Remove obsolete community-notes sections from current Zoomcamp lessons

Outcome

Remove the obsolete “Community notes” contribution furniture from every current, non-legacy Zoomcamp curriculum and prevent it from returning through zoomcamp-ops. This is a source-content and source-contract change, not a website-template suppression.

Normative references:

• zoomcamp-ops/docs/shared-curriculum-v2.md
• zoomcamp-ops/STRUCTURE.md
• zoomcamp-ops/templates/unit.md
• _docs/runbooks/course-content-push-and-pull.md
• _docs/architecture/database-only-content.md

Boundary and content decision

• “Current, non-legacy” means the public root curriculum: for schema v2, root NN-module material selected by course.yaml.current_cohort/curriculum: current; for schema v1, the existing root current curriculum graph. It includes lesson Markdown and GitHub-facing current module indexes, and live or self-paced delivery.
• legacy: true, curriculum: github_archive, and historical cohorts/<id> trees are excluded and immutable. Stock Markets Analytics remains schema v1; removing its one obsolete current-root section does not authorize a schema migration.
• Remove normalized Community notes headings, “Did you take notes?”/“Add links above this line” prompts, empty sentinels, and note links whose only purpose is that obsolete contribution block.
• Do not silently discard a resource that is part of the taught material. During inventory, a link with independent curricular value must either be moved under a specific topic/resources heading with a short reviewed rationale or be escalated for course-owner decision. Generic personal-note collections are removed.
• Machine Learning’s 86 ordinary ## Notes lesson sections contain substantial inline learner-authored summaries/code and are explicitly preserved. They are not contribution placeholders and must not be matched by the new rule.

Scope

1. Inventory every occurrence, case-insensitively, across current root lesson files and current module README.md files in all six public Zoomcamp repositories; classify empty prompt, note-link collection, or potential curricular resource. The grooming audit found 32 affected current files: AI Dev Tools 4, Data Engineering 7, LLM 5, Machine Learning 9, MLOps 6, and Stock Markets Analytics 1; implementation must reconcile to that baseline or explain drift.
2. Apply the reviewed removal/reclassification in each of those six source repositories.
3. Update zoomcamp-ops templates, structure/curriculum guidance, checker, diagnostic catalogue, fixtures, and tests so current v2 material cannot reintroduce this obsolete section/prompt. The current templates/unit.md instruction that community notes “stay” must be removed/replaced.
4. Push focused reviewed commits per source repository and a separate zoomcamp-ops contract commit, then sync the website from accepted source SHAs and verify rendered current lessons.

Non-goals

• Editing any legacy/archive cohort or migrating the schema-v1 Stock Markets course.
• Removing ordinary authored uses of the words “community” or “notes” in lesson prose.
• Deleting topic-specific external resources merely because a community member created them.
• Removing or rewriting Machine Learning’s 86 substantive ## Notes lesson sections.
• Hiding source content with CSS/template filters or adding a website-side sanitizer/fallback.
• Combining changes with actual-homework publication (#423) or video/navigation work (#425).
• Redesigning lesson pages.

Dependencies and safety

• Land the zoomcamp-ops contract/checker change before or with source cleanups so regressions fail CI. Source repositories must consume the accepted checker revision according to their existing workflow.
• AI Dev Tools and Data Engineering had unrelated dirty work during grooming. Use clean pinned worktrees or coordinate ownership; never stash, overwrite, or mix those edits.
• #423 and #425 overlap some repositories but are not hard dependencies. Keep separate commits, validation, sync runs, and rollback units.

Acceptance criteria

• A repository/path inventory reconciles the 32-file/six-repository grooming baseline and records why each matched link/block was removed, retained/reclassified, or escalated.
• Current root material contains no normalized Community notes heading and no solicitation/sentinel such as “Did you take notes?”, “Share community notes…”, or “Add … notes above this line.”
• Note links that existed only inside the obsolete contribution block are removed. Any retained external resource has a topic-specific placement and a recorded curricular rationale.
• No file under a legacy: true, github_archive, or historical cohort tree changes. Stock Markets changes only its one reviewed current-root obsolete section; its schema/route/content model does not migrate.
• All 86 substantive Machine Learning ## Notes sections remain present and byte-stable unless a source-owner-approved move is recorded; the checker treats them as valid content.
• zoomcamp-ops no longer recommends Community notes in templates/unit.md or normative guidance. New current v2 lessons/module indexes containing the normalized obsolete heading or sentinel fail with a named, actionable diagnostic; archive fixtures remain accepted.
• Checker tests cover heading case/level variations, exact/near sentinel phrases, v1/v2 current lesson and module-index scope, ordinary prose false positives, substantive ## Notes, retained topic-specific resources, and archive exclusion.
• Every affected source repository passes its own checks plus the accepted zoomcamp-ops checker. Commits are focused and pushed to main; SHAs and rollback commits are recorded.
• Website sync/import from those SHAs succeeds and replays as a no-op. Rendered current lesson bodies contain no obsolete block; database-owned source provenance points to the accepted commit.
• Current lesson/module routes, headings, images, code links, navigation, and legacy routes remain unchanged apart from the reviewed content removal/reclassification.

Verification scenarios

Repository

1. Run the inventory before and after; after-state is zero obsolete matches in current scope, 86 preserved ML ## Notes sections, and an unchanged count/digest for archive scope.
2. Add checker fixtures for current and archived versions of the same heading/prompt; assert only current content is rejected.
3. Run each affected repository’s prescribed checks from a clean worktree, then run the website course-source parser/import tests.
4. Sync each accepted SHA twice into an isolated database and compare row identities/counts/provenance.

Browser

1. Open representative current lessons that previously showed an empty prompt, a note-link collection, and a reclassified curricular resource; compare with the source decision table.
2. Verify desktop and 390px mobile in light/dark mode: no blank section gap, stray divider, or broken adjacent heading/navigation remains.
3. Open a representative legacy/archive cohort and verify its route/content is unchanged.

Blocker rule

If a link’s curricular value cannot be determined from the surrounding current source, stop on that link and request course-owner disposition rather than deleting it by assumption.

Evidence to attach

• Before/after inventory with source SHAs and classifications.
• zoomcamp-ops contract/checker test output and per-course check output.
• Website sync/replay output and representative rendered screenshots.

Read the linked zoomcamp-ops docs, especially templates/unit.md, STRUCTURE.md, docs/shared-curriculum-v2.md, and the checker guidance, then inventory the six repositories against the 32-file baseline. Use the current-root and archive boundaries before changing source Markdown. Done means reviewed removals or reclassifications, preserved ML Notes and archives, passing checker tests, and successful website sync verification.