lanthanides.io — Site Audit & Investment-Readiness Gaps

lanthanides.io — Site Audit & Investment-Readiness Gaps

Audit date: 2026-05-31 · Branch: main · Audited against current main (HEAD 2a0a1b9; a prior pass mis-stated its base as 56e980f, which predated /framework/) Scope: the current static Jekyll repo (no Next.js yet). Read-only inventory. This document is the reference that later migration prompts read. It is factual and cites file paths. Where the older docs/improvement-plan.md (dated 2026-04-08) disagrees with the repo, the repo wins — that plan is stale (it references supply-risk.html, heatmap.html, search.js, heatmap.js, _sass/_heatmap.scss, alternative_supply.yml, header.html, all of which no longer exist).

Re-evaluation note (2026-05-31): a prior pass of this audit was written against the stale base commit 56e980f, which predates d55ad05 (“feat: add /framework/ — REE import/export operational reference”) and 0628880/44fdc1d (the regulatory-monitor automation). That pass omitted (a) the served /framework/ page from the inventory (§1.4), the permalink contract (§2), the navigation (§1.3), and the crown-jewel discussion (§6); and (b) the regulatory-monitor pipeline (.github/workflows/regulatory-monitor.yml

• scripts/scraper|triage|notify) that keeps the regulatory data fresh. This revision adds both and corrects the nav link count (8 → 9). Every other count and claim below was re-verified against the working tree and stands unchanged.

1. Site map & inventory

1.1 Configuration & infrastructure

1.2 Layouts (_layouts/)

1.3 Includes (_includes/)

1.4 Pages (pages/ + index.html)

1.5 Collections

1.6 Data files (_data/)

1.7 JavaScript (assets/js/)

1.8 Styles (_sass/) + media

• 24 SCSS partials imported by assets/css/main.scss (_variables, _reset, _mixins, _typography, _layout, _header, _footer, _home, _cards, _badges, _tables, _panels, _dashboard, _movements, _element-detail, _elements-page, _news-page, _article, _about, _interactive, _regulatory-banner, _periodic-nav, _responsive, _print).
• assets/images/: favicons (ico/16/32/svg/apple-touch/android-chrome 192+512), logos (48/128/png), og-default.png (exists), article images china-rare-earth-controls-{600,1200}.jpg, site.webmanifest.
• assets/data/: elements.json (Liquid-generated from _data/elements/), fluctuations.json (a duplicate of _data/fluctuations.json).

1.9 Out-of-build tooling (not served)

scripts/*.py (validate, normalize, generate, import invoices/offers, detect movements, build snapshots, fluctuations, source breakdown, placeholder generator), outreach/*.py (consent-gated supplier intake), invoices/, logs/, prompts/. Excluded via _config.yml/.gitignore.

Regulatory-monitor pipeline (out-of-build, but a real asset — see §5/§6): scripts/run_monitor.py + scripts/run_state.json (orchestrator + dedup/seen state), scripts/scraper/monitor.py + scripts/scraper/sources.yaml (polls MOFCOM homepage, gov.cn, and RSS/Atom feeds for REE-relevant announcements), scripts/triage.py (rule-based 1–10 significance scorer, biased low to avoid false-positive alerts), scripts/notify/{telegram,email}.py (alert channels). Driven by .github/workflows/regulatory-monitor.yml; its automated commits are authored by the regulatory-monitor bot.

2. Routing / permalinks (MUST be preserved by the migration)

Every public URL the current site serves. This is the contract the Next.js migration must honor (redirects if any change).

Top-level pages

• / — home
• /dashboard/
• /prices/
• /elements/
• /regulatory/
• /framework/ — served page (pages/framework.md); was missing from the prior pass. Part of the contract.
• /movements/
• /news/
• /methodology/
• /about/
• /sources/
• /404.html

Element collection (31) — /elements/<Symbol>/ with original-case symbol: /elements/La/, /elements/Ce/, /elements/Pr/, /elements/Nd/, /elements/Sm/, /elements/Y/, /elements/Sc/, /elements/Eu/, /elements/Gd/, /elements/Tb/, /elements/Dy/, /elements/Ho/, /elements/Er/, /elements/Tm/, /elements/Yb/, /elements/Lu/, /elements/Te/, /elements/V/, /elements/Sb/, /elements/W/, /elements/Bi/, /elements/Mo/, /elements/Zr/, /elements/Ta/, /elements/Nb/, /elements/Co/, /elements/Li/, /elements/Ga/, /elements/Ge/, /elements/Se/, /elements/In/

