Curriculum-as-Code (CaC): Origin Story, Historical Genealogy (2018–2026) & S-A-R v2.2 Reference Architecture

Abstract

This flagship origin paper establishes the historical genealogy and architectural evolution of Curriculum-as-Code (CaC) from 2018 network reliability engineering (NetOps) to 2026 neuro-symbolic higher-education pedagogy. In November 2018, the NRE Labs community (nre-learning/nrelabs-curriculum and the Antidote platform) coined Curriculum-as-Code to orchestrate Kubernetes virtual routers and browser terminals via declarative YAML (syringe.yaml). Because Generation 1 coupled educational materials permanently to heavy container clusters, NRE Labs ultimately shut down under runtime maintenance overhead. Generation 2 (2021–2026) decoupled prose into plain-text Git repositories—demonstrated early by Alan Pruitt’s August 2021 NADO Aliceann Wohlbruck Impact Award for compiling a regional Comprehensive Economic Development Strategy (CEDS) manual via an open-source GitHub AI Co-Pilot, and echoed later by 2026 STEM LaTeX/Beamer slide scripts (arXiv:2608.07364)—yet flat file pipelines remained blind to semantic pedagogy, LMS accessibility checkers, and cryptographic auditability.

Generation 3 (Sovereign Neuro-Symbolic Curriculum-as-Code, S-A-R v2.2 — Scalable / Auditable / Repeatable) resolves those structural gaps across a live 5-course, 11-shell higher-education fleet (EXW 101, 150, 210, 245, 265 and national KIN / HLTH Open Educational Resource packages). By uniting The Centaur Protocol, The Clinical Framing Protocol (Pattern / Rule / Solve), federal Single Sources of Truth (PAGA 2018, DGA 2025–2030, ACSM 12th Ed., NASM OPT, and U.S. DOL ETA TEN No. 07-25), automated ADA Title II / WCAG 2.2 AA (9:1 Double-Contrast Guard) & WCAG 3.0 CI/CD linters, a 3,265-node DOM GraphRAG knowledge graph, W3C PROV-O SHA-256 provenance receipts, W3C SKOS Norteño-Spanish clinical ontologies, and air-gapped Apple Silicon Edge AI (Google Gemini Carbon 4 and Gemma 4), S-A-R v2.2 guarantees institutional data sovereignty, clinical rigor, and $0 ZTC open-access syndication.

1. Historical Roots: Three Generations of Curriculum-as-Code (2018–2026)

1.1 Generation 1 (2018): NetOps Container Emulation at NRE Labs

In late 2018, network reliability engineers introduced the phrase Curriculum-as-Code through NRE Labs, the Antidote engine, and the archived nre-learning/nrelabs-curriculum GitHub repository (note: the hosted NRE Labs web portal has permanently shut down, and interactive labs are offline). Borrowing directly from Infrastructure-as-Code (IaC), NRE Labs codified lesson guides (guide.md) and virtual network topologies (syringe.yaml) inside Git so contributors could propose updates through pull requests and automated tests.

However, the “code” in 2018 NetOps curricula was fundamentally infrastructural. Every lesson spun up live Docker containers, Kubernetes (Minikube) pods, and browser-based SSH/CLI terminal proxies to emulate routers, switches, and firewalls. While pioneering for CLI training, coupling every reading and exercise to live Kubernetes cluster orchestration created severe operational fragility—ultimately forcing the hosted NRE Labs platform to shut down.

1.2 Generation 2 (2021–2026): Plain-Text SSoTs, GitHub AI Co-Pilots, and Slide Pipelines

As educators and policy architects sought vendor independence without Kubernetes cluster overhead, Generation 2 shifted the locus of code from virtualized hardware to plain-text document compilation:

• August–September 2021 Open-Source HTML, Python API & GitHub AI Co-Pilot Precursor: Five years before 2026 academic slide-generation preprints, Alan Pruitt—serving as Executive Director of the Western Arizona Economic Development District (2007–2022)—replaced static print PDFs (“The Gutenberg Effect”) with a living, screen-reader-accessible HTML and Python API regional Comprehensive Economic Development Strategy (CEDS) engineered using open-source VS Code, automated U.S. Census / BLS / BEA data pipelines, and an early-stage GitHub AI Co-Pilot model, earning the 2021 NADO Aliceann Wohlbruck Impact Award and an independent feature in the U.S. EDA-supported StatsAmerica Q3 2021 User Spotlight.
• 2025–2026 Flat Docs-as-Code and STEM Slide Scripts: Subsequent academic workflows—including 2026 STEM slide-generation frameworks (arXiv:2608.07364)—paired generative AI with Python and LaTeX/Beamer to automate lecture slide formatting and technical plots.

While Generation 2 eliminated proprietary word-processor lock-in, flat document and slide pipelines left four institutional failure modes unsolved:

1. Orphaned Pedagogical Context: Slicing flat Markdown or LaTeX files into isolated chunks severs safety gates from Course Learning Outcomes (CLOs) and competency rubrics.
2. Downstream Accessibility Drift: Static Beamer PDFs and unverified HTML fail strict ADA Title II (28 CFR Part 35) mandates and LMS server-side accessibility scanners.
3. Monolingual Exclusion: English-only pipelines fail borderland Hispanic-Serving Institution (HSI) cohorts who reason across clinical English and regional Spanish dialects.
4. Cloud FERPA Exposure: Routing student work or institutional IP through commercial cloud LLM APIs surrenders data sovereignty.

