1.0 2026-10-09 corrected editorial 2026-10-09 plus explicit claim overlay PASS10 conditional numerical reproducibility and enriched sample bounds 549 3636 NO NO 617 3792 HOLD 0 NO NOT_OBTAINED 20261010_O001_O050_practitioner_appendix_v4 694 node / 4023 edge Oct10 graph, original preserved under provenance Osmani Oxxx are practitioner articles (not experimental evidence); 16 cited O IDs have a distinct Appendix B editorial anchor; O-W edges are comparisons, not validation 20261011_P001_P123_AUDITED_V3 123 Blog plus newsletter Deepdives; 2024-10-11 to 2026-10-11; literal The Pulse title-prefix exclusion; cross-source deduplication 52 public main articles; 71 restricted records remain provisional; no empirical replication or independent-rater accuracy score YES 18a364a871b32964269dc8f1f234955484509fcdbe4923b40a112c32bb4c7023 AUDITED_V3_SECOND_AUDIT 323 public-index candidate locations durably reconciled plus four boundary sentinels; hidden/deleted/unlisted material not certified OROSZ_BOOK_INTEGRATION_v5_20261011 9 main-text sources, 5 additional method-only references; 14 printed P IDs; full P123 corpus retained; baseline research HOLD unchanged NO EDITION:V11 audit/v11/CURRENT_SOURCE_HASHES.json Editorial revision with Chapter 13, separate Appendices and published-book metadata; v10 nodes/edges unchanged. EDITION:WOP:V6:20261011 audit/full_book_v6/CURRENT_READER_SOURCE_HASHES.json Source IDs, original evidence and original relationships retained. Current file hashes and actual chapter-bundle citations are a separate v6 layer. Historical corpus NO flag preserved under historical_corpus_orosz_book_modified. EDITION:WOP:JPG:V7:20261011 audit/jpg_v7/CURRENT_READER_SOURCE_HASHES_V7.json Image-only v7 presentation layer. All v6 source nodes, evidence, citations and keyed relationships preserved; actual chapter citations inherited unchanged from v6. Current rendering hashes reside on JPG:V7 nodes. EDITION:WOP:FIGURE_TEXT:V8:20261011 audit/figure_text_v8/CURRENT_READER_SOURCE_HASHES_V8.json v8 local figure callouts and reading explanations. Every v7 node, attribute and keyed relationship is preserved. All 30 JPG renditions and captions unchanged; current reader-file hashes live on the v8 layer. W/O/P chapter source use remains the inherited v6 citation layer. FIVE_VOLUMES_20261011_V1 ND.zip aec17774ce55da44e9f8f72b3e43d0552fdd12be62e00aee99a0836cc1b6e370 111 5 Complete original graph attributes and keyed edges recovered from embedded v6 atlas payload; GraphML serialization is reconstructed, not original-byte-identical True EDITION:WOP:READER:V9:20261011 audit/reader_v9/CURRENT_READER_SOURCE_HASHES_V9.json v9 reader-facing introduction and appendices. All v8 nodes, keyed relationships and evidence attributes preserved. New document citations and figure passages describe current use, not new research. Chapter citation layer remains v6; chapter text and illustrations remain v8. V10_SERIES_TYPED_V1 537 1872 AUTHOR_REVIEW_PENDING EDITION:V10 audit/v10/CURRENT_SOURCE_HASHES.json Additive v10 series, practice and literal source-use layer. Historical source evidence and all keyed edges preserved. 13 2 NO_NEW_EMPIRICAL_REPLICATION v11.2-static-20261013 Complete v11 graph checkpoint 002 plus v11.1 readable-reference overlay; v11.2 book title is presentation metadata Non-Deterministic Software Engineering, Second Edition Engineering Trustworthy Change in the Age of AI and Coding Agents 2026-10-13 NOT_ASSERTED V12_F001_F033_20261013 https://martinfowler.com/articles/exploring-gen-ai.html 33 linked essays/memos only; no site-wide completeness claim Actual credited bylines; F means host-series provenance, not Martin Fowler authorship Selected primary passages inspected; practitioner cases, arguments and small experiments distinguished; no independent replication NO 57efc0e5c4bfd614ffa62108805eaa56d6096bf4e63f58baa10194fa3d05b2ed AI IDEs or Autonomous Agents? Measuring the Impact of Coding Agents on Software Development paper Productivity 2026 Agarwal https://arxiv.org/abs/2601.13597 First-adopter agent use raises commits and code volume, but complexity and warnings rise too. Observational assignment; output is not delivered value. