Every data correction Afrimintel has made since April 2026 · Last updated: 22 April 2026 · Editorial responsibility: Nikesh Patel, Honorary Consul of Rwanda in Mauritius
Afrimintel publishes every data correction it has made. Every source we have replaced. Every assumption we have retracted. Every error caught by us or flagged by a reader. We do this because most commercial mineral intelligence platforms do not, and the distinction matters to anyone making capital decisions on African minerals.
Filter by type, scope, or date. Every entry names the original claim, the correction, and the source that drove the change. If you spot something we have not caught, email nikesh@afrimintel.com.
This is not an apology page. Corrections are not weaknesses — they are evidence that the platform is operating on primary-source discipline rather than aggregated secondary sources. A platform that never corrects is not a platform that is verifying.
Earlier corrections are reflected in the current data but are not individually logged here. The log begins from the point at which this discipline became continuous.
Filter
Loading…
No matching entries. Try broadening the filter or resetting.
What counts as a logged correction
Entries fall into five classes. A correction is logged when any of the following is true:
Data — a specific factual claim (grade, tonnage, discovery date, CEO name, operator, resource figure) was changed.
Integrity — content caught as fabricated or unsourced was removed or replaced with a sourced version.
Methodology — the way a score, tier, or metric is constructed was changed in a way that affects published numbers.
UX / layer — a user-facing surface was fixed where it had been silently misrepresenting data (e.g. stale counts, rendering bugs hiding content).
Version — a platform or dataset version label was corrected to match the current state.
Release-level technical refactors, cosmetic styling work, and new feature additions are NOT logged here — they appear in the release CHANGELOG files. This page is specifically for changes where a reader would be entitled to know the claim has moved.
How to contribute a correction
If you read any Afrimintel record or output and believe it is wrong, send the specifics and a citable source to nikesh@afrimintel.com. Corrections from external readers are acknowledged in the Source field of the relevant entry. The platform's data discipline gets better when people who know more than we do tell us.
snippet added to all 6 missing pages, placed after matching the existing pattern. Coverage now 11 of 11 HTML pages. Pageview tracking will fire on every public surface. Methodology Record tier definitions section rewritten to surface the three-tier framework empirically: (a) explicit count line stating 18 IG + 6 RG + 88 SR + 112 total with grade_class field reference; (b) new RESEARCH-GRADE section documenting criterion (named operator + named primary-source citation, but no published reserve/resource statement against an internationally recognised reporting code), intended use (orientation only, not citable for capital allocation), and the six current Rwanda assets — Rutongo, Nyakabingo, Gatumba, Bugarama-Bukunzi, Musanze, NW Rwanda Gold Belt — sourced to RMB and US Dept of Commerce Rwanda Mining Guide; (c) upgrade path stated; (d) stale 'sub-tier classification in progress / scheduled for future release' paragraph removed. Plausible custom event taxonomy (dossier-view, tool-started/completed, document-download, conversion events per the install spec) remains a Phase B follow-up — basic pageview tracking is sufficient for closing the coverage gap in this deploy.",
source: "Internal scan during platform-items review post-v1.0.39 deploy. Plausible coverage gap surfaced by grep audit of presence across HTML pages. Tier section drift surfaced by comparison of methodology page text against the actual grade_class values in js/data.js — the page was documenting a state superseded by the v1.0.18 schema migration that gave records machine-readable tier fields. Both corrections are platform-discipline fixes (no editorial decision required); the published tier criteria are derived empirically from what the data already contains."
},
{
date: "2026-05-04",
title: "v1.0.39 — DCF Test Battery v1.0 spec published at /methodology/dcf-test-battery; platform forward-link activated",
type: "methodology",
scope: "/methodology/dcf-test-battery.html (new); /methodology/index.html; platform.html DCF pane",
was: "DCF Test Battery v1.0 spec existed as an internal artifact and was promised in the v1.0.38 audit log entry to publish at /methodology/ as part of the W3 deploy. The Known Limitations panel added in v1.0.38 forward-linked to /methodology/ generically because the dedicated page did not yet exist on the platform — a sophisticated reader following the link would land on the methodology index, not on the battery itself. The methodology index page made no mention of tool reliability testing and did not link to the battery.",
now: "New page published at /methodology/dcf-test-battery.html rendering the v1.0 spec — 15 tests across 5 categories (monotonicity, boundary, sensitivity, reconciliation, pre-specified failure mode), Kamoa-Kakula reference asset with five published reconciliation targets ($5.5bn through $20.2bn NPV), Day 30 tool limitations explicitly documented, change log including the drafted v1.1 amendment under editorial review and the scheduled Q3 v2.0 expansion under named external reviewer. Page matches the dark editorial aesthetic of the methodology index. Platform.html Known Limitations panel updated — forward-link now points at /methodology/dcf-test-battery.html directly. Methodology index updated with new 'Tool reliability testing' section between Audit source protocol and Commodity price methodology, providing context and routing to the battery. Cycle 1 first-cycle results companion document publishes at a companion URL on editorial sign-off; the v1.1 spec amendment integrates into this page on the same trigger. Note: the test count is reported as 15 on this page (verified count of test specifications C1.1 through C5.2). The v1.0.38 audit log entry stated 13 tests — that figure was inherited from the v1.0 spec's own change log section, which is at variance with the spec's actual test list. The 15 figure is authoritative.",
source: "v1.0.38 audit log entry promised W3 publication; this is that publication. Spec content is unchanged from the v1.0 internal artifact approved April 2026 — no fabrication, no new claims. Editorial responsibility unchanged: Nikesh Patel."
