in the source summary.
• Updated wiki/index.md: sources 36→37

2026-07-05 — Concept synthesis: bottleneck dynamics + surgical edits

Created (1):

• population-bottleneck — New concept page: population bottlenecks in clonal evolution. Core contributions: the bottleneck paradox (deeper response → branching relapse, Miething 2019), compression-progress resolution (forced decompression → renewed exploration), cross-domain evidence from ecological invasion (Geng et al. 2016 — plasticity compensates for low genetic diversity), category-theoretic structure (Buehler functorial mapping between domains), falsifiable predictions (bottleneck-severity threshold, plasticity as mediator, non-monotonicity of adaptive outcomes).

Updated (4):

• clonal-evolution — Added “Therapy-induced clonal bottlenecks” subsection to Clinical Significance: Miething (2019) on myeloma bottleneck dynamics, bottleneck paradox, lenalidomide dissociation of clinical benefit from clonal selection.
• clonal-sweep — Added “Therapy-induced bottlenecks as incomplete sweeps”: therapy creates forced sweeps that, unlike natural sweeps, can produce branching re-diversification at relapse.
• dual-regime-evolution — Added “Ecological Evidence” section: Geng et al. (2016) provides independent-domain evidence that phenotypic plasticity compensates for low genetic diversity — structural parallel to the dual-regime model’s claim for the epigenetic regime.
• compression-progress-evolution — Added “The bottleneck paradox” to therapeutic implications: deeper decompression → renewed exploration → branching relapse; therapeutic corollary (optimal bottleneck is partial, not maximal).

Cross-domain synthesis:

• First cross-domain concept page drawing on both a cancer source (Miething 2019) and an ecological source (Geng 2016)
• The ecology→cancer mapping is formalized via buehler2011-reoccurring-patterns functorial commutativity
• Both source summaries include cross-domain functor tables

Gaps filled:

• The wiki previously had no dedicated concept page on bottleneck dynamics despite their centrality to therapy response and relapse
• The wiki previously had no ecological sources — Geng is the first independent-domain evidence source
• The bottleneck paradox resolves an apparent tension between compression-entrenchment (monoclonal = entrenched = worse outcome) and the observation that deeper responses produce more diverse relapses

Verification:

• L1 lint: manual wikilink check — 0 new broken wikilinks; all new pages have valid frontmatter and tags. Full lint script crashed during content-hash phase (pre-existing issue, not related to new content).
• Updated wiki/index.md: sources 36→37, concepts 33→34

2026-07-05 — Source-summary ingestion: Walens, Lin, Damrauer, et al. (2020)

• Pre-extraction with pymupdf4llm to /tmp/walens2020.md (2,432 lines)
• Created source-summary: walens2020-adaptation-selection-clonal-evolution — evidence level III (controlled experimental mouse model with mechanistic follow-up), confidence: medium.
• Key content: Cellular barcoding tracks clonal dynamics during primary tumor growth, Her2/neu-induced regression, residual disease, and spontaneous recurrence in MTB;TAN mouse model. Two distinct routes to recurrence identified: (1) Met amplification-driven clonal dominance (~50%, sensitive to crizotinib), (2) IL-6-Jak/Stat3-dependent polyclonal reactivation (~50%, sensitive to Jak inhibitors). Progressive decrease in clonal diversity continues during residual disease — dormancy is not static. EMT identified as an adaptive (non-selective) response to oncogene inhibition.
• Strongest direct experimental evidence for therapy-induced bottleneck dynamics in wiki corpus. Dual-route finding maps onto dual-regime-evolution: Met amp = genetic/Darwinian route, Jak/Stat = epigenetic/transcriptional/plasticity route. Links to population-bottleneck, compression-progress-evolution, therapy-resistance.
• Updated wiki/index.md: sources 37→38

2026-07-05 — Source-summary ingestion: Mikutenaite et al. (2025)

• User added research article to raw/papers/Mikutenaite2025_ClonalEvolutionandtranscriptional.pdf
• Pre-extraction with pymupdf4llm to /tmp/mikutenaite2025.md (1,913 lines)
• Created source-summary: mikutenaite2025-clonal-evolution-transcriptional-plasticity — evidence level III (multi-region human study with snRNA-seq + WGS integration), confidence: medium.
• Key content: Integrated snRNA-seq + WGS from 43 spatially distinct samples across 5 prostate cancer patients. Both monophyletic and polyphyletic metastatic dissemination observed. Ongoing clonal evolution in primary tumor after metastatic spread. Transcriptional plasticity enables metastatic adaptation without new mutations — convergent activation of AR-independent, WNT, and JAK-STAT pathways across spatially distinct metastases.
• Strongest direct human cancer evidence for dual-regime-evolution: transcriptional (non-Darwinian) adaptation operates alongside genetic (Darwinian) evolution. Integration of snRNA-seq + WGS provides methodological template for dual-regime empirical testing.
• Updated wiki/index.md: sources 38→39