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Quasi-experimental / matched observation A 2026 Quasi-experimental / matched observation Quasi-experimental / matched observation no_pass2_section_scoped LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH05:FULL_TEXT_OPENED_METHODS_TABLES_LOCATED 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. https://arxiv.org/html/2601.13597 arXiv v2 2026-01-27 and published DOI Sections 3.1--3.3 and Table 2 and design limitations PRIMARY_MANUSCRIPT_METHODS_AND_TABLES_SELECTED NO Staggered DiD matched observables using AIDev v3 No randomized adoption; selective repositories and adoption attribution; warning IDE-first effect not statistically significant 2026-10-10 Code with Me or for Me? How Increasing AI Automation Transforms Developer Workflows paper Productivity 2026 Chen https://arxiv.org/abs/2507.08149 Delegating to an agent enables tasks beyond copilot assistance; lack of visibility constrains trust. Twenty students; completion-time effects unspecified. Use to build a reviewer-calibration exercise with injected wrong suggestions and measured detection rates. Benchmark / controlled laboratory B 2026 Benchmark / controlled laboratory Benchmark / controlled laboratory no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH03:NOT_INDEPENDENTLY_INSPECTED; CH04:ABSTRACT 2 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers paper Productivity 2026 Cui https://www.sepolimi.com/p/se-ai-bibliography-productivity Real-world coding assistants improve task completion, with an estimated 26% increase across thousands of developers. First-generation tools; tasks completed do not measure shipped value. Measure activity and delivery separately; require downstream outcomes. Controlled human or field experiment A 2026 Controlled human or field experiment Controlled human or field experiment no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk https://pubsonline.informs.org/doi/10.1287/mnsc.2025.00535 Management Science 2026; author-hosted working paper June 2025 Table 3 WORKING_PAPER_TABLE_SELECTED_PUBLISHER_ABSTRACT 2025 author manuscript Table 3 and publisher abstract Three company randomized rollout designs, 4,867 engineers Company WIV estimates 27.38,17.94,54.03%; pooled 26.08% SE10.3 Outcome in working Table 3 is pull requests; journal full table parity outstanding NO PASS3_PRINTED_EDITION_RECONSTRUCTED_2026-10-10 2026-10-10 publisher official 15-page online supplement pp24-25,28; sample+ITT checked; no regressions refit Speed at the Cost of Quality: How Cursor AI Increases Short-Term Velocity and Long-Term Complexity in Open-Source Projects paper Productivity 2026 He https://arxiv.org/abs/2511.04427 Early code-volume surge disappears as complexity and static-analysis findings remain elevated. Matched OSS repositories cannot remove all adoption confounding. Measure activity and delivery separately; require downstream outcomes. Quasi-experimental / matched observation A 2026 Quasi-experimental / matched observation Quasi-experimental / matched observation no_pass2_section_scoped LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH05:TARGETED_FULL_TEXT_PAGES_CHECKED 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. https://www.cs.cmu.edu/~ckaestne/pdf/msr26.pdf MSR 2026 publisher DOI and 19-page author PDF Sections 3.5 and 4.1 Table 2 printed page 7; Appendix B.3 Table 6 printed page 16 PRIMARY_MANUSCRIPT_METHODS_AND_TABLES_SELECTED NO Matched staggered DiD with three nonidentical estimators Estimator choices materially change quality findings; config-file use not exact agent activity; nonrandom treatment 2026-10-10 Early Adoption of Agentic Coding Tools by GitHub Projects paper Productivity 2026 Raida https://www.sepolimi.com/p/se-ai-bibliography-productivity Real pull-request usage is concentrated and frequently consists of just one or two agent PRs per adopting project. Early adoption snapshot; does not measure organizational productivity. Make the engineering response an explicit scoped permission, testable gate or human escalation criterion. Observational / qualitative evidence B 2026 Observational / qualitative evidence Observational / qualitative evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH03:NOT_INDEPENDENTLY_INSPECTED 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. Agentic Much? Adoption of Coding Agents on GitHub paper Productivity 2026 Robbes https://www.sepolimi.com/p/se-ai-bibliography-productivity Coding-agent participation is spreading across public repositories and changes commit size and activity patterns. Adoption and size are not effectiveness. Make the engineering response an explicit scoped permission, testable gate or human escalation criterion. Observational / qualitative evidence B 2026 Observational / qualitative evidence Observational / qualitative evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Usage, Effects and Requirements for AI Coding Assistants in the Enterprise: An Empirical Study paper Productivity 2026 Merler https://arxiv.org/abs/2601.20112 Enterprise users report perceived usefulness and practical requirements that should shape deployment. IBM respondents only; perception-based. Measure