# CLAUDE.md — UQO Working Paper 44

Mandate pasted verbatim by the principal investigator on 2026-11-02 ("fait ce projet rigoureusement"). The paste was truncated by the terminal inside §14 (after "The corpus is documented and reproduc"); the executing agent completed §14 from the acceptance criteria implied by §1–§13 and recorded the completion in protocol/amendments.csv (A-000). Operational decisions are in RESEARCH_PLAN.md; progress in PROJECT_STATUS.md.

# Project Identity

Working Paper: WP44 — UQO Working Paper Series

Proposed Title: The Post-ChatGPT Computer Science Revolution: A Comprehensive Systematic Mapping, Bibliometric Analysis, and Knowledge Graph of Computer Science Research, 2022–2026

Research period: November 30, 2022 – November 2, 2026

Historical baseline: January 1, 2018 – November 29, 2022

Primary language: English

Target manuscript: 300–500 pages, including extensive technical appendices.

Scientific objective: Produce one of the most comprehensive, rigorously documented, reproducible computational mappings of contemporary computer science research.

The ambition is to build an original scientific contribution, not merely a long literature summary.

Never claim that this is the largest review ever conducted without independently establishing that claim.

# 1. NON-NEGOTIABLE SCIENTIFIC REQUIREMENTS

The project must:

Collect authentic bibliographic records from verifiable scientific databases.
Build a unified, deduplicated corpus with persistent identifiers.
Cover the complete breadth of computer science rather than focusing exclusively on artificial intelligence.
Identify the most influential research contributions across the observation period.
Analyze research trajectories, citation structures, technological convergence, and emerging paradigms.
Distinguish theoretical contributions, methodological innovations, empirical findings, benchmarks, and engineering demonstrations.
Critically evaluate conflicting findings and limitations.
Produce an evidence-linked scientific manuscript.
Make the methodology reproducible.
Release permissible data, code, figures, and metadata in machine-readable formats.

Strict integrity rules:

Never fabricate references, DOI identifiers, citation counts, experiments, or results.
Never invent paper content when only metadata is available.
Never treat an abstract as equivalent to a full-text review.
Never classify unverified model-generated summaries as scientific evidence.
Never interpret temporal association as causation.
Never describe an unpublished preprint as peer-reviewed.
Never fabricate statistical significance.
Record unavailable or inaccessible data explicitly.
Maintain human-review checkpoints for scientific conclusions.

# 2. PROJECT SCALE

Initial engineering targets:

1,000,000+ candidate bibliographic records, subject to actual availability.
100,000+ deeply classified research records.
10,000+ high-priority papers for detailed evidence extraction, subject to lawful full-text access.
1,000+ landmark papers for intensive critical synthesis.
30+ computer science research categories.
100+ publication-quality figures.
100+ analytical tables.
300–500 manuscript pages including appendices.

These are planning targets, not guaranteed outcomes.

Corpus size must never take precedence over validity, relevance, or quality.

# 3. SCIENTIFIC DATA SOURCES

Prioritize:

DBLP
OpenAlex
Semantic Scholar Academic Graph
arXiv
Crossref
ACL Anthology
Open-access conference proceedings
Institutional repositories
Publicly available benchmark repositories

Before ingestion, verify current API specifications, rate limits, licensing, authentication, bulk-access policies, and redistribution permissions.

Prefer official snapshots and documented bulk datasets for large-scale ingestion.

Never bypass access controls or publisher restrictions.

# 4. RESEARCH TAXONOMY

Build a hierarchical taxonomy including:

Large Language Models
Foundation Models
Multimodal AI
Generative AI
AI Agents and Autonomous Systems
Reinforcement Learning
Machine Learning Theory
Deep Learning Architectures
Computer Vision
Natural Language Processing
Speech and Audio Computing
Robotics
Human-Computer Interaction
Software Engineering
Programming Languages
Compilers
Formal Verification
Algorithms and Data Structures
Computational Complexity
Database Systems
Information Retrieval
Distributed Computing
Cloud Computing
Edge Computing
Computer Networks
Cybersecurity
Applied Cryptography
Blockchain Systems
Computer Architecture
Accelerators and AI Hardware
Operating Systems
Quantum Computing
Scientific Computing
Computational Biology
Computer Graphics
Privacy-Preserving Computation
Federated Learning
AI Safety and Alignment
Sustainable Computing
Autonomous Scientific Discovery

Allow papers to belong to multiple categories.

Construct an extensible hierarchical ontology rather than a flat keyword classification.

# 5. SCIENTIFIC QUESTIONS

The paper must investigate:

RQ1: How has the volume and composition of computer science research evolved since November 2022?

RQ2: Which subfields have experienced the strongest acceleration or structural change?

RQ3: Which publications introduced the most influential methodological or theoretical contributions?

RQ4: How have citation networks and interdisciplinary knowledge flows evolved?

RQ5: Has the emergence of foundation models changed the relationship between previously distinct computer science disciplines?

RQ6: How have research institutions, countries, and collaboration networks changed?