2026-07-05 — Concept synthesis: Walens 2020 + Mikutenaite 2025

Updated (2):

• population-bottleneck — Added “Direct Experimental Evidence” section: Walens et al. (2020) cellular barcoding study demonstrating dual-route recurrence (Met amp clonal dominance vs. Jak/Stat polyclonal). Strongest direct experimental evidence for bottleneck dynamics in the wiki corpus.
• dual-regime-evolution — Added “Direct Genomic Evidence” section: Mikutenaite et al. (2025) snRNA-seq + WGS study — transcriptional plasticity enables metastatic adaptation without new mutations; both regimes operate simultaneously; convergent transcriptional programs across metastases. Strongest direct human cancer evidence for dual-regime model. Methodological template for dual-regime empirical testing.

Gaps filled:

• The dual-regime model now has evidence from three independent domains: theoretical (Gabora 2017), ecological (Geng et al. 2016), and direct human cancer genomics (Mikutenaite et al. 2025)
• The bottleneck concept now has evidence from clinical observation (Miething 2019, myeloma) and direct experimental manipulation (Walens et al. 2020, cellular barcoding in mouse model)
• Mikutenaite provides the methodological template for separating genetic from transcriptional adaptation — exactly what the dual-regime empirical test design (deferred) requires

Corpus state: 39 sources, 34 concepts

2026-07-05 — Compression Progress Olog

Operations:

• Built the category-theoretic ontology log (olog) for the Schmidhuber (2009) compression progress framework, following the Buehler/Spivak (2011) hierarchical olog methodology (buehler2011-reoccurring-patterns).
• Wiki pages created (1):
  • compression-progress-olog.md — Full olog with 23 objects across 5 hierarchical levels (Data/Information, Compressor, Cognitive/Agent, Action/Exploration, Evolutionary Mapping), 16 arrows (compression, aesthetic, exploration, evolutionary), 8 commutativity conditions, Mermaid diagram, and functorial mapping notes to cancer and ecology domains.
• Key formalisms captured:
  • Level 1: DataSequence, Regularity, Noise, KolmogorovComplexity
  • Level 2: Compressor, CompressedRepresentation, CompressionQuality, CompressionProgress
  • Level 3: CuriousAgent, IntrinsicReward, SubjectiveBeauty, SubjectiveInterestingness, LearningCurve
  • Level 4: Exploration, Exploitation, Discovery, BoringPredictability, BoringRandomness
  • Level 5 (evolutionary): GenomeAsCompressor, FitnessAsCompressionQuality, MutationAsCompressorPerturbation, SelectionAsCompressionEvaluation, ClonalSweepAsDiscovery, CancerAsDecompression
• Commutativity conditions (8):
  1. CompressionProgress = d(Beauty)/dt — the core identity
  2. Interestingness Gradient — interestingness as derivative of beauty
  3. Exploration-Exploitation Boundary — determined by compression progress gradient
  4. Boredom Dichotomy — boring = fully compressed OR incompressible noise
  5. Discovery as Discontinuity — step-change in compression above threshold
  6. Evolution-Compression Isomorphism — functorial mapping with 5 subdiagrams; where it commutes (strong selection) and breaks (passengers, decompression inverse)
  7. Self-Referential Limit — genome as both compressor and data; endofunctor constraint
  8. Bottleneck as Forced Decompression — shallow vs. deep decompression determines exploration regime
• Cross-references: Links to compression-progress-evolution, schmidhuber2009-compression-progress, buehler2011-reoccurring-patterns, clonal-evolution, clonal-sweep, punctuated-evolution, population-bottleneck, dual-regime-evolution.
• Index updated: concepts 34→35.

2026-07-05 — Concept page: Intratumor Heterogeneity (major revision)

Operation:

• Fully revised intratumor-heterogeneity — the most-referenced missing page in the wiki (14+ pages link via wikilink). The existing page was a 300-word skeleton (3 sources, 4 related links) created 2026-06-16. Replaced with a ~5,500-word comprehensive concept page drawing on 8 sources across the full corpus.