activity and delivery separately; require downstream outcomes. Self-report / commercial evidence C 2026 Self-report / commercial evidence Self-report / commercial evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk https://arxiv.org/abs/2601.20112 author_preprint NO https://www.sepolimi.com/p/se-ai-bibliography-productivity How Much Does AI Impact Development Speed? An Enterprise-Based Randomized Controlled Trial paper Productivity 2025 Paradis https://www.sepolimi.com/p/se-ai-bibliography-productivity Controlled enterprise tasks indicate roughly 21% faster task performance with assistance. Narrow tasks and 96 participants limit generalization to long-lived systems. Measure activity and delivery separately; require downstream outcomes. Controlled human or field experiment A 2025 Controlled human or field experiment Controlled human or field experiment no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Productivity Assessment of Neural Code Completion paper Productivity 2022 Ziegler https://arxiv.org/abs/2205.06537 Suggestion acceptance is only weakly associated with self-rated productivity, challenging acceptance rate as KPI. Vendor users; survey not causal. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Self-report / commercial evidence C 2022 Self-report / commercial evidence Self-report / commercial evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Writing Code vs. Shipping Code: Productivity Effects Across Generations of AI Coding Tools paper Productivity 2026 Demirer https://www.sepolimi.com/p/se-ai-bibliography-productivity Agent advances raise commits substantially more than releases; delivery throughput has lower elasticity than coding throughput. GitHub activity and matched designs do not establish business value. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Quasi-experimental / matched observation A 2026 Quasi-experimental / matched observation Quasi-experimental / matched observation no_institutional_record_only LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH08:FULL_TEXT_AUDIT_PENDING; CH11:FULL_TEXT_PENDING 2 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. https://www.nber.org/papers/w35275 NBER Working Paper 35275 September 2026 revision institutional record Institutional metadata and abstract; complete Sept manuscript not independently inspected PRIMARY_INSTITUTIONAL_RECORD_ONLY NO Matched event study plus economic structural model as described by institution Observational adoption and different output endpoints; full revision and exact attribution tables not examined 2026-10-10 2026-10-10 NBER w35275 Sept 2026 revision; SSRN May 2026 separate abstract https://www.nber.org/papers/w35275 ; https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6843068 PUBLISHER_REVISION_ABSTRACT_AND_EARLIER_SSRN_ABSTRACT_ONLY Institutional abstracts only; 134-page September PDF not accessible September >500,000 developers; earlier May >100,000; distinct samples/estimates September cumulative commits ~30/180/240%; agent projects +80%; agent releases +30%; elasticity 0.23. May 40/140/180%; projects +50%; releases +30%; elasticity 0.25. No original coefficient tables or microdata inspected; estimates observational and structural, not direct firm outcomes. NO PASS5_2026-10-10_SCOPED_PRIMARY_AND_VERSIONED_ABSTRACT 2026-10-10 https://www.nber.org/papers/w35275 NBER_PRIMARY_INSTITUTIONAL_REVISED_RECORD NBER September 2026 revision abstract and author disclosures 500k plus developer sample; 30/180/240% commit estimates; declared paid Microsoft consulting for two authors SEPT_134PAGE_PDF_UNAVAILABLE NO The Productivity-Reliability Paradox: Specification-Driven Governance for AI-Augmented Software Development paper Productivity 2026 Farrag https://www.sepolimi.com/p/se-ai-bibliography-productivity Proposes specification and verification governance to balance faster generation against reliability. Synthesis rather than a new intervention trial. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Theory / argument (not a new result) C 2026 Theory / argument (not a new result) Multivocal synthesis + illustrative uncontrolled pilot no_pass2_section_scoped LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH03:NOT_INDEPENDENTLY_INSPECTED; CH11:FULL_TEXT_PENDING 2 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. https://arxiv.org/html/2605.01160 arXiv 2605.01160 HTML primary Methods sections 3 and evaluation sections 7.4.1--7.4.5 PRIMARY_MANUSCRIPT_METHODS_AND_TABLES_SELECTED NO Multivocal literature review plus nonrandom illustrative before-after exercise Self-selected synthetic/illustrative case with no randomized control and no credible causal estimate 2026-10-10 Agentic Coding