},
{
date: "2026-05-04",
title: "v1.0.38 — DCF tool finalisation: input unit disambiguation + Known Limitations panel published",
type: "methodology",
scope: "DCF / NPV Model pane — platform.html",
was: "DCF tool form labels did not disambiguate contained-vs-recovered metal: 'Annual Production (kt/yr or koz/yr)', 'Commodity Price ($/t or $/oz)', 'OPEX ($/t produced)'. The math (calcDCF in js/app.js) requires production = contained metal (= ore × grade) and OPEX per contained tonne — but a user feeding published Kamoa-Kakula values directly (recovered metal output, $/lb cash cost) computed an NPV $1.74bn away from the math-correct interpretation at the same Kakula DFS inputs ($9.81bn vs $8.06bn). Tool also offered no published documentation of its five structural omissions: no production ramp-up, no sustaining capex, no depreciation tax shield, no by-product credits, no multi-phase capex / state participation / working capital / stripping ratio variance. Cycle 1 reconciliation against published Kakula DFS NPV target of $5.5bn produced +78.3% over-credit at literal first-five-years cash cost input — diagnostic, but invisible to the user before running the tool.",
now: "Three input labels disambiguated: 'Annual Production — contained metal (kt/yr or koz/yr)', 'Commodity Price ($/t or $/oz, real)', 'OPEX ($/t of contained metal)'. Each input now carries help text below clarifying conversion from common alternative units ($/lb-recovered = ×2204.62 ×recovery%; $/t-recovered = ×recovery%; C1 cash cost = + treatment/refining/transport). Closed-by-default 'Known limitations & structural omissions' panel placed above the form, documenting the five categories of structural omission with their practical effect on NPV. Forward-link to /roadmap.html for Q3 2026 expansion under independent reviewer audit. No DCF math change — calcDCF and runDCF unchanged; the labels are now in agreement with the existing post-results methodology note (which had always confirmed the contained-metal convention). CANONICAL_VERSION bumped from v1.0.37 to v1.0.38 across all 7 references in pre-deploy-audit.js, platform.html, about.html, roadmap.html, and js/app.js.",
source: "DCF Test Battery v1.0 first-cycle results v2 (4 May 2026, dual-column reference + platform run against published Kamoa-Kakula targets). Reference Python DCF over-credited NPV by +35.5%; platform DCF over-credited by +78.3% at literal cash-cost input. Implied reconciling LoM cost: $1.06/lb-recovered on reference, $1.51/lb-recovered on platform — the 45c spread reflects the platform's structural omissions. Engineering recommendations ENG-1 (input unit disambiguation) and ENG-5 (documented limitations) shipped as immediate copy-only patch; ENG-2/3/4 (math-side fixes for ramp-up, sustaining capex, depreciation shield) deferred to Q3 paid-independent-reviewer scope. Editorial decision: ship documented-limitations rather than rapid math changes — preserves acquirer-perceived position as 'tool that documents its boundaries and is scheduled to extend them under independent review' rather than 'silent failure' or 'rushed engineering before sprint close'."
},
{
date: "2026-05-03",
title: "v1.0.21 — shared injection helper introduced; production code and audit infrastructure unified to single source of truth",
type: "integrity",
scope: "AI provenance contract — js/ai.js, tests/audit/provenance-contract-battery.js, js/ai-injection-helper.js",
was: "Production AI code (ai.js) and the audit test battery maintained parallel system-prompt construction logic. When the v1.0.18 schema migration updated production but not the test battery, the battery silently sent stripped data to the model — producing a false 80% compliance reading. Drift between audit infrastructure and production AI was structurally possible.",
now: "Single shared module js/ai-injection-helper.js exports buildSystemPromptLines(). Both production code and test battery require the helper and call the same function. Schema-version-stamped (v1.0.18). New pre-deploy pipeline check (Check 18) verifies helper exists, exports the canonical functions, is consumed by both surfaces, and loads in the correct script order in platform.html. Drift bug class is now structurally impossible — schema migrations touch one function, both surfaces inherit. Tag-detection regex also fixed in helper to handle both bare [VERIFIED] and dated [VERIFIED 2026-05-02] forms; previously all 18 v1.0.18 IG records were injecting as [SOURCED] instead of [VERIFIED] due to date-suffix mismatch.",
source: "Internal architectural review prompted by false-80% test reading on first v1.0.18 schema run. Refactor pattern: extract duplicated logic to a UMD module loadable from both browser and Node contexts."
},
{
date: "2026-05-03",
title: "Methodology page Table 4 reconciliation published live (v1.0.20)",
type: "methodology",
scope: "/methodology/ index.html",
was: "Methodology page summary-only — formula stated, Opportunity Multiplier range stated, but per-province MSP/IC/Opp/Score breakdown not displayed. Table 4 only available in the downloadable PDF, which itself had previously drifted from data.js values until the v1.0.18 fix.",
now: "Full Table 4 displayed inline on /methodology/ with all 13 provinces, MSP, IC, Opp, Score, and IC Source columns. Reconciliation discipline statement added explicitly: all 13 province scores reconcile to the formula (MSP × 0.6 + IC × 0.4) × Opp within ±0.05 rounding tolerance, enforced by pre-deploy reconciliation script. Fraser Institute vintage transparency note added (Botswana/Zambia 2025; others 2024 with INFERRED tags where applicable). Methodology page version bumped to v1.1.",
source: "Editorial decision to give institutional reviewers (BGS, DFI partners) the full reconciliation visible without requiring PDF download. The decision arose from preparing the BGS Director outreach — a national-survey-grade reviewer should see methodology rigour on first click, not after a download step."
},
{
date: "2026-05-02",
title: "v1.0.18 schema rebuild — IG-18 dossier finance-grade re-audit cycle complete",
type: "data",
scope: "AFRICA_DEPOSITS — all 18 INTELLIGENCE-GRADE deposit dossiers",
was: "IG-18 dossier pool inherited from earlier development cycles. A spot-check audit identified material drift across most records: 2 SEVERE multi-dimensional errors (Manono operator wildly outdated; Colluli operator three years stale), 4 SEVERE single-dimensional errors (Sentinel tonnage 2.4× actual; South Deep order-of-magnitude tonnage error; Venetia grade ~half actual; Motheo operator misattributed to MMG when Sandfire still owns), 6 records with material drift or missing operator structure (Mogalakwena, Sukari, Khoemacau, Tenke Fungurume, Kansanshi, Obuasi), 2 records with correct numbers but missing operational context (Kamoa-Kakula seismic event; Kipushi production restart), and 1 reserve-category mislabel (Balama).",
now: "All 18 IG records rewritten with new finance-grade schema. New schema fields: separate res_mt_reserve / grade_reserve / res_type_reserve and res_mt_resource / grade_resource / res_type_resource (replacing single triplet); proved_mt / probable_mt / measured_mt / indicated_mt / inferred_mt sub-categories; reserve_basis / resource_basis (100% project basis vs attributable disclosure); operator_interest (source-qualified ownership detail); dispute_status enum for contested assets; transparency_flag for opaque or recovery-stage operators; recent_events; last_reviewed (mandatory, pre-deploy gate blocks IG records older than 180 days). Manono Lithium District restructured as multi-claim contested asset with parent record + 2 sub-records (Manono Lithium SAS NE under construction; Dathcom-AVZ South in arbitration). 13 of 18 dossiers required material correction; all corrections individually documented in companion files in /audit-log/ directory.",
source: "Editorial verification cycle 2 May 2026, conducted dossier-by-dossier against most-recent primary sources (NI 43-101 / JORC / S-K 1300 technical reports, operator annual reports, regulatory filings). Tranche files: 2026-05-02-ig-dossier-spotcheck.md, -tranche2.md, -tranche3-FINAL.md."