Core contributions:

• Definition distinguishing ITH from inter-tumor and inter-patient heterogeneity — three nested levels of diversity
• Genetic ITH components: SNVs (VAF → CCF → SMF pipeline with formulas), CNAs (logR/BAF measurement), SVs (chromothripsis 22.3%, predominantly clonal), subclonal architecture (linear/branching/neutral/punctuated)
• Non-genetic ITH: epigenetic heterogeneity (Lamarckian inheritance via reader-writer complexes), transcriptional heterogeneity (Mikutenaite 2025 — plasticity without new mutations), phenotypic heterogeneity (Walens 2020 — dual-route recurrence), microenvironmental heterogeneity (spatial niche partitioning)
• Mermaid flowchart: sources → types → measurement → clinical significance → compression-entrenchment prediction
• Measurement pipeline: VAF → CCF formula → clonal/subclonal classification (CCF ≥ 0.85) → SMF. Detection limits: CCF_min ≈ 3/√(depth), ~7-doubling time bias, panel vs exome vs WGS. Confounders: TMB, purity, copy number, depth.
• Evolutionary mode table (4 × 3): linear (low ITH, sequential sweeps), branching (moderate ITH, clonal interference), neutral (high ITH, 1/f²), punctuated (low subclonal ITH despite high aneuploidy)
• Compression-entrenchment hypothesis: U-shaped ITH-outcome relationship with formal statistical specification (β₂ > 0 in Cox model SMF + SMF²)
• ITH in the cancer-evolution-olog: IntratumorHeterogeneity object, key arrows (diversity, hasITH, constrains, enables), cross-domain functor mappings (F-C4: low ITH ≠ low relapse risk; G: compressor diversity → ITH mapping)
• Cross-reference to contradiction registry CN-017 (pre-treatment vs. post-treatment monoclonality)
• Nine-item limitation section

Evidence base:

• Sources: 3 → 8 (Nowell 1976, Greaves & Maley 2012, McGranahan & Swanton 2017, Turajlic et al. 2019, PCAWG 2020, Tarabichi et al. 2021, Walens et al. 2020, Miething 2019)
• Related pages: 4 → 17
• Tags: foundations → foundations + selection + methods + clinical
• Confidence: high (unchanged — now supported by 8 independent sources)

Wiki stats: 39 sources, 38 concepts

2026-07-05 — Source-summary ingestion: Stejskal et al. (2023) ctDNA biology review

Operations:

• Pre-extraction with pymupdf4llm to /tmp/stejskal2023.md (2,041 lines)
• Created source-summary: stejskal2023-ctdna-biology-review — evidence level V (narrative review), confidence: medium, verification_status: unverified.
• Key content: Comprehensive review of ctDNA and ctRNA biology — passive release mechanisms (apoptosis → 167 bp ladder pattern; necrosis → large kbp fragments; CIN → micronuclei/double minutes; CTC breakage), active release (exosomes, microvesicles, AGO2/HDL protein complexes), degradation/clearance (minutes to 1-2 hour half-life), modulating factors (radiation, senescence, hypoxia, paracrine signaling), molecular determinants of shedding, and clinical applications (MRD detection, tumor characterization, treatment monitoring, resistance detection).
• Connection to wiki: Provides biological foundation for why ctDNA works as an ITH measurement modality. ctDNA captures multi-clonal information from subclones missed by tissue biopsy; the short half-life enables real-time evolutionary snapshots. Sample type considerations (FFPE vs. fresh vs. cfDNA) for the compression-entrenchment ITH-outcome test design: ctDNA fragmentomics (shorter <145 bp), differential shedding rates by cancer type (<50% brain/renal/prostate/thyroid, >75% many others), clonal hematopoiesis as false-positive confounder requiring matched normal filtering. ctRNA provides complementary non-Darwinian (transcriptional/epigenetic) regime information relative to ctDNA’s Darwinian (genetic) regime — connecting to dual-regime-evolution.
• Updated wiki/index.md: sources 42→43

Key decisions:

• Frontmatter related links to ctdna, intratumor-heterogeneity, clonal-evolution, therapy-resistance, cell-free-dna, liquid-biopsy, minimal-residual-disease — several are planned-but-uncreated pages that will need concept pages
• Did NOT create concept pages for ctdna, liquid-biopsy, cell-free-dna, or minimal-residual-disease at this time. These are deferred for future concept synthesis once more liquid biopsy sources are in the corpus.
• The source summary’s Relevance to Clonal Evolution section connects explicitly to the compression-entrenchment ITH-outcome test design (sample type considerations, cfDNA measurement modality, clonal hematopoiesis confound, ctRNA dual-regime connection).
• The Stejskal review is positioned as the primary biological reference for why ctDNA works — before methodological references (Abbosh 2017 phylogenetic ctDNA tracking, Mouliere 2018 fragmentomics, Avanzini 2020 mathematical modeling) are ingested.

2026-07-05 — Ma, Liu, Liu, et al. (2018) ingestion: Pan-cancer pediatric genome/transcriptome

Source summary: wiki/sources/ma2018-pediatric-pancancer.md

What was ingested: First pan-cancer analysis of 1,699 pediatric leukemias and solid tumours (B-ALL, T-ALL, AML, neuroblastoma, Wilms tumour, osteosarcoma) with WGS, WES, and transcriptome data processed under a uniform analytical framework.