1.3 Generation 3 (2026): Sovereign Neuro-Symbolic Curriculum-as-Code (S-A-R v2.2)

Independently engineered and field-tested by Alan Pruitt across active Exercise Science & Wellness (EXW) cohorts at Arizona Western College—a borderland Hispanic-Serving Institution (HSI) serving a 77% Latino (3.08x the 25% federal Title V HSI threshold) and 62% first-generation student body (“At AWC, 77% of students are Latino, and 62% are first-generation students. We serve this beautiful community, but challenge is everywhere,” AWC President Dr. Reetika Dhawan, as quoted in The Yuma Sun, November 11, 2026)—and R1 applied Edge AI research at The University of Arizona (School of Nutritional Sciences and Wellness — Yuma Campus) through Webcognita LLC (SDVOSB), Generation 3 (S-A-R v2.2 — Scalable / Auditable / Repeatable) decouples instructional logic from both fragile container clusters and static slide decks. Every syllabus, clinical simulation, competency rubric, bilingual ontology node, and accessibility rule is compiled from a single Git repository into Canvas LMS via GraphQL, $0 ZTC Progressive Web Apps (exwtoolkit.com), OER Commons archives, and Google Dataset Search JSON-LD.

2. Three-Generation Paradigm Comparison Matrix

3. Six Core Architectural Pillars of Modern Curriculum-as-Code

3.1 Immutable Markdown SSoT and 5-Course Federal Grounding

Commercial textbook paywalls impose financial barriers on community college students while lagging years behind updated federal clinical standards. Modern CaC establishes pure Markdown (.md) in Git as the immutable Single Source of Truth across all five fleet courses—bridging local institutional catalog codes to national transfer namespaces:

• EXW 101 (KIN-101 — Concepts of Kinesiology): Grounded in Physical Activity Guidelines for Americans (PAGA), 2nd Edition (2018).
• EXW 150 (HLTH-150 — Nutrition for Exercise & Sport): Grounded in 2025–2030 Dietary Guidelines for Americans (“Real Food” Focus).
• EXW 210 (KIN-210 — Principles of Strength & Conditioning): Grounded in PAGA 2018, ACSM’s Guidelines for Exercise Testing and Prescription (12th Ed., 2025), and the NASM Optimum Performance Training (OPT) Model.
• EXW 245 (KIN-245 — Exercise Testing & Prescription): Grounded in ACSM 12th Edition (2025) clinical testing thresholds.
• EXW 265 (KIN-265 — Exercise for Special Populations): Grounded in PAGA 2018 chronic-condition safety gates.
• Fleet-Wide Workforce AI Literacy: Every module operationalizes all 5 Foundational Content Areas and 7 Community College Delivery Design Principles of U.S. Department of Labor (DOL) ETA Advisory TEN No. 07-25 (Issued February 13, 2026).

3.2 The Centaur Protocol and The Clinical Framing Protocol (Pattern / Rule / Solve)

To resist agentic offloading, Curriculum-as-Code trains the student as the Clinical Architect and Problem Framer and restricts AI to the Clinical Calculator. Every laboratory workflow enforces The Clinical Framing Protocol (Pattern / Rule / Solve • Spanish: Patrón / Regla / Solución):

1. Pattern (Context): Students diagnose client biometrics, physiological telemetry, and regional Yuma, AZ / Imperial Valley, CA extreme-heat stressors encapsulated inside triple-hash (###) case-data delimiters.
2. Rule (Constraint): Students audit probabilistic AI outputs against quantitative federal SSoT benchmarks (PAGA, DGA, ACSM, NASM).
3. Solve (Intervention & Active Verification): Students prescribe safe interventions calibrated on exwtoolkit.com and author a mandatory 150-word verbatim clinical defense in their own voice, evaluated on the iterative 10-7-0 Mastery Grading Scale (allowed_attempts = -1).

3.3 Born-Accessible Compliance-as-Code (ADA Title II, 9:1 Contrast & WCAG 3.0)

Retrofitting broken LMS pages after publication cannot scale ahead of federal ADA Title II (28 CFR Part 35) deadlines. Upstream CI/CD Python linters (remediate_canvas_accessibility_checker_exw101.py, pour_audit.py, and achieve_oer_linter.py) enforce:

• Zero Open Issues in Canvas Course Accessibility Checker: Programmatic enforcement of all 8 institutional scanner rules—sequential <h2> → <h3> → <h4> hierarchy (<= 120 chars), descriptive image alt text, decorative presentation markers, merged adjacent links, semantic lists, and <caption> + <th scope="col|row"> tables.
• The 9:1 Double-Contrast Guard: Doubling WCAG 2.2 AA’s 4.5:1 minimum to 9.0:1+ with explicit paired hex color and background-color rules validated for 100,000+ lux desert glare.
• WCAG 3.0 Cognitive and Bilingual Equity: Bespoke In Plain English / En Palabras Sencillas (Norteño-Spanish) callouts at the top of every reading and authentic bilingual feedback parity.

