DROP 4 — Masterclasses Section + Stories + Whitepapers
**88 publishable files** routed into a new `/masterclasses` section plus `stories` and `whitepapers`.
**16 files quarantined** in `_review/` — do not publish without a decision (see §4).
Source: `2050 Planet` Desktop folder, 124 root files.
`NWE_backups/` (7,198 hourly code backups) ignored — not content.
---
1. What's in this drop
```
content/
masterclasses/
early-learners/ 13 files ages ~6-11
gen-2050/ 19 files ages ~12-18
professionals/ 33 files adult
whitepapers/ 16 files
stories/ 7 files narrative fiction
_review/ 16 files NOT FOR PUBLISH — see §4
manifest.json full source→dest map with word counts
```
**Total words routed: ~230,000.**
---
2. The architecture finding — read this before building
The content is **not** three flat lists. It is **19 topic families**, most written
deliberately at three reading levels. Example:
| Family | Early Learners | Gen 2050 | Professionals | Whitepaper |
|---|---|---|---|---|
| solar | `sunny-power` | `solar-for-real` | `solar-energy-masterclass` | `the-solar-imperative` |
| algorithms | `what-is-an-algorithm` | `algorithms-masterclass` | `algorithms-masterclass-complete` | `the-algorithmic-age` |
| ai-datacenters | `where-does-ai-live` | `ai-datacenters-uae` | `ai-datacenters-masterclass` | `ai-datacenters-uae-whitepaper` |
**8 families have a complete early→gen→pro ladder.** That laddering is the single most
valuable structural asset in this content set and the section must expose it — a learner
should be able to sit on one topic and move up a tier, and a ministry reviewer should be
able to see one subject covered cradle-to-grave on a single screen.
Build the section with **two axes**: browse by tier (the three tracks) *and* browse by
family (the ladder). Do not ship tier-only navigation — it discards the ladder.
Gaps worth noting: `carbon`, `tokenization`, `cities`, `futures`, `ecosystems` are
professional-only. `ai-tools` and `chatbots` have no early-learner rung.
---
3. Claude Code instructions — copy-paste
```
Build the Masterclasses section from the Drop 4 content package.
CONTENT
Copy content/masterclasses/, content/stories/, content/whitepapers/ into the
repo content tree. Every file already has YAML frontmatter:
title, slug, family, tier, readingMinutes, wordCount, status, locale.
Do NOT copy _review/ — those are quarantined pending a decision.
Before writing, check the naming convention of each destination folder with
ls | sort -V | tail -5
Some existing folders use N-slug.md numeric prefixes, others plain slugs.
Match whatever is already there; these files ship as plain descriptive slugs.
ROUTES — additive only, never remove a menu item
/masterclasses section landing
/masterclasses/[tier] early-learners | gen-2050 | professionals
/masterclasses/[tier]/[slug] article
/masterclasses/topics/[family] the ladder view — all tiers for one topic
Mirror the existing section architecture (page.tsx / [slug]/page.tsx / registry
or index pattern) — copy whatever tech-ai/ does rather than inventing a new shape.
REQUIREMENTS
- Turquoise design tokens, existing type scale, no new colours
- RTL-safe markup throughout (Arabic + Urdu are mandatory targets)
- Three Scenarios pattern where source content supports it
- readingMinutes surfaced on cards
- Tier badge + family badge on every article card
- Ladder component on article pages: "also available at" linking sibling tiers
BRAND WALL — locked decision #1
Zero user-facing mentions of Claude or Anthropic. 32 files were flagged; the
resolution is NOT a blind find-and-replace (see README §5 — "Claude Shannon"
must survive). Apply the decision recorded there, then verify:
grep -rin "claude\|anthropic" content/ app/ components/ \
| grep -vi "claude shannon\|philanthropic"
Expected result: zero rows.
REPORT BACK
files copied, new total article count, any frontmatter that failed to parse,
brand-wall grep output.