Key findings incorporated:

• Pediatric cancers have median mutation rates 0.17–0.79 per Mb, 10–100x lower than adult epithelial cancers
• 142 driver genes, only 45% shared with adult pan-cancer studies; CNAs/SVs constitute 62% of driver events
• 11 mutational signatures identified, including UV-light signature in 8 aneuploid B-ALLs
• Chromothripsis in 11% of samples across all histotypes
• 20% of expressed mutations exhibit allele-specific expression; truncating mutations suppressed, hotspot mutations elevated
• Novel KRAS isoforms in 70% of leukaemias

Concept pages updated: clonal-evolution, driver-mutation, mutational-signature, intratumor-heterogeneity — each received Ma2018 in their sources field and updated updated dates.

Key decisions:

• The summary emphasizes pediatric cancers as a distinct evolutionary regime — low mutation burden, developmental origins, CNA/SV-driven evolutionary steps
• The “Relevance to Clonal Evolution” section connects to punctuated-evolution (chromothripsis), subclonal architecture despite low mutation rates, and ASE as an underappreciated evolutionary mechanism
• Evidence level set to III (controlled study / large cohort) pending source verification agent dispatch
• Companion Gröbner et al. (2018) CNS pan-cancer study noted but not ingested

2026-07-05 — Source-summary ingestion: Spina, Bruscaggin, Cuccaro, et al. (2018) ctDNA in Hodgkin lymphoma

Operations:

• Pre-extraction with pymupdf4llm to /tmp/spina2018.md (1,023 lines) — following PDF pre-extraction rule.
• Created source-summary: spina2018-ctdna-hodgkin-lymphoma — evidence level III (controlled cohort study), confidence: medium, verification_status: unverified.
• Key content: ctDNA mirrors HRS cell genetics (87.5% sensitivity for biopsy-confirmed mutations) enabling noninvasive genotyping of cHL. STAT6 identified as most frequently mutated gene (~38%). Two distinct clonal evolution patterns documented: (1) under chemotherapy/brentuximab vedotin, ancestral clones persist and acquire new mutations at relapse (shallow bottleneck, branching from persistent ancestral clone); (2) under nivolumab, ancestral clones are suppressed and replaced by novel clones (deep bottleneck, rapid re-diversification). ctDNA quantification after 2 ABVD courses predicts outcome: >2-log drop associated with cure, <2-log drop with progression. ctDNA resolves inconsistent interim PET/CT results.
• Connections to existing wiki concepts: therapy-resistance (ancestral reservoir persistence through chemotherapy), population-bottleneck (two bottleneck severities — shallow under chemotherapy, deep under immunotherapy), clonal-evolution (treatment-dependent clonal trajectories), intratumor-heterogeneity (subclonal mutations detected in ctDNA but missed by microdissected biopsy).
• Updated wiki/index.md: sources 43→44
• Lint: not run (pre-existing script crash in content-hash phase).

2026-07-05 — Source-summary ingestion: Hintz et al. (2006) — Hopeful monster in flower development

Operations:

• Pre-extraction with pymupdf4llm to /tmp/hintz2006.md (1,193 lines).
• Created source-summary: hintz2006-hopeful-monster-flower — evidence level V (narrative review/synthesis paper with preliminary original data), confidence: medium, verification_status: unverified.
• Key content: Covers the Spe (“Stamenoid petals”) homeotic variety of Capsella bursa-pastoris (shepherd’s purse) as a candidate for Goldschmidt’s hopeful monster. Petals transformed into stamens by a single co-dominant mutation at one locus. Spe plants persist in stable wild populations across Europe with no detectable reproductive fitness disadvantage. Proposed molecular mechanism: ectopic expression of Class C MADS-box genes (AGAMOUS orthologue) in the second floral whorl. Bridges molecular developmental genetics (ABC model, MADS-box genes) with field ecology (population monitoring, fitness comparisons, insect visitation).
• Cross-domain connections: Spe hopeful monster structurally parallels cancer hopeful monsters (chromothripsis, WHIM-09) — both involve large-effect mutations producing radically altered but viable phenotypes, bypassing gradual intermediate states. Connected to hopeful-monster, clonal-evolution, punctuated-evolution, intermediate-clones, mcdermott2015-chromothriptic-cure-whim.
• Updated wiki/index.md: sources 45→46

2026-07-06 — Ingest Lambert & Patterson (2024) + double-categorical enrichment of olog infrastructure.