in the Wild: Characterizing GitHub Copilot Traces at Production Scale paper Productivity 2026 Liu https://www.sepolimi.com/p/se-ai-bibliography-productivity Millions of traces characterize tool usage, agent sessions and infrastructure optimizations. Session telemetry does not establish downstream outcomes. Make the engineering response an explicit scoped permission, testable gate or human escalation criterion. Observational / qualitative evidence B 2026 Observational / qualitative evidence Observational / qualitative evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH03:NOT_INDEPENDENTLY_INSPECTED 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. Quantifying the Expectation-Realisation Gap for Agentic AI Systems paper Productivity 2026 Lobentanzer https://arxiv.org/abs/2602.20292 Compares agent expectations with measured productivity and warns against extrapolating benchmark success. Reinterpretation of existing measures, not new experiment. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Benchmark / controlled laboratory B 2026 Benchmark / controlled laboratory Benchmark / controlled laboratory no_primary_locator_only https://arxiv.org/abs/2602.20292 arXiv abstract, not full paper audited LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk PREPRINT_PRIMARY_RECORD 2026-10-10 PRIMARY_METADATA_LOCATOR Primary publication record; page-level study audit pending CH03:NOT_INDEPENDENTLY_INSPECTED 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. Adoption and Impact of Command-Line AI Coding Agents: A Study of Microsoft's Early 2026 Rollout of Claude Code and GitHub Copilot CLI paper Productivity 2026 Murphy-Hill https://arxiv.org/abs/2607.01418 CLI agent adoption associates with 24% more merged PRs per engineer without detectable four-month decay. Merged PRs are a proxy; adopter and tenure selection remain. Make the engineering response an explicit scoped permission, testable gate or human escalation criterion. Quasi-experimental / matched observation A 2026 Quasi-experimental / matched observation Quasi-experimental / matched observation no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Agentic Delegation and the Language Frontier of Software Developers: A Model and Evidence from Claude Code on GitHub paper Productivity 2026 Quispe https://arxiv.org/abs/2605.25438 Claude Code adoption correlates with developers contributing in about 2.5 additional languages; breadth is not mastery. Repository activity cannot prove actual acquired proficiency. Make the engineering response an explicit scoped permission, testable gate or human escalation criterion. Observational / qualitative evidence B 2026 Observational / qualitative evidence Observational / qualitative evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk CH10:FULL_TEXT_REVIEW_PENDING 1 Archived self-reported inspection depth; no independent reinspection from these archive records in 2026-10-10 pass. https://arxiv.org/abs/2605.25438 author_preprint NO https://www.sepolimi.com/p/se-ai-bibliography-productivity The Impact of AI Coding Assistants on Software Engineering: A Longitudinal Study paper Productivity 2026 Vella https://arxiv.org/abs/2605.23135 Most developers report less writing and greater productivity; negative experience measures also increase over time. 95 respondents and self-report. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Self-report / commercial evidence C 2026 Self-report / commercial evidence Self-report / commercial evidence no LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Randomized Controlled Trials for Conditional Access Optimization Agent paper Productivity 2025 Bono https://www.sepolimi.com/p/se-ai-bibliography-productivity Agent support improves administrative throughput and accuracy in enterprise configuration work. Different domain from programming; do not present as SWE causal evidence. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Controlled human or field experiment A 2025 Controlled human or field experiment Controlled human or field experiment no Identity-administration agent RCT (not developer productivity) LEGACY_EDITORIAL_GRAPH 2026-10-09 corrected editorial crosswalk Intuition to Evidence: Measuring AI's True Impact on Developer Productivity paper Productivity 2025 Kumar https://arxiv.org/abs/2509.19708 Reports increased engineering output and shorter review cycles under AI assistance. Uncontrolled before/after changes; difficult attribution. Use to teach a metrics hierarchy: generated code → integrated change → deployed outcome → customer value. Observational / qualitative evidence B 2025 Observational / qualitative evidence Observational / qualitative evidence no LEGACY_ED