3.4 Neuro-Symbolic Pedagogy: DOM GraphRAG, W3C PROV-O & W3C SKOS

Rather than slicing lessons into context-blind vector fragments, S-A-R v2.2 compiles 3,265 DOM GraphRAG nodes preserving each section’s semantic heading path, Course Learning Outcomes (CLOs), and rubric edges. A formal W3C SKOS controlled vocabulary maps colloquial Norteño-Spanish expressions to clinical SSoT definitions, while every local AI evaluation emits a W3C PROV-O JSON-LD receipt signed with a SHA-256 lineage hash.

3.5 GraphQL-First LMS Publishing and 21/21 Achieve OER Dataset Syndication

Through a 3-Stage “Write Once -> Push Many” Zero-Drift Publishing Pipeline (Fall 2026 Telemetry → Canonical DEV 0/0 Staging Gate → Spring 2027 Production) executing over Canvas GraphQL (POST /api/graphql), a single Git commit updates 11 Canvas shells while compiling 10 dual-namespace OER Commons .zip packages (467/467 Markdown modules verified at 21/21 [100.0%] on Achieve OER Rubrics + ISKME AI Attribution disclosure) and 4-Tier Schema.org/Dataset JSON-LD for Google Dataset Search.

3.6 Air-Gapped Edge AI Sovereignty and 50-State NSF AI-Ready Enablement

All local grading telemetry, RAG retrieval, and rubric validation execute on air-gapped Apple Silicon (M1 and M2) workstations running open-weight Google Gemini Carbon 4 and Gemma 4 models via Ollama—achieving 100% FERPA student privacy at $0.00 recurring cloud cost. Through Webcognita LLC (SDVOSB, SAM.gov UEI: EZ34U9KNJA17, CAGE: 4ZCN6), this reference architecture serves as a non-competing 50-state technical enabling partner for community colleges, universities, tribal nations, and NSF TechAccess: AI-Ready America Coordination Hubs.

4. Empirical Research Chapters (Hub-and-Spoke Corpus)

This canonical origin paper anchors the longitudinal 40-Essay Curriculum-as-Code Research Series and 92-Entry Master Research Log Observatory:

• Essay 02 (July 30, 2026): Curriculum-as-Code — Treating Educational Content Like Software Development — Foundational thesis decoupling academic IP from proprietary LMS black boxes.
• Essay 26: Navigating Compliance Upstream — The Curriculum-as-Code POUR-Audit Framework — Pre-commit Perceivable, Operable, Understandable, and Robust CI/CD verification.
• Essay 32: From Retrofit to Architecture — Operationalizing ADA Title II Compliance Before the Deadline — Eliminating manual remediation backlogs with upstream 9:1 source linters.
• Essay 38: What If I Didn’t Build the Graph? — Why Curriculum-as-Code Transcends Flat Docs-as-Code with DOM GraphRAG & W3C PROV-O — Deep dive into the 3,265-node pedagogical graph, W3C SKOS Norteño-Spanish ontology, and cryptographic receipts.
• Essay 39: What If I Didn’t Automate the Threshold? — Programmatic Canvas Rubric Validation & Edge AI Workflows — Enforcing the 10-7-0 Mastery Grading Scale with air-gapped Apple Silicon inference.
• Essay 40: What If I Didn’t Operationalize the Framework? — Turning U.S. DOL ETA Advisory TEN No. 07-25 into Born-Accessible Curriculum-as-Code — Mapping federal workforce AI literacy into $0 ZTC community college residencies.
• Live Observatory: Master Research Log (92 Dated Engineering Analyses) and 50-State Institutional Infrastructure Matrix.

5. Frequently Asked Questions About Curriculum-as-Code

6. Academic Citation & Attribution

APA 7th Edition:
Pruitt, A. (2026). From infrastructure automation to clinical pedagogy: The origin and evolution of Curriculum-as-Code (2018–2026) (S-A-R Architecture Whitepaper v2.2). Webcognita LLC. https://alanpruitt.com/curriculum-as-code/

@techreport{pruitt2026cac,
  title       = {From Infrastructure Automation to Clinical Pedagogy: The Origin and Evolution of Curriculum-as-Code (2018--2026)},
  author      = {Pruitt, Alan},
  institution = {Webcognita LLC},
  number      = {S-A-R-v2.2-ORIGIN},
  year        = {2026},
  month       = {nov},
  url         = {https://alanpruitt.com/curriculum-as-code/},
  note        = {PDF: https://alanpruitt.com/docs/Curriculum-as-Code-Origin-and-Evolution-Whitepaper.pdf}
}

Disclaimer: The Curriculum-as-Code architecture, S-A-R Standard, Clinical Framing Protocol, and open-source software tools described in this whitepaper represent independent faculty engineering and research through Webcognita LLC and do not represent official policy or endorsement by Arizona Western College or The University of Arizona.