• Created source-summary: lambert2024-double-functorial-semantics — evidence level VII (ACT 2024 conference paper, peer-reviewed proceedings), confidence: high, verification_status: verified.
• Key content: Double-categorical ologs unify functional ologs (Spivak & Kent, 2012) and relational ologs (Patterson, 2017). A double olog has types, terms (arrows), predicates (proarrows), and judgments (cells) — the full logic-fibered-over-type-theory structure. Queries (select, filter, join) are internal double-categorical operations rather than external adjoint functors. Tabulators enable automatic type creation from relations. Partial maps formally characterized via R†⊗R ≤ id (Theorem A.1).
• Added §Double-Categorical Enrichment to cross-domain-functors: catalogs 12/84 arrows across 3 domain ologs that are semantically relations, notes 24 commutativity conditions exist as external prose, and describes what a double-functorial upgrade of F, G, H would entail. Tier 2 (formalize) and Tier 3 (rebuild) deferred.
• Updated cancer-evolution-olog: added functional-only framework limitation (7/49 arrows are relations forced as functions).
• Updated ecology-invasion-olog: added §9 Limitations (4/19 arrows are relations forced as functions; single-source; deterministic; static). Old §9 renumbered to §10.
• Updated compression-progress-olog: added functional-only framework limitation as bullet 7 (decompresses arrow is non-unique, semantic relation).
• Updated wiki/index.md: sources 49→50.
• Sources: 49→50. Concepts: 39 (unchanged — source summary only).
• Deferred: Tier 2 (formalize proarrows, express CCs as cells, sketch double-functorial F upgrade). Tier 3 (rebuild ologs as double ologs with instance data). Recorded in memory.

2026-07-06 — Tier 2 double-categorical formalization (continued).

• Added ## Double-Categorical Annotation sections to all three domain ologs, cataloguing 17 proarrows total (cancer: 8 reclassified + 4 new = 12; ecology: 4; compression: 1) using Criterion D (functional-in-principle test).
• Added ### Commutativity Conditions as Cells and ### Double-functorial upgrade of F subsections to cross-domain-functors §Double-Categorical Enrichment.
• Three cell types constructed: Cell Type 1 — implication between composed proarrows (F-C4, ecology olog). Cell Type 2 — mechanism-level coherence with constants (CC3.7, cancer olog). Cell Type 3 — documented failure of functorial preservation (G-CC6, no cell exists for passenger accumulation).
• Double functor F specified: F₀ (unchanged 12-object mapping), F₁ (unchanged 15-arrow mapping), F₁’ (4 ecology proarrows → 4 new cancer proarrows), F₂ (3 ecology cells → 3 testable cancer predictions). Naturality constraint documented.
• Four new cancer proarrows defined: enables (EpigeneticPlasticity ⇝ Relapse), predictsOutcome (IntratumorHeterogeneity ⇝ Relapse), relapsesAs (TumorCellPopulation ⇝ Relapse), appearsIn (GenomeState ⇝ TumorCellPopulation).
• Updated revision histories and frontmatter dates on all four olog pages.
• Spec: docs/superpowers/specs/2026-07-06-double-categorical-formalization-design.md. Plan: docs/superpowers/plans/2026-07-06-double-categorical-formalization.md.
• Sources: 50 (unchanged). Concepts: 39 (unchanged — annotations are additive, no new pages).

2026-07-11 — Ingest Burns et al. (2015) + update APOBEC-mutagenesis concept page.

• Created source-summary: burns2014-apobec3b-pathological-consequences — evidence level V (narrative review), confidence: medium, verification_status: unverified.
• Key content: Canonical review from the Harris lab synthesizing the evidence establishing APOBEC3B as the primary mutagenic APOBEC family member in cancer (2013–2015). Covers APOBEC family overview, technical challenges, APOBEC3B nuclear localization and 5’TCA/5’TCG substrate preference, global cancer type analyses (bladder, cervix, lung, head/neck, breast), uracil processing mutagenic outcomes (transitions, transversions, DSBs), APOBEC3B deletion polymorphism, HPV-mediated upregulation, and hypomutator/hypermutator therapeutic strategies.
• Updated APOBEC-mutagenesis: added Burns 2015 to sources and related frontmatter. Updated the “Burns et al. papers not in corpus” note to reflect that Burns 2015 is now in the corpus (Burns 2013 Nature paper remains a gap). Added revision history entry.
• Fills a known corpus gap — the APOBEC concept page explicitly noted that direct source-summaries for Burns et al. would strengthen the evidence trail for the APOBEC3B-centric to APOBEC3A-centric consensus shift documented by Petljak et al. (2022).
• Sources: 50→51. Concepts: 39 (unchanged).

2026-07-11 — Ingest remaining 3 raw/articles/ items via parallel subagents.