},
{
date: "2026-05-02",
title: "Pre-deploy pipeline expanded to 18 checks; new v1.0.18 schema-aware checks block deploy on basis disclosure / staleness / dispute representation gaps",
type: "integrity",
scope: "scripts/audit/pre-deploy-audit.js",
was: "Pre-deploy pipeline ran 14 checks at v1.0.12 — covered permanent exclusions, hardcoded secrets, wildcard CORS, counterparty hallucinations, methodology reconciliation, JS syntax, HTML structure. Did not enforce v1.0.18 schema discipline (basis disclosure, refresh cadence, dispute representation).",
now: "Pipeline expanded to 18 checks. New checks 15-18: IG dossier staleness (flags any IG record without last_reviewed or older than 180 days); Reserve/Resource basis disclosure (flags IG records with reserve or resource numbers but no basis field); Disputed asset surface (flags records with dispute language in operator field but no dispute_status flag); Shared injection helper present and consumed by both production and audit infrastructure (added 3 May at v1.0.21). Round-number drift check updated to use new IG criterion. Bundle now passes 0 CRITICAL, 0 SIGNIFICANT.",
source: "Pipeline architecture extended in lockstep with v1.0.18 schema migration. Each new check addresses a discipline gap identified in the 2 May audit cycle."
},
{
date: "2026-05-02",
title: "Counterparty hallucination scrub — false BGS partnership claim removed (v1.0.10)",
type: "integrity",
scope: "Platform copy — js/app.js partnership opportunity tag, platform.html BGS reference",
was: "js/app.js contained 'BGS Africa programme welcomes Afrimintel as a data dissemination partner — dataset reach in return for attribution. Nikesh: Honorary Consul of Rwanda is direct entry point for BGS Africa partnership.' platform.html contained 'BGS Africa — will replace editorial coverage estimates with validated geological survey data for 8+ provinces.' Both claims were fabrications — no such partnership existed; BGS outreach was at LinkedIn-DM stage with no response. Both surfaces violated the Counterparty Extension to the Hallucination Guard.",
now: "Both surfaces rewritten with strict status language matching the Outreach Tracker. Current platform copy: 'Afrimintel has identified the BGS Africa Programme as a primary data partnership target. Status: outreach in dialogue (May 2026) — initial response received from BGS leadership; email follow-up underway. No partnership confirmed.' The Honorary-Consul-as-pathway claim was also scrubbed — title is verifiable institutional standing, but no documented diplomatic introduction has occurred.",
source: "Counterparty Extension to Hallucination Guard, audited 2 May 2026. Outreach Tracker v1.2 records actual BGS engagement state."
},
{
date: "2026-05-02",
title: "Comparable transactions count corrected from 30 to 28 (round-number drift)",
type: "data",
scope: "Marketing copy — platform.html sample dossier preview",
was: "Platform copy claimed '30 verified African mineral transactions 2016-2025'. Pre-deploy pipeline run after introduction of round-number-drift check found COMPARABLE_TRANSACTIONS array actually contains 28 records, not 30.",
now: "Marketing copy updated to '28 African mineral transactions 2016-2025'. The word 'verified' also dropped — per-record source verification is W2 work and not yet complete; restoring 'verified' requires editorial sign-off on each transaction's primary-source citation.",
source: "Pre-deploy pipeline Check 10 (round-number drift). Internal audit caught the discrepancy on first pipeline run after the check was added."
},
{
date: "2026-05-02",
title: "AI provenance contract test battery shipped + 95% compliance baseline established (v1.0.13)",
type: "methodology",
scope: "tests/audit/provenance-contract-battery.js",
was: "v1.0.11 release strengthened the aiAskStrata system prompt with a 7-clause Provenance Contract instructing the model on no-fabrication, refuse-when-absent, tier-tagging discipline. Contract was shipped untested — claims that the contract was 'enforced' rested on architecture, not measurement.",
now: "20-query test battery shipped: 5 IN-DB (facts present in database), 5 ABSENT (deposits/figures NOT in database — model must refuse), 5 ADVERSARIAL (questions designed to provoke fabrication of counterparties or fictitious data), 5 TIER-MIX (questions requiring distinguishing INTELLIGENCE-GRADE vs SPATIAL-REFERENCE). Initial run showed 70% compliance; investigation found 6 false-positive pass-criteria bugs in the battery itself. Battery v1.1 fixed: comma-tolerant number matching; fail criteria scoped to non-refusal text only; fabrication patterns require proximity to queried fictitious deposit name. Re-run produced 95% compliance — verdict: contract may be marketed as 'enforces'.",
source: "Build-out per v1.1.0 sprint plan finding: 'AI provenance contract has never been measured.' Test battery converts contract from instruction to measured compliance."
},
{
date: "2026-05-02",
title: "Runtime self-audit on AI outputs — defence-in-depth layer added (v1.0.12)",
type: "integrity",
scope: "js/audit-runtime.js",
was: "v1.0.11 Provenance Contract was instruction-only — the model received the contract in the system prompt but no runtime enforcement existed if the model produced a response that violated it. A 95% compliance rate means 5% of responses could fabricate without detection at runtime.",
now: "Runtime self-audit module added. Runs after every aiAskStrata response, performs lightweight client-side compliance checks against the v1.0.11 Provenance Contract, surfaces inline audit card with response. Three card states: green (no automated flags), amber (warning — e.g. numeric claims without provenance tier tag), red (critical — counterparty hallucination, forward-looking partnership claim, or permanent-exclusion term detected). Defence-in-depth alongside the system-prompt contract: contract instructs the model, runtime audit catches violations the contract failed to prevent.",
source: "v1.0.12 release notes. Defence-in-depth pattern recommended after the 95% compliance reading established that 5% residual fabrication risk needed runtime mitigation."