Article collection (5) — /news/<slug>/:

• /news/chinas-invisible-rare-earth-wall/
• /news/chinese-export-licence-timelines/
• /news/ebay-rare-earth-market-research-april-2026/
• /news/mofcom-october-2025-suspension/
• /news/retail-premium-puzzle/

Machine-readable / feeds / SEO

• /sitemap.xml (jekyll-sitemap)
• /feed.xml (jekyll-feed; <link type="application/atom+xml" rel="alternate" href="https://www.lanthanides.io/feed.xml" title="lanthanides.io — Rare Earth Prices, Export Controls &amp; Strategic Materials Intelligence" /> in head.html)
• /movements.xml (custom Atom feed, permalink: /movements.xml)
• /assets/data/elements.json (open-data export, permalink: /assets/data/elements.json)
• /assets/data/fluctuations.json (open-data export)
• /robots.txt
• /humans.txt
• /assets/images/site.webmanifest

⚠️ Migration note: /elements/<Sym>/ uses case-sensitive symbols. Several pages also deep-link /methodology/#display-price, /methodology/#provenance-chain, /methodology/#data-sources-breakdown, /methodology/#oxide-to-metal — preserve those anchors. The /framework/ page also exposes in-page anchors (#pricing, #us-side-tariff-stack-may-14-2026) for its own internal navigation — preserve them too.

3. Visualization audit

Context that colors every “data sufficiency” cell below: the price corpus is essentially two collection days — 2026-03-15 and 2026-04-04 (with a sprinkle of bulk on 04-04). fluctuations.json confirms it: nearly every window carries confidence: "low" and distinct_days_in_window: 2, and several elements have distinct_days: 1 with all windows null. The codebase itself says so (pages/dashboard.html:209 “Most current windows span only two distinct observation days, so confidence is generally low”). Any widget that draws a line, trend, or “% move” from ≤2 points is structurally choppy/misleading and is flagged accordingly.

4. Student-project signals (and where they get fixed)

Concrete, fixable tells that make a strong dataset read as a class project rather than a fundable venture.

4.1 .edu contact address (credibility)

mironovb@berea.edu appears in 3 public places: pages/about.html:79, CONTRIBUTING.md:220, humans.txt. A .edu address signals “student, not company.” Fix: replace with a domain address (e.g. data@lanthanides.io / hello@lanthanides.io). → content/brand phase.

4.2 Stray scratch files at repo root (professionalism)

Untracked junk in the working tree: chat.md, combined.txt, combined 2.txt, combined 3.txt, .arun/. (logs/, prompts/, run-all.sh are correctly gitignored.) Also committed bytecode dirs outreach/__pycache__/, scripts/__pycache__/ despite __pycache__/ being in .gitignore — they were committed before the ignore rule. Fix: delete the scratch files; git rm -r --cached the __pycache__ dirs. → repo-hygiene / migration setup phase.

4.3 Two parallel data stores with diverging numbers (data integrity)

• Prices: price_records.json (238) vs price_history/*.yml (285 observations). The “238 price records” stat (home/about) contradicts methodology’s “285 observations.”
• Elements: element_catalog.yml (canonical, drives all server pages) vs _data/elements/*.yml (drives only elements.json → ledger.js). With JS on, /prices/ can show different prices than with JS off. Fix: collapse to a single canonical store in the migration; derive everything else from it; label every count. → data-model phase (high priority).

4.4 Broken / placeholder data shipping (data integrity)

_data/elements/La.yml is malformed placeholder: literal schema strings source: "Description of source(s)", applications_summary: "Brief description of primary uses.", broken indentation, no price_per_kg. It flows into assets/data/elements.json and the interactive Ledger. A demo dir (_data/demo/) and a generate_placeholder_data.py script also sit in the repo (demo is build-excluded, but its presence is a tell). Fix: regenerate/repair La.yml from canonical data; remove _data/demo/ and the placeholder generator from the shipped tree. → data-model phase.