• Created source-summary: asco-gu-2025-ctdna-next-generation — evidence level VII (conference report), confidence: low, verification_status: unverified. ASCO GU 2025 presentation on epigenomic ctDNA detection technologies; flags 5 missing concept pages (ctdna, clonal-hematopoiesis, cell-free-dna, minimal-residual-disease, liquid-biopsy).
• Created source-summary: cfdna-vs-ctdna-explained — evidence level VII (educational blog post), confidence: low, verification_status: unverified. Thermo Fisher explainer distinguishing cfDNA from ctDNA; 18 wikilinks all resolve.
• Created source-summary: okf-local-knowledge-base — evidence level VII (Substack article), confidence: low, verification_status: unverified. Extended Brain (2026) on building local knowledge bases via Karpathy’s LLM Wiki pattern and Google’s Open Knowledge Format. Structurally mirrors this wiki’s own architecture; identifies OKF as potential file-format layer for double ologs with arbitrary frontmatter keys for categorical metadata. Tag management critique directly relevant to wiki’s tag taxonomy governance.
• Sources: 51→53. Concepts: 39 (unchanged — article ingest only, no new concept pages).

2026-07-11 — Ingest Bian (2025) + KG construction stack design note.

• Created source-summary: bian2025-llm-kg-construction — evidence level V (narrative survey), confidence: medium, verification_status: unverified. ICAIS 2025 survey of LLM-driven knowledge graph construction across ontology engineering, knowledge extraction, and knowledge fusion. Zero biological content — earns its place as architectural self-awareness: the wiki itself is a KG, and Bian’s taxonomy provides a diagnostic framework.
• Created design note: docs/superpowers/specs/2026-07-11-kg-construction-stack.md. Synthesizes OKF (Layer 1: file format), Bian (Layer 2: algorithmic population), and Lambert & Patterson (Layer 3: formal semantics) into a three-layer knowledge infrastructure stack. Classifies the wiki’s architecture in Bian’s taxonomy (top-down, schema-based, human-AI collaborative) with strength/weakness analysis.
• Updated okf-local-knowledge-base Relevance section with three-layer stack reference and links to Bian and the design note.
• Scope discipline: No concept page created for KG construction — the concept has no biological instantiation and would sit orphaned. Architectural insight confined to the design note where infrastructure analysis belongs.
• Sources: 53→54. Concepts: 39 (unchanged).

2026-07-11 — Ingest Yuan et al. (2026) K2V + update design note to 4-layer stack.

• Created source-summary: yuan2026-k2v-knowledge-verification — evidence level VI (primary research + experiments), confidence: medium, verification_status: unverified. K2V extends RLVR to knowledge-intensive domains through fill-blank KG-based QA synthesis, checklist-style reasoning verification, and answer-gated reward mechanism. Zero biological content — earns its place as the automated verification layer in the wiki’s infrastructure stack.
• Updated design note: docs/superpowers/specs/2026-07-11-kg-construction-stack.md — expanded from 3-layer to 4-layer architecture. K2V inserted as Layer 3 (Automated Verification) between Bian/Layer 2 (Algorithmic Population) and Lambert/Layer 4 (Formal Semantics). Maturity table, capability comparison, and implications sections updated.
• Sources: 54→55. Concepts: 39 (unchanged).

2026-07-11 — Ingest Gerlinger 2012 + full pipeline integration test.

• Ingested gerlinger2012-ith-branched-evolution — the landmark NEJM paper establishing branched evolution and ITH through multi-region sequencing. Evidence level VI (n=4 case series, corrected from initial III). Verified with caveats (5/5 claims verified; Merlo citation year noted).
• Full upgraded pipeline executed successfully: Phase 0 pre-extraction → Phase 1 source summary → Phase 2 verification → Phase 2.5 KG check → Phase 3 concept synthesis. KG now at 118 triples (24 new from Gerlinger). Zero contradictions — Gerlinger introduces new entities (VHL, SETD2, KDM5C, MTOR) not overlapping with existing KG.
• Bug found and fixed: kg-check.sh argument parsing ($1 was --new-triples, not the file path). Fixed to handle --new-triples flag. Also added Format: line skip for agent-produced triple files.
• Updated branching-evolution: added Gerlinger 2012 as primary source (the paper that introduced the concept). Updated intratumor-heterogeneity: added Gerlinger to sources and related links.
• Architecture review: pipeline is working. Pre-extraction → verification → KG check → concept synthesis all operational end-to-end. Kg-check.sh bug was the right kind of bug — caught during integration testing, not in production.
• Sources: 55→56. Concepts: 39 (unchanged). KG: 95→118 triples.

2026-07-11 — Domain audit checklist + verification backfill.