```
---
4. `_review/` — 16 quarantined files, decision needed
**Near-duplicate pairs** (token overlap in parentheses) — pick one, discard or merge:
| Keep candidate | Quarantined variant | Overlap |
|---|---|---|
| `ai-system-types-complete-guide` | `ai-system-types-masterclass` | 0.83 |
| `fintech-masterclass` | `fintech-masterclass-udemy`, `fintech-masterclass-updated` | 0.89 |
| `ai-abundance-manifesto` | `ai-abundance-manifesto-v1` | 0.58 |
| `metaverse-masterclass` | `metaverse-masterclass-v1` | — |
| `metaverse-guide` | `metaverse-no-cap` | 0.60 |
| `copy-nature-self-study` | `copy-nature-genz-self-study`, `natures-super-secrets-self-study` | — |
`complete-masterclass-UNTITLED` overlaps `the-monetary-system` at 0.65 but has no
self-evident subject — it needs a human read before it can be titled or placed.
**Also quarantined, needs a home rather than a decision:** `15c-challenge-overview`,
`freiburg-case-study`, `water-innovations`, `stormwater-einstein-roof`,
`amazon-living-time-machine`, `ai-amazon-symbiosis`, `ai-tools-2025-complete`.
These are strong standalone articles that belong in existing sections
(`environment/`, `tech-ai/`) rather than in Masterclasses.
---
5. Brand wall — this one needs your call, not a script
143 real occurrences across 32 files. **A blind scrub is the wrong move**, for two reasons:
**False positives that must survive.** `The_Algorithmic_Age_White_Paper` cites
**Claude Shannon** as a Dartmouth 1956 organiser — that is the founding history of
the field and removing it would be a factual error. `Earth_Future_White_Paper`
triggered only on **philanthropic**. Neither is a violation.
**The real occurrences are mostly vendor-comparison tables** — rows like
"Blog posts → ChatGPT, Claude, Jasper" and curriculum items comparing assistants
head-to-head. Deleting one vendor from a comparison table while leaving ChatGPT,
Copilot, Jasper and Gemini in place doesn't produce neutral content; it produces
content with a conspicuous hole in it, and a reader who knows the landscape will
notice the omission. For a platform whose credibility rests on being an
authoritative educational reference for a ministry audience, that's a real cost —
worth weighing against the brand-wall rule rather than resolving on autopilot.
Three options:
1. **Generic substitution** — "Claude" → "frontier assistants", "Anthropic API" →
"commercial LLM APIs". Preserves the teaching point, loses specificity. Safest
against the locked rule.
2. **Delete the vendor rows entirely.** Cleanest compliance, thinnest content.
3. **Keep third-party vendor names as neutral market fact, scrub only where the
text reads as endorsement or as TERRA's own provenance.** Highest editorial
quality; requires accepting that decision #1 governs TERRA's self-description
rather than all third-party mention.
I've applied **none** of these — the files in `content/` still contain the original
text. Tell me which option and I'll run it, or run it yourself with the grep in §3
as the acceptance test.
The 5 heaviest files are `ai-creation-guide` (23), `ai-tools-masterclass-complete`
(15), `the-last-human-intern` (13), `ai-tools-2025-complete` (12),
`ai-tools-masterclass` (11) — 74 of the 143 occurrences sit in those five.
---
6. Processing notes
- **12 byte-identical duplicates** removed via md5sum before conversion
(the `(1)`/`(2)` browser re-downloads). `The_Algorithmic_Age_White_Paper (1).docx`
was *not* identical and was kept for review.
- **2,055 headings recovered.** Most Word sources used bold-styled paragraphs
instead of Heading styles, so pandoc emitted zero `#` marks —
`Solar_Energy_Masterclass` had 56 bold pseudo-headings and no real structure.
Recovery script promotes standalone `**bold**` lines to `#`/`##`/`###` using
MODULE/PART/SECTION patterns, caps, and length heuristics. Table rows and
emphasis inside prose are skipped. **Spot-check the long files** — heuristics
are good, not perfect.
- Not converted: `AI_Datacenters_Masterclass.pptx`, `AI_Tools_Masterclass_2025.pptx`,
`New World Curriculum Deck.pdf`, 2 curriculum-planning `.xlsx`, `286.jpg`.
The two xlsx are planning artefacts, not publishable content. The PDF deck and
pptx files may be worth converting separately if you want them as downloads.