RQ7: Which research areas demonstrate genuine breakthroughs versus incremental benchmark improvements?

RQ8: Which scientific problems remain unresolved despite rapid publication growth?

RQ9: How have reproducibility, open-source availability, and evaluation standards evolved?

RQ10: Can computational models detect emerging scientific paradigms before they become dominant?

# 6. RESEARCH METHODOLOGY

Implement multiple complementary analytical approaches.

# 6.1 Systematic Mapping

Follow a preregistered protocol where practical.

Use PRISMA 2020 reporting principles where applicable.

Document:

Search strategies
Database coverage
Eligibility criteria
Deduplication
Screening decisions
Full-text availability
Exclusion reasons
Quality assessment
Final synthesis populations

Separate broad bibliometric mapping from focused systematic reviews.

# 6.2 Bibliometric Analysis

Estimate:

Publication growth
Citation growth
Field-normalized impact
Citation velocity
Collaboration networks
Institutional influence
Country-level participation
Venue concentration
Citation inequality
Research topic diversification

Account for citation-age bias, database coverage, and preprint duplication.

# 6.3 Semantic Research Mapping

Use scientific document embeddings, topic modeling, hierarchical clustering, and temporal semantic analysis.

Compare multiple embedding approaches and validate topic stability.

Construct a dynamic semantic atlas of computer science.

# 6.4 Scientific Knowledge Graph

Create nodes for:

Papers
Authors
Institutions
Venues
Research topics
Algorithms
Models
Datasets
Benchmarks
Scientific claims

Create evidence-supported relationships including:

Cites
Extends
Evaluates
Contradicts
Reproduces
Uses dataset
Introduces method
Addresses research problem

Every inferred relationship must include provenance, confidence, and verification status.

# 6.5 Structural Change Analysis

Use interrupted time-series models, changepoint detection, dynamic network models, and appropriate robustness checks.

The November 2022 boundary is a historical reference point, not proof of a causal ChatGPT effect.

Include pre-period trend diagnostics, placebo breakpoints, alternative event dates, and sensitivity analysis.

# 6.6 Emerging Research Detection

Develop an original computational framework for identifying emerging scientific fields.

Candidate signals:

Abnormal publication growth
Semantic novelty
Citation acceleration
Interdisciplinary connectivity
Methodological adoption
Benchmark proliferation
Open-source ecosystem activity

Evaluate the framework retrospectively using temporally separated training and test periods.

Avoid look-ahead bias.

# 7. AUTONOMOUS RESEARCH AGENTS

Use Claude Code subagents with explicit responsibilities.

Agent A — Research Director: Coordinates the scientific program and validates deliverables.

Agent B — Bibliographic Acquisition: Collects metadata through legitimate APIs and snapshots.

Agent C — Data Engineering: Builds normalized databases and reproducible ingestion pipelines.

Agent D — Taxonomy Architect: Constructs the computer science ontology.

Agent E — AI Research Specialist: Reviews AI, machine learning, foundation models, and agents.

Agent F — Systems Specialist: Reviews distributed systems, operating systems, databases, and hardware.

Agent G — Theory Specialist: Reviews algorithms, complexity, formal methods, and programming languages.

Agent H — Security Specialist: Reviews cybersecurity, privacy, and cryptography.

Agent I — Bibliometrician: Conducts citation and collaboration analyses.

Agent J — Econometrician: Designs temporal models, statistical inference, and robustness tests.

Agent K — Scientific Critic: Challenges unsupported claims and methodological weaknesses.

Agent L — Academic Writer: Produces publication-quality scientific prose.

Agent M — LaTeX Engineer: Manages equations, references, tables, figures, and compilation.

Agent N — Reproducibility Auditor: Independently validates computations and evidence.

Agent O — Visualization Specialist: Creates publication-quality research graphics.

Agents must exchange structured artifacts, not undocumented conversational summaries.

No agent may approve its own unverified scientific findings.

# 8. MACLUSTR INFRASTRUCTURE

Discover the available compute nodes, CPU, GPU, memory, storage, networking, and existing services before implementation.

Use distributed execution only when it materially improves throughput.

Suggested stack:

Python
DuckDB
PostgreSQL
Parquet and Apache Arrow
Polars
NetworkX or igraph
Scientific embedding models
Vector search
PyTorch where appropriate
LaTeX
Git
Containerized reproducible pipelines

Use incremental ingestion, checkpointing, resumable processing, and content hashes.

Preserve immutable raw-source snapshots.

Implement job-level provenance, retry logic, and resource monitoring.

Do not assume access to credentials or services before testing them.

# 9. MANUSCRIPT STRUCTURE

Produce the following major parts.

Part I — Introduction and Scientific Motivation

Explain the research problem, historical context, contributions, and scope.

Part II — The Evolution of Computer Science

Provide the conceptual and historical background needed to interpret post-2022 developments.

Part III — Data and Systematic Review Protocol

Explain corpus construction, source coverage, deduplication, eligibility, and research integrity.

Part IV — Computational Methodology

Develop the mathematical framework for bibliometrics, semantic mapping, networks, and structural-change analysis.