• Created domain audit criteria spec: docs/superpowers/specs/2026-07-11-domain-audit-checklist.md. Eight general criteria (C1–C8) adapted from K2V’s checklist-style verification for clonal evolution domain claims: evidence hierarchy awareness, correlation vs. causation, detection limits, primary vs. review source, population/context specificity, uncertainty quantification, contradiction awareness, mathematical model assumptions. Each criterion includes binary-verifiable instantiation questions and red flags. Designed for use in Phase 2 source verification, L2 audits, and L3 audits.
• Pre-extracted 15 PDFs with pymupdf4llm for verification backfill. Two verification batches (9+9 agents) dispatched covering all 18 sources with verification_status: unverified. Agents apply domain criteria C1, C4, C5 and update frontmatter directly.
• Verification backfill complete. 18 sources verified (9 verified + 9 verified_with_caveats). Issues found and fixed: 2 tag corrections (hintz2006, miething2019: primary-research → correct type), 1 title year correction (satas2022: 2022→2021), 1 evidence level downgrade (mikutenaite2025: III→IV, n=5 study), 3 extraction retries (mikutenaite, walens, weng2026). Domain criteria C1, C4, C5 caught real issues structural lint would miss.
• Final state: 29 verified + 14 verified_with_caveats = 43/56 sources verified. 0 unverified, 0 flagged, 0 rejected.

2026-07-11 — Update system architecture document to reflect all session upgrades.

• Updated docs/superpowers/specs/2026-06-28-llm-wiki-system-architecture.md: added Layer 6 (Knowledge Infrastructure) to the runtime layer model; added new §8 Knowledge Infrastructure Stack with four-layer architecture diagram, rationale for each layer choice, architecture decisions record, and limitations; updated §7.3 QA Pipeline with domain criteria C1-C8 integration; updated §7.4 Subagent Dispatch with verification backfill process; added §7.8 Verification Status Tracking; added 3 new design principles (9.10, 9.11, 9.12); updated component map, file inventory, and references; renumbered sections 8→9, 9→10, 10→11, 11→12.

2026-07-11 — Tag co-occurrence graph + dynamic schema adaptation.

• Created scripts/tag-graph.sh: extracts tags from all 96 wiki pages, builds co-occurrence matrix (80 pairs), identifies clusters via greedy thresholding, generates Mermaid graph with color-coded categories. Two dominant clusters: theoretical/methods (7 tags) and empirical/clinical (6 tags). Six isolated tags. Zero dead tags, zero orphan tags — taxonomy is healthy.
• Created scripts/propose-tags.sh: implements AdaKGC-inspired dynamic schema adaptation. Detects orphan tags, classifies by scope, generates structured proposals with signal-strength recommendations (≥3 strong, 2 moderate, 1 weak). Human-in-the-loop by design. Also detects dead tags for pruning. Current state: zero orphan tags — taxonomy perfectly synced with usage.
• Updated architecture doc: added §7.9 Tag Co-occurrence Graph and §7.10 Dynamic Schema Adaptation. Added both scripts to component map and file inventory. These operationalize Layer 2 (algorithmic population — dynamic schema adaptation) and Layer 1 (file format — OKF-inspired tag visualization gap).
• Remaining deferred improvements: KG-based wiki self-testing (#5, Phase 0 entity pages complete), Tier 3 double olog rebuild (#6, deferred indefinitely).

2026-07-11 — Entity page seeding (Phase 0 of KG-based wiki self-testing).

• Created 16 new entity pages: APOBEC3B, APOBEC3A, TP53, EGFR, KRAS, CDK4, IGF2, RB1, CXCR4, ESR1, AR, AID (genes), ctDNA, cfDNA (methods), breast cancer, prostate cancer, lung cancer, colorectal cancer (cancer types), cisplatin, imatinib, lenalidomide (drugs). Each page includes external identifiers (HGNC, NCBI Gene, UniProt, DrugBank) and wikilinks to source summaries. Total entities: 2→18.
• These serve as canonical KG node identifiers for Phase 1 (KG construction from Key Findings). Without entity pages, entity resolution across sources was impossible — the KG had no nodes to resolve to.
• Updated index.md: Entities section expanded from 2 to 18 entries.
• Updated Tier 3 memory file: deferred indefinitely (cost/benefit analysis concluded).

2026-07-11 — Phase 1: KG construction from Key Findings.

• Dispatched 3 parallel agents to extract (subject, relation, object) triples from 29 verified source summaries’ Key Findings sections. 16 of 29 sources produced biological or methodological triples; 13 produced none (pure theory/infrastructure papers or non-biological domains).
• Consolidated output: wiki/knowledge-graph/triples.md — 94 unique triples, tab-separated format with provenance and evidence level.
• Created entity page for hodgkin-lymphoma (5 triples, most frequent new cancer type).
• Unresolved entities flagged (20): NRAS, MYC, WT1, APC, SMAD4, ERBB2, MET, STAT6, FOLH1, BRCA1, bladder-cancer, cervical-cancer, head-and-neck-cancer, multiple-myeloma, WHIM-syndrome, crizotinib, nivolumab, DeCiFer, PyClone — need pages in Phase 1.5.
• Entities: 18→19. KG operational at 94 triples from 16 sources.
• Architecture: wiki/knowledge-graph/ directory created as new wiki component.

