Historical learning-support review: edition 0.7.0
Archived maintainer evidence for PR 4. For learner entry, use start here.
Edition 0.7.0, October 25, 2026. Implemented locally from the integrated 0.6.0 course. Changes address the critical review's ten improvement areas.
| Review area | Implemented evidence |
|---|---|
| Entry preparation | Python Foundation first, eight-item readiness diagnostic, three routes and a file/table bridge |
| Conceptual scaffolding | Worked mental models, numerical examples, glossary and construction in all 14 notebooks |
| Independent construction | Guided pages, required purposeful changes, diagnosed failures and transfer evidence |
| Mastery and feedback | Module-specific misconception questions, explanations and developing/meets/extends criteria |
| Execution support | Preflight, explicit bootstrap, acquisition/hosted/troubleshooting guidance and private-source access limits |
| Distribution navigation | Source-browser download links lead to artifact access; generated site retains actual local downloads |
| Statistical reasoning | Validation calibration tables, sample-size/interval exercise and explicit independence assumptions |
| Comparable model evidence | Same-ID/same-content lexical versus LLM comparison; base/adapted generation alongside completion loss |
| IT transfer and depth | Real UCI phishing records, grouped split, four models, separate pipeline stages and combined notebook |
| Synthesis and teaching validation | Project milestones, 14-module pacing, private source reconciliation, pilot instrument and browser regression checks |
Local verification
Core Python 3.12 on Linux/WSL: 76 tests discovered, 66 passed, 10 framework tests intentionally skipped. The separate framework run passed its original nine checks; the added matched-generation check passed separately. Those checks cover actual TensorFlow training/reload, local generation, adapters, public CLIs, app forms and HTTP health. The complete updated suite is also required in commit-specific CI.
All 14 registered notebooks execute in fresh matching kernels: 56 core and 48 framework teaching code cells. The separate phishing notebook adds six executable cells and uses the same staged functions as its public CLI. Source notebooks retain no outputs. Real data-cache hashes, extracted-source CLI execution and isolated installed-wheel inference/schema/cache checks passed locally.
Source Markdown and generated HTML links are checked separately. Generated site, HTTP routes, byte-matched archive membership and browser evidence are checked after final source staging. Consult commit-specific CI for final page/file counts and the exact pushed commit; intermediate counts are not substituted for final evidence.
Teaching limits still requiring observation or ownership
Learner piloting and workload validation, human assistive-technology/real-zoom/notebook-interface checks, live Codespaces/Colab/Binder sessions, institutional outcome approval, current Canvas policies need actual owner/human evidence. The provided instruments and configurations are not reported as completed learner studies or live provider verification. Python Foundation supports the no-programming-entry design and does not create a formal prerequisite or silently replace an institutionally assigned resource.
Local prompts and authentic benchmark implementations do not establish every broader original objective: closed-provider API execution and full agent-framework implementation remain explicit adoption decisions. No production, accreditation or WCAG-conformance claim is made. Publication/visibility changes are separate from source integration.