execution-plan/notebooks/source-checks.md · UPDATED 2026-10-01
Notebook source checks — 2026-09-23
The weekly notebooks reuse the existing local specifications, v6 chapters and resource cards. This is a targeted verification log, not a claim that every external card, paper, video or dataset was read or tested. Historical verification dates in the original materials remain unchanged.
| Source | Check and use |
|---|---|
| Google source import guidance | Official page retrieved. Text, Markdown and pasted text are supported source routes. Linked pages are not recursively imported; check the imported source content. Product/interface wording may differ from the programme’s NotebookLM name. |
| OpenStax Calculus 5.1 | Official section retrieved; optional foundation repair for accumulation and area approximations. The notebook’s pump and biomass examples are original teaching examples, not research results. |
| SciPy solve_ivp | Official API page retrieved. Use the documented function interface and examples with your installed version; the plan’s particular predator–prey example may differ by documentation version. |
| scikit-learn common pitfalls | Official guide retrieved; supports the preprocessing/leakage checks. No installed-version compatibility or project experiments were tested by browsing it. |
| Conformal prediction tutorial | Primary paper record retrieved. Use sections 1–2 and verify the applicable calibration rule and assumptions before implementing; shifted-data coverage is not assumed. |
| FLIP author repository | Verified working author repository via the original FLIP paper. The inherited github.com/J-SN/FLIP returned 404. New notebooks use the corrected locator where that URL occurred. Existing master plans/cards remain preserved. Inspect task-specific split guidance before comparing scores. |
| Europe PMC API documentation | Main documentation request returned 403 in the browsing tool. The provider’s development documentation was readable and points to the official EBI search endpoint. Do not assume the development site is a production data source; verify current production access when executing. |
| statsmodels mixed effects | Official documentation retrieved. Consult model assumptions and examples for the specific field-trial design. This does not validate a future analysis. |
| Zhang, Chung and Oldenburg 1999 | Primary publisher record located; Broad-hosted paper PDF accessible. Reference for Z-prime/control separation, not a universal assay acceptance policy. |
| arXiv endorsement | Official guidance retrieved. Recheck category eligibility at execution; endorsement, moderation and peer review are distinct. |
| JOSS submissions | Official guidance retrieved. Eligibility and submission requirements must be checked again against the actual software at Week 36. |
The original weekly plans already correct several aspirational master-plan statements. The notebooks retain those corrections: IndPenSim is simulated; lag requires an explicit mechanism; DO alone may not identify kLa; exact SINDy recovery is not guaranteed; computational rankings are not wet-lab validation; external events need actual evidence.
For an inaccessible source, record the title, attempted URL, access date and the exact gap. Use a free primary-source substitute addressing the same concept, or keep the relevant task pending. Never describe a search link as an inspected source or fabricate an exact video timestamp.
No ## parts in this file — use Edit whole file to change it.