},
{
date: "2026-05-02",
title: "Site security hardening — wildcard CORS removed from 5 proxy functions, hardcoded beta codes removed from netlify.toml, ai-proxy origin gate fixed",
type: "integrity",
scope: "netlify/functions/*.js, netlify.toml",
was: "ai-proxy.js (consuming Anthropic API key) had no origin gate and wildcard Access-Control-Allow-Origin: '*' — open relay to Anthropic's API for any internet caller. validate-code.js used startsWith() for origin matching, vulnerable to prefix-bypass (e.g. 'localhost.evil.com' would match 'http://localhost'). netlify.toml comments contained the actual beta access codes (STRATA2026, BETA01-05) committed to source. news.js, afdb-proxy.js, poi-query.js, ungm-proxy.js — all wildcard CORS, all open relays to Afrimintel's compute budget.",
now: "ai-proxy.js: exact-match origin allowlist via URL.origin parsing (no prefix bypass), CORS restricted to afrimintel.com / www.afrimintel.com, sanitised upstream error echo (no information disclosure). validate-code.js: same origin-gate fix; localhost / 127.0.0.1 removed from production allowlist; duplicate origin entry removed. netlify.toml: hardcoded codes removed, replaced with reference to Netlify dashboard env var. 5 proxy functions (news, afdb-proxy, poi-query, ungm-proxy, commodities): wildcard CORS replaced with afrimintel.com origin restriction. news.js: ReDoS hardening — input bounded to 500KB, regex bounded with {0,5000}, iteration count bounded to 50 max, 8s fetch timeout.",
source: "Adversarial security audit prompted by the question 'why would Karen not partner with Afrimintel after seeing the platform?' — security-audit posture became one of the five hesitancies to address."
},
{
date: "2026-05-02",
title: "Operations Dashboard removed from public platform (admin-internal pane was localStorage-gated only)",
type: "integrity",
scope: "platform.html — Operations Dashboard pane and supporting code",
was: "Operations Dashboard pane (lines 2220-2325 in v1.0.8 platform.html) displayed bootstrap progress, subscriber concentration, revenue health — internal P&L scaffolding. Visibility was gated only by localStorage access code (NIKESH), not server-side. Any user inspecting JavaScript could trivially trigger the pane. Pane shipped to all users, only hidden by client-side check.",
now: "Pane removed entirely from platform.html. Associated nav-injection JS (checkAdminAccess function and pick() override that created the Ops nav button on admin code entry) deleted from app.js. Dead supporting code (~150 lines: GRO_RULES, STRATA_SUBSCRIBERS_KEY, getSubscribers, saveSubscribers, addSubscriber, renderOpsDashboard, renderGroRules, DISQUALIFIERS, renderDisqualifiers, toggleDisqualifier) deleted in follow-up cleanup pass. Pre-deploy pipeline now flags any reintroduction of p-ops-dashboard or related identifiers.",
source: "Sprint task 1.11 — Day 30 success criterion 3 (Operations Dashboard not publicly accessible)."
},
{
date: "2026-05-02",
title: "Partnership Outreach pane removed from public platform (internal sales-pipeline scaffolding was rendering to all government-role users)",
type: "integrity",
scope: "platform.html — Partnership Outreach pane",
was: "p-partnership pane exposed pre-drafted email templates and 'entry angle' notes for BGS, Fraser Institute, AfDB, USGS to any authenticated government-role user. Internal sales-pipeline scaffolding rendered to public surface. The DRAFT EMAIL framing protected the body language but not the surrounding strategy notes; email body texts described Afrimintel's user base in operational terms ('currently serves DFI investors, government geologists, junior mining companies') that implied commercial scale Afrimintel did not yet have. Per GRO failure pattern: 'optimising for investor narrative over user pain'.",
now: "Pane fully removed: p-partnership div deleted from platform.html, renderPartnership function deleted from app.js (40 lines), partnership removed from PANELS_CONFIG.government.panels in data.js, partnership label removed from NAV_LABELS.government, renderPartnership call removed from init sequence.",
source: "Audit Phase 5 finding 1 May 2026 — internal sales scaffolding exposed to public surface violated Counterparty Extension."
},
{
date: "2026-05-01",
title: "v1.0.9 rebrand close — STRATA → Afrimintel substitution across HTML/JS/CSS, 5 PDFs reissued, 42 internal markdowns deleted from deploy bundle",
type: "version",
scope: "Bundle-wide rebrand — platform.html, js/*.js, css/strata.css → afrimintel.css, all 5 customer-facing PDFs",
was: "Platform launched April 2026 under working name STRATA. Domain afrimintel.com registered 22 April 2026 with rebrand pass scheduled but not yet applied. Bundle contained STRATA references throughout HTML, JS internal identifiers, CSS file names. 42 internal working markdowns at bundle root were Netlify-served by default — anyone could read full strategic documents including hallucinated counterparty references.",
now: "5 customer-facing PDFs (Methodology, Category Paper, Acquirer Brief, API Reference, Competitor Gap) renamed and rebranded with cleaned content. 42 internal markdowns at root deleted from deploy bundle (legacy CHANGELOGs, D1-D5 deliverables, E1 Commercial Brief, SYNTHESIS reports, audit drafts) — these contained Thomas references, Asterimo references, Bibemi references, hallucinated counterparty claims. STRATA → Afrimintel substitution applied bundle-wide (case-sensitive, brand only — historical audit-log entries and internal localStorage keys preserved per documented policy). CSS files renamed in lockstep. _redirects file written with 6 301 redirects for legacy STRATA-branded URLs. methodology/soc2-roadmap.md deleted (contained Wood Mac / S&P / Glencore as enterprise prospects in operational-relationship language — Counterparty Extension violation).",
source: "Sprint W1 close (24 April → 1 May). Companion file: audit-log/v1.0.9-entries.md (15 entries detailing every change between v1.0.5 and v1.0.9)."