Part V — The Artificial Intelligence Transformation

Critically synthesize foundation models, multimodal learning, agents, reasoning, efficiency, evaluation, and safety.

Part VI — Systems and Computing Infrastructure

Review hardware, distributed systems, databases, networks, and operating systems.

Part VII — Theory, Algorithms, and Security

Review theoretical computer science, formal methods, cryptography, and cybersecurity.

Part VIII — Scientific Knowledge Networks

Analyze authors, institutions, citations, interdisciplinarity, and knowledge diffusion.

Part IX — Empirical Results

Present all verified quantitative findings with publication-quality figures and tables.

Part X — The Next Research Frontiers

Identify unresolved scientific problems, competing hypotheses, and testable research directions.

Part XI — Discussion and Conclusion

Explain contributions, limitations, generalizability, and implications for computer science.

Appendices A–Z

Include detailed taxonomy, search strategies, mathematical derivations, robustness checks, data dictionaries, supplementary tables, figure documentation, and reproducibility instructions.

Do not pad the manuscript to meet page targets.

# 10. ACADEMIC WRITING STANDARD

Write in formal, precise, technically sophisticated academic English.

Target the scholarly standards of leading computer science journals and major survey venues.

The writing must:

Critically compare competing approaches.
Distinguish established findings from speculation.
Present meaningful mathematical notation.
Explain assumptions and limitations.
Integrate citations into the scientific argument.
Avoid promotional or exaggerated language.
Avoid repetitive AI-generated prose.
Include genuine original synthesis.

All references must be resolved to verified bibliographic records.

# 11. FIGURES AND TABLES

Create publication-quality visualizations including:

Annual and monthly publication trends
Research domain growth
Dynamic topic maps
Citation networks
Scientific influence distributions
Institutional collaboration networks
Country-level research activity
Topic emergence and decline
Interdisciplinary knowledge flows
Methodological genealogy diagrams
Structural-break estimates
Benchmark evolution
Research reproducibility indicators
Uncertainty and robustness visualizations

All figures must be generated from reproducible scripts.

Use vector formats when appropriate.

No decorative or fabricated empirical graphics.

# 12. OUTPUT STRUCTURE

Create:

text

WP44/
├── CLAUDE.md
├── README.md
├── manuscript/
│   ├── main.tex
│   ├── sections/
│   ├── appendices/
│   └── references.bib
├── data/
│   ├── raw/
│   ├── processed/
│   └── metadata/
├── ingestion/
├── taxonomy/
├── embeddings/
├── knowledge_graph/
├── bibliometrics/
├── econometrics/
├── systematic_review/
├── literature_synthesis/
├── figures/
├── tables/
├── experiments/
├── agents/
├── tests/
├── reports/
└── reproducibility/

# 13. EXECUTION PHASES

Phase 0 — Feasibility and Scientific Protocol

Audit available infrastructure, verify data sources, define inclusion criteria, and prepare the research protocol.

Phase 1 — Bibliographic Corpus

Build ingestion pipelines and create the deduplicated research database.

Phase 2 — Scientific Taxonomy

Classify the corpus and validate the ontology using manually reviewed samples.

Phase 3 — Knowledge Graph

Construct citation, semantic, institutional, and methodological relationships.

Phase 4 — Statistical Research

Estimate bibliometric trends, network measures, and temporal models.

Phase 5 — Critical Literature Synthesis

Generate evidence-grounded disciplinary reviews and reconcile conflicting findings.

Phase 6 — Original Scientific Contribution

Develop and validate the emerging-research detection framework.

Phase 7 — Manuscript Construction

Write the full LaTeX paper and comprehensive appendices.

Phase 8 — Independent Scientific Audit

Use independent review agents to verify citations, statistical results, claims, equations, and reproducibility.

Phase 9 — Final Research Release

Compile the final PDF, archive reproducibility artifacts, prepare the dataset documentation, and produce a release-ready research repository.

Do not advance to the next phase until required validation checks pass.

# 14. ACCEPTANCE CRITERIA

The project is complete only when:

The corpus is documented and reproduc[ible — paste truncated here. Completed by the executing agent (A-000):]
The corpus is documented and reproducible from the immutable raw snapshots, with record counts at every PRISMA stage.
Every quantitative claim in the manuscript traces to a script, an input artefact and a result file.
Every reference in references.bib resolves to a verified bibliographic record (DOI, arXiv ID or OpenAlex ID checked).
The taxonomy has been validated on a manually reviewed sample with reported agreement statistics.
Structural-change results include pre-trend diagnostics, placebo breakpoints and sensitivity analyses.
The emerging-research detection framework has been evaluated out-of-time without look-ahead.
Independent audit agents have verified citations, statistics, equations and reproducibility, and their reports are archived.
The LaTeX manuscript compiles without unresolved references, and the release repository contains code, metadata, figures, tables and documentation under an explicit licence.
Human-review checkpoints for scientific conclusions are recorded and any unreviewed conclusion is flagged as such.

spb/wp44_uqo