4.5 Apologetic “sparse data” framing (positioning)

The site repeatedly apologizes for thin coverage: methodology.md:406 “## Why Some Pages Show Sparse Data,” dashboard copy “We don’t hide thin coverage,” “Insufficient data yet,” “confidence is generally low.” Honesty is good; apology is not — it reads as a student hedging. Fix: reframe transparency as a feature (“provenance-first; we show our work”) and lead with the regulatory intelligence that is deep, rather than foregrounding what’s missing. The existing /framework/ page already models this confident, declarative, decision-oriented voice — use it as the in-repo tone reference. → content/positioning phase.

4.6 No clear value proposition / “who is this for” above the fold (positioning)

The home hero (“Rare earth & critical metal prices, with provenance.”) states what but not for whom or why now. .impeccable.md names the audience (investors, journalists, analysts) but the page never does. There is no above-the-fold hook tying the (world-class) export-control tracking to a decision a reader is trying to make. Fix: add a one-line audience + value statement and route the eye to the regulatory tracker / movements. → home redesign phase.

4.7 Design diverges from its own brief (polish)

.impeccable.md commits to IBM Plex Sans/Mono + Source Serif 4 (“serif headings… load-bearing,” “monospace for all numeric data”). The shipped site loads Inter + JetBrains Mono and no serif at all (head.html:13). The “Foreign Affairs / Goldman note” editorial identity is therefore not actually implemented. Fix: decide the type system deliberately in the migration and apply it consistently. → design-system phase.

4.8 site.webmanifest is broken + off-brand (polish/bug)

assets/images/site.webmanifest points icons at /periodicpricing/assets/images/… — a stale GitHub-Pages project base path that 404s on the root domain (www.lanthanides.io). Its theme_color: #1a1d23 and background_color: #ffffff also contradict the brand tokens used elsewhere (theme-color #1A5C6B in head.html; paper #F5F2EC in the brief). Fix: correct icon paths to /assets/images/… and align colors. → polish / PWA phase.

4.9 Thin-data charts implying trends (covered in §3)

Items #1, #2, #3, #4, #5 above. The migration should treat “≤2 distinct days ⇒ no line/no %move” as a hard rendering rule. → visualization phase.

4.10 Orphaned data + dead includes (codebase smell)

_data/news.yml, _data/key_developments.yml, _data/snapshots/, and _includes/fluctuation-fallback.html are referenced by no template. They suggest abandoned iterations. Fix: delete or re-wire during migration. → cleanup phase.

4.11 Minor: contributor dropdown lists untracked elements

price-update.yml offers Pm, Hf, Re (not in the 31-element catalog). Fix: trim to tracked elements (or commit to adding them). → low priority.

5. Assets to preserve (do not lose in the migration)

These are real, hard-won, and differentiating. Carry every one forward.

SEO & structured data

• The four _includes/structured-data-*.html (WebSite + FAQPage, Product+Offer, Article, BreadcrumbList) — comprehensive JSON-LD with live counts.
• jekyll-seo-tag-style meta (title/desc/keywords/canonical/OG/Twitter) in head.html, including the per-layout OG title logic.
• /sitemap.xml (jekyll-sitemap) and /feed.xml (jekyll-feed).
• The custom /movements.xml Atom feed (movements.xml) — hand-rolled, valid, category-tagged.
• robots.txt, humans.txt, favicons + site.webmanifest (fix paths per §4.8), og-default.png, CNAME.

Open data

• assets/data/elements.json and assets/data/fluctuations.json (sitemap: false, stable permalinks) — the public open-data export. Preserve the URLs; fix the underlying store (§4.3/4.4).

Licensing

• Dual CC BY 4.0 (content/data) + MIT (code), surfaced in LICENSE, footer.html, head.html rel="license", and the Atom <rights>.

Contributor pipeline

• .github/ISSUE_TEMPLATE/{price-update,data-correction,market-note}.yml + PULL_REQUEST_TEMPLATE.md.
• .github/workflows/community-intake.yml — the manual-dispatch, approved-label-gated, PR-opening “two human checkpoints” flow (and the matching outreach/community_intake.py). This consent-/review-gated provenance model is itself a credibility asset and is documented on /methodology/.