},
{
date: "2026-05-01",
title: "Five flagged editorial items resolved — Category Paper + Acquirer Brief content corrections",
type: "data",
scope: "afrimintel_category_paper.md, afrimintel_acquirer_brief.md",
was: "Category Paper opening section contained 6 unsourced numerical claims: '$2.3 trillion undiscovered endowment by 2040'; '54% land mass undercovered'; '20 years lag'; 'Kalahari Platform 52% / Saharan Metacraton <15% surface coverage'; section heading 'The problem in three numbers'. Acquirer Brief contained placeholder text '[SOURCE: USGS / IEA estimates — cite precisely before circulation]' and '[SOURCE: USGS Mineral Commodity Summaries 2024 — verify exact percentages before citation]'.",
now: "Category Paper substitutions: $2.3T → $8.6T (sourced to AFC Compendium of Africa's Strategic Minerals, February 2026, verified across 8+ secondary citations). 54% → reframed qualitatively as 'significantly under-explored' with World Bank 2025 + AFC 2026 citations. 20 years → reframed as 'decades behind' with World Bank 2025 citation. Kalahari/Saharan coverage figures softened to qualitative comparative claim. Section heading changed from 'three numbers' to 'three signals'. Acquirer Brief placeholders replaced with sourced figures: Sub-Saharan Africa holds approximately 30% (IMF SSA REO April 2024); USGS MCS 2025 figures for cobalt (DRC 76%), copper (DRC 3.3 Mt + Zambia 680 kt of 23.0 Mt world), graphite (Mozambique 75 kt + Madagascar 89 kt of 1.63 Mt world); IEA 2025 demand projections.",
source: "Sprint task 1.21. All sourced figures independently verified against primary sources via web search May 2026."
},
{
date: "2026-05-01",
title: "Methodology Table 4 reconciled to formula — Kalahari 9.2 → 9.6, Lufilian 8.2 → 8.4, OM range corrected to 1.05-1.30 across all surfaces",
type: "methodology",
scope: "js/data.js PROV array, methodology PDF Table 4, /methodology/ index.html, platform.html scoring tooltip, app.js Opp tooltip rendering",
was: "Methodology PDF Table 4 component values (MSP/IC/Opp) disagreed with data.js implementation values for ~12 of 13 provinces. Formula (MSP × 0.6 + IC × 0.4) × Opp did not compute the displayed scores from the displayed components for multiple rows: Saharan Metacraton showed (6.2 × 0.6 + 3.8 × 0.4) × 0.80 = 4.19, displayed 6.2. Two displayed scores carried stale-update bugs: Kalahari Platform 9.2 (computed 9.61); Lufilian Arc 8.2 (computed 8.37). Opportunity Multiplier range published 4 different ways across 4 surfaces: methodology PDF Table 1 said 0.80-1.25; platform.html tooltip said 0.8× – 1.3×; methodology HTML said 1.0-1.3; app.js rendering said 1.05-1.30; actual data.js values ranged 1.05-1.30.",
now: "Methodology PDF Table 4 component values rewritten to match data.js exactly. Kalahari score updated 9.2 → 9.6. Lufilian score updated 8.2 → 8.4. All 13 provinces now reconcile to the formula within ±0.05 rounding tolerance. Opportunity Multiplier range corrected to 1.05-1.30 across methodology PDF, methodology HTML, platform.html tooltip, and app.js renderings. Pre-deploy reconciliation script enforces the lock at every build.",
source: "Sprint task 1.20. v1.0.9 methodology PDF discrepancy was identified during a pre-deploy adversarial audit; the fix landed in v1.0.10 with build-time reconciliation gate."
},
{
date: "2026-05-01",
title: "Permanent-exclusion scrubs — Bibemi references removed from platform.html and js/data-t4.js",
type: "integrity",
scope: "platform.html line 1374 (Field Corrections Programme card), js/data-t4.js line 158 (orogenic Au signature record)",
was: "Two surfaces carried Bibemi references. platform.html: 'The Bibemi geophysical correction (April 2026) is an example of this process.' data-t4.js: '[Field correction: Bibemi latitude, NE Pan-African fabric, Cameroon — April 2026]'. Both violated permanent-exclusion list per project brief.",
now: "platform.html: 'Field corrections submitted by senior exploration geologists working on live African projects are examples of this process.' data-t4.js: bracket deleted entirely; geological signature description stands without source attribution. First-pass replacement on data-t4.js attempted 'submitted by senior exploration geologist on active project' which on review still functioned as oblique reference; second-pass deletion of the entire bracket adopted instead.",
source: "Pre-deploy pipeline Check 1 (permanent exclusions) blocks any reintroduction. Bundle-wide grep returns zero matches in deployed surfaces."