2026-07-11 — Phase 2: Contradiction detection + KG consistency check.

• Created scripts/kg-check.sh: reusable contradiction detection script. Accepts optional --new-triples argument for checking new ingestions against the existing KG. Uses simple heuristics for incompatible relations (upregulated vs downregulated, mutated vs wildtype, confers_resistance vs confers_sensitivity). Synonym detection for same-meaning different-verb pairs.
• Ran initial check on 95 triples: 0 contradictions, 6 synonym flags. The wiki’s verified sources are internally consistent. Synonym flags (ctdna “captures” vs “assesses” ITH; rb1/esr1 “enriched_in” vs “under_selection_in” breast-cancer) suggest future relation vocabulary standardization.
• Phase 2 deliverable: contradiction detection is operational. New ingestions can be checked against the KG by running bash scripts/kg-check.sh --new-triples /tmp/new-triples.md.
• Next: Phase 3 — integrate into ingest pipeline as optional Phase 2.5.

2026-07-11 — Phase 3: Pipeline integration.

• Updated CLAUDE.md § Ingest workflow: inserted Phase 2.5 (KG consistency check) between Phase 2 (verification) and Phase 3 (concept synthesis). Fixed numbering throughout (Phases now have sequential step numbers 1-13).
• The check is automatic but non-blocking: runs on every ingestion, produces a report, human decides on contradictions. Infrastructure/theory papers without biological triples skip this phase.
• Updated architecture doc: KG check script and KG triples file added to component map.
• #5 complete. All 4 phases delivered: entity pages (Phase 0), KG construction (Phase 1), contradiction detection (Phase 2), pipeline integration (Phase 3).

2026-07-11 — Phase 2 verification of 4 new source summaries.

• asco-gu-2025-ctdna-next-generation: verified. All 5 claims cross-referenced against original article; all verbatim matches. Evidence level VII (conference report).
• cfdna-vs-ctdna-explained: verified_with_caveats. One hallucination found and fixed — Entities Referenced section incorrectly claimed the article cited Abbosh et al. (2017). Corroborating TRACERx/PCAWG wikilinks in the Relevance section are the wiki’s own cross-references, not the article’s citations. Evidence level VII (commercial vendor blog).
• okf-local-knowledge-base: verified_with_caveats. Two minor cosmetic issues: stage-count inconsistency in Core Argument (3 vs 5) and title formatting appends citation metadata. No factual errors. Evidence level VII (Substack article).
• burns2014-apobec3b-pathological-consequences: verified_with_caveats. All 5 claims cross-referenced correctly against extracted PDF. DOI 10.4103/2319-4170.148904 resolves, PubMed PMID 25566802, S2 Corpus ID 41634764. Biomedical Journal indexed in PubMed/MEDLINE, Scopus, SCIE. Three caveats: (1) source PDF filename says “apobec2b” instead of APOBEC3B — immutable per raw/ rule, noted; (2) PDF lives in raw/articles/ rather than raw/papers/ — noted; (3) Core Argument collapses 6 cancer types to 5 organ-level entries while Key Finding 3 correctly lists all 6 — low priority. Evidence level V (narrative review).

2026-07-17 (session 2)

• Ingested Naranjo-Ortiz & Gabaldon (2019) — Fungal evolution: major ecological adaptations and evolutionary transitions (Biological Reviews). Narrative review synthesizing fungal ecological diversity through comparative genomics. 9 numbered conclusions spanning zoosporic ancestry, terrestrialization scenarios, mycoparasitism, animal/plant pathogenesis, lignin degradation, lichenization, black fungi, and the yeast lifestyle. Created source summary naranjo-ortiz2019-fungal-evolution. Updated cross-domain-functors: added potential fifth domain (fungal evolution → cancer evolution functor T) with 6-paragraph structural isomorphism analysis and mapping table. Updated ecology-invasion-olog: added fungal evolution as methodological source (minimum bar). Updated dual-regime-evolution: added fungal lifestyle switching as cross-kingdom plasticity parallel (minimum bar). Verified 2026-07-17: All 5 claims cross-referenced against extracted PDF — verified. Venue: Biological Reviews (IF ~10-12, Scopus/WoS indexed). DOI resolves. S2: 294 citations. Evidence level V (narrative review). KG check skipped: infrastructure/theory paper — fungal ecology domain, no cancer-biology triples to check against existing KG. Index updated (67 sources). L1 lint clean (0 errors).

2026-07-11 (session 2)

• Ingested Burns et al. (2013) — APOBEC3B is an enzymatic source of mutation in breast cancer (Nature). Created source summary burns2013-apobec3b-breast-cancer. Updated APOBEC-mutagenesis (corpus gap filled, Burns papers callout updated, source and related links added). Updated apobec3b entity page (added 2013 discovery paper as source). Index updated (57 sources).