Regulatory-monitoring automation (the freshness engine behind the crown jewel)

• .github/workflows/regulatory-monitor.yml + scripts/scraper/ (monitor.py, sources.yaml — Chinese-gov + RSS pollers), scripts/triage.py (rule-based significance scorer), scripts/notify/{telegram,email}.py, scripts/run_monitor.py + run_state.json (dedup state). Runs every 6h and alerts only on critical announcements. This is what makes “updated as announcements land” literally true; carry the pipeline (and its bot-commit cadence) into the migration’s ops story even though scripts/ itself isn’t served.

Content (the substance)

• All 31 element pages (_elements/*.md) — sourced applications, market/supply stats, notes, properties, references.
• All 5 articles (_articles/*.md) — footnoted, publication-quality.
• The /framework/ operational reference (pages/framework.md) — ~450 lines of sourced import/export decision support (regulatory tiering, US tariff stack, realised-price model, landed-cost decomposition, channels, 2026 trigger calendar). A near-unique companion to the regulatory tracker; preserve verbatim and keep the /framework/ URL.
• element_catalog.yml, price_records.json, policy_events.yml, regulatory/*.yml, source_registry.yml, site_settings.yml, methodology.md — the data and the rules that govern it.
• The two-price (Retail Reference / Bulk Benchmark) model and the price-selection.html logic.

6. Crown jewel — the MOFCOM/GAC regulatory tracker

The strongest, least-replicable asset on the site is the Chinese export-control tracker (pages/regulatory.html, _data/regulatory/*.yml, _data/policy_events.yml) — now a cluster with its operational companion /framework/ and an automated monitoring pipeline (fourth bullet below) — and the migration should make it the centerpiece, not a tab buried mid-nav.

Why it’s the differentiator:

• Depth and primary sourcing. _data/regulatory/mofcom_18_2025.yml (and siblings) carry the announcement number, the Chinese reference (商务部 海关总署公告2025年第18号), issuing authority, issue/effective dates, exact affected elements and controlled forms, the legal basis (Article 15, Export Control Law of the PRC; 2024 Dual-Use Items Regulations), the 45-working-day review period, and suspension state (e.g. “No. 18 was never suspended; Oct-2025 Nos. 55–62 suspended until 28 Nov 2026”). policy_events.yml strings 11 events from the July-2023 Ga/Ge controls through the Feb-2026 Japan sanctions into one coherent timeline.
• It answers the question the audience actually has. The brief’s primary user (investor/journalist/analyst) needs “is this element controlled, by which announcement, since when, and is it suspended?” — and almost no English-language source assembles this with citations. The element pages already reflect it inline (Dy.md banner cites No. 18/2025 + the 28-Nov-2026 suspension key date), and the dashboard/regulatory snapshot quantify it.
• It’s defensible. Prices are thin and will always be contestable; the regulatory corpus is deep, dated, sourced, and updated as announcements land. It is the thing a funder can verify in five minutes and not find elsewhere.
• It now ships with two force-multipliers the prior pass missed. (1) The /framework/ operational reference (pages/framework.md) converts the tracker’s what/when into procurement how — three-tier classification (uncontrolled / general-licence / case-by-case), the live post-IEEPA US tariff stack, a multiplicative realised-price model, four-axis landed-cost decomposition with worked examples, and a 2026 decision-trigger calendar — all reconciled against the same _data/regulatory/*.yml corpus, in ~450 lines of confident, sourced, decision-oriented prose (the clearest in-repo counter-example to the apologetic voice of §4.5). (2) An automated monitoring pipeline (.github/workflows/regulatory-monitor.yml → scripts/scraper/ → scripts/triage.py → scripts/notify/) polls Chinese-government sources every six hours, scores each item’s regulatory significance, and fires Telegram/email alerts only on critical announcements — the machinery that makes “updated as announcements land” literally true. Treat tracker + framework + monitor as one crown-jewel cluster; nav already keeps the first two adjacent (Regulatory → Framework).

Implication for the migration: elevate the tracker to a hero position (home above-the-fold + a first-class destination), keep its data model intact, and let the (transparent) thin price data play a supporting role behind the regulatory intelligence rather than leading with apologies about sparse coverage (§4.5).

End of audit. Later prompts: treat §2 (permalinks) as a hard contract, §3 decisions as the visualization backlog, §4 as the investment-readiness punch-list, §5 as the do-not-break list, and §6 as the product’s north star.