},
{
date: "2026-04-22",
title: "Rwanda integrity audit — Rutongo grade figure softened from [VERIFIED] to [INFERRED] (v1.0.8)",
type: "integrity",
scope: "Rwanda — Rutongo Tin Mines deposit record (js/data.js COUNTRY_PROFILES and AFRICA_DEPOSITS)",
was: "A pre-outreach integrity audit of the Rwanda country profile, conducted immediately after the private-circulation briefing PDF to the Rwanda Mines, Petroleum and Gas Board, tested every specific technical claim in the AI-generated Rwanda brief against named public sources. Three of four specific claims verified cleanly against the RMB's own website (rmb.gov.rw — CEO Francis Kamanzi's Rutongo tour entry): (a) Rutongo production average of ~80 tonnes of cassiterite per month; (b) 25.4% of Rwanda's total mineral production in 2023; (c) operational history since the 1920s. Alice Uwase's career sequence (Ngali Mining → Head of Mining, Petroleum and Gas Exploration Department → Deputy CEO from June 2024 → CEO from 18 July 2025) verified via RMB's own Management page and multiple independent news sources. The fourth claim — 'average ore grade 0.58% cassiterite' at Rutongo — carried a [VERIFIED: RMB Apr 2024; Trinity Metals Group] tag but could not be traced to any named public source after three targeted searches (RMB website, Trinity Metals corporate site, 2025 Trinity Metals media tour coverage, USGS Minerals Yearbook 2017-2019 Rwanda chapters, ScienceDirect studies on Rwandan tin-tantalum ore, industry interviews). The figure is directionally plausible within observed Rwandan cassiterite ore grade ranges (0.067–0.469% at Musha-Ntunga and Kamonyi per ScienceDirect 2021; broader 0.5–1% context per Asian Metal industry interview 2025) but plausible is not the same as sourced.",
now: "Rutongo grade figure retained in the data record, tag softened from [VERIFIED] to [INFERRED] with explicit disclosure. Scale field in COUNTRY_PROFILES Rwanda → known_deposits → Rutongo Tin Mines now reads: 'Rwanda's largest tin producer. ~80 tonnes Sn per month (2023); contributed 25.4% of national tin production 2023. Operational since the 1920s. Average ore grade reported as ~0.58% cassiterite [INFERRED — no primary public source located Apr 2026; pending verification].' Source field now reads: '[VERIFIED: RMB website — production tonnage, 2023 share, operational history; grade figure INFERRED only]' — distinguishing verified claims from the inferred claim at the field level. Matching edit applied to the Rutongo entry in AFRICA_DEPOSITS (res_source field) so the map marker carries the same honest epistemic status as the country profile. No content removed. Grade figure preserved with honest epistemic label so a subsequent researcher with access to Trinity Metals internal reporting or RMB archived technical files can promote the tag back to [VERIFIED] with citation.",
source: "Integrity audit conducted 22 April 2026 evening after briefing PDF sent to Alice Uwase, CEO RMB. Verification sources: rmb.gov.rw/updates/news-detail/ceo-francis-leads-a-tour-of-rutongo-mines (production and 2023 share); rmb.gov.rw/about/management (Alice Uwase career sequence); The New Times, iGihe, ChimpReports (Uwase appointment); ZoomInfo (career verification). Sources searched and not found for grade figure: rmb.gov.rw, trinity-metals.com, Trinity Metals 2025 media tour coverage (Mining Outlook, Africa Outlook), USGS Minerals Yearbook Rwanda chapters 2017/2018/2019, NS Energy Business Rutongo technical page, TINCO Investments legacy pages, International Tin Association project tables, ScienceDirect 2021 Rwandan tantalum-tin ore study (Ntunga and Kamonyi samples). The audit itself is an instance of the Afrimintel epistemic-tag integrity protocol functioning as designed — a claim was challenged, tested against public record, retained with honest labelling rather than silently removed."
},
{
date: "2026-04-22",
title: "Regional framing neutralised — DRC-Rwanda border dynamics, M23 references, conflict minerals language (v1.0.5)",
type: "integrity",
scope: "Platform-wide — regional security framing and AI prompts",
was: "Pre-launch review for RMB audience identified specific phrasings across the platform that could be read as taking sides in Central African regional dynamics. Four specific items: (1) An ESG risk entry labelled 'M23/DRC conflict zones (Eastern DRC)' with 'M23 advance through Goma region' — naming of specific non-state armed actor in commercial risk context. (2) A Congo Craton ESG block citing 'ICGLR Red Flag list identifies conflict minerals' — 'conflict minerals' is terminology Rwanda strategically disputes because it has historically been used to implicate Rwanda's 3T sector by association with eastern DRC. (3) A forward scenario in Afrimintel methodology stating 'DRC conflict extends westward from Kivu into the craton interior' — the framing 'DRC conflict' implicitly accepts a contested characterisation. (4) AI-generated Rwanda briefings framing regional dynamics with words like 'contagion' and 'instability' that position Rwanda as a passive victim or contaminated party. None of these were factually wrong, but all used politically sensitive framings that could be problematic in a platform under review by Rwandan government officials.",
now: "Five specific changes applied. (1) Eastern DRC ESG risk reframed without naming M23 — now reads 'Eastern DRC security dynamics' with neutral commercial language covering North Kivu, South Kivu, Ituri operational constraints and explicit reference to the 4 December 2025 Washington Accords peace framework. (2) Congo Craton ESG reframed to remove 'conflict minerals' language — now describes 'Security constraints on mining operations in Eastern DRC. Operators align with ICGLR Regional Certification Mechanism and ITSCI traceability standards for 3T minerals.' (3) Westward-extension forward scenario reframed to 'Eastern DRC security constraints extend westward... commercial access structurally narrows' — neutral operational language. (4) DRC country profile infrastructure note reframed: 'operational constraints in North Kivu, South Kivu, and Ituri provinces continue to affect Eastern DRC commercial logistics despite the 4 December 2025 Washington Accords peace framework.' (5) AI country-intelligence prompt extended with REGIONAL FRAMING RULE instructing the LLM to avoid words like 'contagion', 'instability', 'conflict minerals' in ways that imply the subject country is a passive victim; to use neutral commercial language for cross-border dynamics; and to avoid naming armed groups unless essential, using neutral descriptors when unavoidable. Policy is symmetric — applies to any cross-border context, not just DRC-Rwanda.",
source: "Pre-outreach review for Rwanda Mines Petroleum and Gas Board (RMB) leadership. All DRC coverage retained — Afrimintel covers DRC comprehensively as a continental platform. Changes are to framing language only, not to factual content."
},
{
date: "2026-04-22",
title: "AI prompt tightened — specificity rule and geopolitical framing guardrail (v1.0.4)",
type: "integrity",
scope: "AI prompts — country intelligence",
was: "v1.0.3 successfully grounded AI country briefings in Afrimintel curated data, resolving the Rwanda hallucination issue (correct Mining Law 58/2018, Alice Uwase as CEO, Trinity Metals, Washington Accords). However, two softer issues remained visible in the deployed output. First, the AI chose to reference Rwanda's export growth as '46.2% YoY' (correct) but did not cite the absolute dollar figure ($2.56B 2025) despite that figure being present in the grounding context. Second, the AI output framed Rwanda as a 'NATO/US critical mineral supply alternative' — technically defensible given the US-Rwanda Framework but inaccurate framing: Rwanda is not a NATO member and Rwanda's strategic positioning is explicitly multi-vector (US, China, Gulf, India, regional). An RMB-level reader could reasonably read 'NATO supply alternative' as Western framing being projected onto Rwandan policy rather than describing Rwanda's actual strategic posture.",
now: "aiCountry() prompt extended with two explicit rules. SPECIFICITY RULE: 'When the Afrimintel data provides specific figures (dollar amounts, production tonnages, grade percentages, dates), USE THE SPECIFIC FIGURES — do not substitute relative phrasing. If export_revenue_note contains a dollar figure, cite the dollar figure not just the growth rate.' GEOPOLITICAL FRAMING RULE: 'Do NOT frame African countries through NATO alignment. Most African countries are not NATO members. Use neutral commercial framing — Western supply chain alternative, non-Chinese supply routing, or name specific partner countries/agreements — rather than NATO supply alternative.' Rule applied globally so this framing is avoided for every country brief Afrimintel generates. Entry Signal section also updated to explicitly instruct citation of dollar figures from export_revenue_note where applicable.",
source: "Issue observed on live v1.0.3 deploy: AI briefing for Rwanda referenced 46.2% YoY growth without the $2.56B figure, and framed Rwanda as a NATO/US critical mineral supply alternative. User flagged both before sending WhatsApp to RMB CEO."
},
{
date: "2026-04-22",
title: "AI country intelligence grounded in curated profile data (v1.0.3)",
type: "integrity",
scope: "AI prompts — country intelligence, country risk",
was: "The AI-generated Country Intelligence briefing and Country Risk Assessment functions were explicitly anchored to the LLM's training-data cutoff — the prompt literally instructed the model to scope output to 'mid-2025 (your training limit)' and the LLM correspondingly produced briefings labelled 'Mid-2025 Reference'. The AI had access only to country name and primary commodity from Afrimintel's curated data; everything else — regulatory framework, operators, ministers, CEOs, export figures, strategic context — was filled in from the LLM's own training memory. For Rwanda, this produced output citing 'Mining and Quarrying Act (Law No. 17/2018)' (incorrect; the current Rwandan governing law is Mining Law No.58/2018) and did not reference Alice Uwase (CEO since July 2025), the US-Rwanda Framework for Shared Economic Prosperity signed December 2025, the 2025 export figure of $2.56B, or Trinity Metals Group as operator of Rutongo and Nyakabingo. The carefully rebuilt Rwanda country profile (v1.0.1) and the rebuilt DRC profile (v1.0.2) rendered correctly in the top summary card but were then undermined by the AI briefing below, which served stale training-data content as an 'institutional briefing'. This was the single most reputationally material issue on the platform: the curated data was correct but the AI-generated layer broadcast training-era references to every reader.",
now: "aiCountry() function rebuilt. Every AI country briefing now receives a structured AFRIMINTEL AUTHORITATIVE DATA block as grounding context — CEO/Minister name, governance framework, Fraser IAI, export figures, active operators, key deposits, infrastructure, strategic context, known data gaps — pulled live from the COUNTRY_PROFILES entry for that country. Prompt explicitly instructs the LLM: 'When the Afrimintel data conflicts with your training data, USE THE AFRIMINTEL DATA, not your training data. Do NOT cite laws, ministers, CEOs, or figures that contradict the Afrimintel data. Do NOT label this briefing as mid-2025 or any date reference.' Same grounding approach applied to aiRiskAssessment() — country risk briefings now anchor in the curated CEO/Minister/strategic context block. Result: AI briefings for Rwanda will now use 'Mining Law No.58/2018' not the hallucinated 17/2018; will reference Alice Uwase, Trinity Metals, $2.56B exports, Washington Accords; will not be labelled 'mid-2025'. Same correction applies to every country with a curated COUNTRY_PROFILES entry — South Africa, Botswana, Namibia, DRC, Rwanda, Zambia, Ghana, Zimbabwe, Kenya. Countries without curated profiles still fall back to training-data-based briefings but with the mid-2025 date anchor removed.",
source: "Issue discovered on live v1.0.2 deploy when inspecting Country Intelligence pane for Rwanda. AI-generated text cited 'Mining and Quarrying Act (Law No. 17/2018)' and was subtitled 'Mid-2025 Reference' — despite the curated Afrimintel Rwanda profile above it showing current Law 58/2018, Alice Uwase as CEO, $2.56B 2025 exports, and Washington Accords Dec 2025 context. Root cause: AI prompt was not given access to the curated profile data and was explicitly instructed to scope to training cutoff. Fix: grounding context injection pattern."
},
{
date: "2026-04-22",
title: "DRC coverage rebuilt to April 2026 reality (v1.0.2)",
type: "data",
scope: "DRC — country profile, deposits, strategic context",
was: "DRC did not have a dedicated COUNTRY_PROFILES entry — coverage was distributed across COUNTRY_COVERAGE action narrative, Lufilian Arc and Congo Craton province records, and AFRICA_DEPOSITS individual deposits. Stale or missing: previous Minister of Mines (Kizito Kapinga) not current; current Minister Louis Watum Kabamba (appointed August 2025, former Ivanhoe Kamoa Copper MD) absent. Cobalt export ban (Feb 2025) and quota regime (Oct 2025 ARECOMS, 96,600t/yr 2026) not captured. US-DRC Strategic Partnership (4 Dec 2025 Washington Accords, 50% copper + 30% cobalt to west via Lobito corridor within 5 years) absent. Xcalibur Smart Mapping $297.8M airborne programme (Feb 2026 launch, 2.7M line-km, 4 provinces, 3-year duration) not reflected. March 2026 CAMI mass revocation of inactive permits absent. KoBold Metals-Manono MoU (July 2025) not captured. Kamoa-Kakula Dec 2025 NI43-101 reserve downgrade from 472 Mt @ 3.94% Cu to 466 Mt @ 2.82% Cu post-May 2025 seismic event — v6.3.43 caught in Lufilian audit; DRC-level profile did not carry this context.",
now: "Full DRC COUNTRY_PROFILES entry created. Minister Louis Watum Kabamba named with career history (former Ivanhoe Mines DRC GM, MD Kamoa Copper SA, led Randgold Kibali development 2010-2014, prior Minister of Industry, University of Lubumbashi Masters). Deputy Minister Godard Motemona Gibolum noted. Strategic context block covers five headline structural shifts: Feb 2025 cobalt export ban, Oct 2025 ARECOMS quota regime (with 10% strategic reserve, 10% royalty advance, 48-hour compliance certificate requirement), Aug 2025 Watum appointment and consultation framework with mining companies, 4 Dec 2025 Washington Accords US-DRC Strategic Partnership with Lobito corridor commitment and DFC-Gécamines-Mercuria JV with US right of first refusal, March 2026 CAMI mass revocation. Known deposits rebuilt: Kamoa-Kakula with current reserve figures, Tenke Fungurume (CMOC+Gécamines 80/20), Kisanfu (CMOC 71%), Kamoto Copper Company (Glencore-Gécamines JV), Mutanda, Sicomines, Kipushi with 2025 output and 2026 guidance, Manono Lithium with KoBold MoU July 2025 plus AVZ tenure dispute, Kibali (Barrick 45% / AngloGold 45% / SOKIMO 10%), Bisie Sn (Alphamin), Chemaf Mutoshi with Virtus Minerals acquisition blocked by Gécamines. Xcalibur Smart Mapping programme detailed in key_surveys. Lobito corridor named as primary infrastructure constraint and commitment. Fraser IAI 18.4 contextualised. COUNTRY_COVERAGE action narrative updated to reflect Xcalibur transformation and regulatory reform.",
source: "Mining Review Africa (Louis Watum Kabamba appointment Aug 2025); The Africa Report (Watum career background, US-DRC negotiations, Chemaf-Virtus block Nov 2025); Bankable Africa (Watum consultation framework Sept 2025); Mining Technology / Fastmarkets / Serrari Group (cobalt quota regime Dec 2025-Feb 2026); Mining.com (US copper sales fivefold increase Dec 2025); IEA Global Critical Minerals Outlook 2025 (DRC cobalt ban context); Cobalt Institute 2024 Market Report (production figures); US Department of Commerce Rwanda Commercial Guide (DRC coverage update Feb 2026); Egmont Institute, Atlantic Council, CSIS, Oakland Institute (Washington Accords Dec 2025 context and controversy); Ivanhoe Mines NI43-101 effective 31 Dec 2025; CAMI March 2026 revocation announcements; Xcalibur Smart Mapping press release Feb 2026."
},
{
date: "2026-04-22",
title: "Rwanda coverage rebuilt for RMB review readiness (v1.0.1)",
type: "data",
scope: "Rwanda — country profile, deposits, strategic context",
was: "Rwanda COUNTRY_PROFILES entry carried stale and incorrect institutional detail: 'GMD established 2011' and 'Mining Law No.13/2018' misstated the RMB formation date (February 2017) and the governing instrument (Presidential Order n° 028/01 of 18/05/2023; Mining Law No.58/2018). Rutongo operator listed as 'Tinco Investments' — superseded by Trinity Metals Group. Nyakabingo tungsten mine (Africa's largest) not named. Export revenue figure $1.75B reflected 2024; 2025 figure of $2.56B (46.2% YoY growth) not captured. Washington Accords of 4 December 2025 — including US-Rwanda Framework for Shared Economic Prosperity signed by Presidents Kagame and Trump — absent from strategic context. CEO not named. AFRICA_DEPOSITS array carried zero Rwanda deposit records, meaning Rwanda rendered with no markers on the Africa Mineral Map preview despite being one of Afrimintel's most data-mature jurisdictions. MINIRENA URLs used for RMB-owned datasets (MINIRENA is the environment ministry; mining sits with RMB under MININFRA). Province tag alternated inconsistently between 'Congo Craton' and 'East African Rift'.",
now: "Full Rwanda rebuild applied before outreach to RMB CEO Alice Uwase. Changes: (1) COUNTRY_PROFILES entry rebuilt — RMB governance corrected (Presidential Order 028/01 of 2023, Mining Law 58/2018), CEO Alice Uwase named with appointment date (18 July 2025) and career history, province tag standardised to 'Karagwe-Ankole Belt (Kibara) + East African Rift flank' reflecting Rwanda's two-province geological reality. (2) Known deposits rebuilt — Rutongo Tin Mines (Trinity Metals Group, 80t/month, 25.4% national production 2023), Nyakabingo Tungsten Mine (Africa's largest tungsten mine, Trinity Metals), Gatumba Mining District, Bugarama-Bukunzi Wolfram, Musanze Carbonatite Context, NW Rwanda Gold Belt, Kivu Li-brine potential. (3) Export revenue updated — $2.56B (2025), +46.2% YoY from $1.75B (2024), source Rwanda Ministry of Finance and Economic Planning. (4) Strategic context added — US-Rwanda Framework for Shared Economic Prosperity (Washington Accords, 4 Dec 2025) alongside DRC-Rwanda peace agreement, Regional Economic Integration Framework (CIER), and US-DRC Strategic Partnership; raw tin became a top Rwanda export to the US by November 2025. (5) Six Rwanda deposits added to AFRICA_DEPOSITS array so they render on the map: Rutongo, Nyakabingo, Gatumba, Bugarama-Bukunzi, Musanze, NW Rwanda Gold Belt — all with operator and source attribution. (6) MINIRENA URLs corrected to rmb.gov.rw across HIGH_DATA_COUNTRIES. (7) Data counts updated across platform: 112 total deposits (was 106), 94 spatial-reference (was 88), 18 intelligence-grade unchanged. (8) Country contact updated from 'mines@minirena.gov.rw' to 'info@rmb.gov.rw' with verified RMB hotline (+250 788 386 220). Rwanda coverage now reflects April 2026 reality and includes strategic positioning RMB team will expect.",
source: "US Department of Commerce Rwanda Mining Guide (Mar 2026); RMB official website (rmb.gov.rw/about/management and /about/overview); The New Times (Alice Uwase appointment July 2025; 3T mineral export growth Feb 2026; Trinity Metals Nyakabingo coverage); The EastAfrican / Zawya (Rwanda mineral exports $2.56B 2025 rising 46.2%); Ecofin Agency (Washington Accords 4 Dec 2025 framework); Atlantic Council, CSIS, Egmont Institute (peace agreement context and controversy); USGS Mineral Commodity Summaries 2020-2024 (tantalum and tin global position)."
},
{
date: "2026-04-22",
title: "Brand transition — STRATA renamed to Afrimintel for public launch (v1.0)",
type: "methodology",
scope: "Platform identity",
was: "The platform was developed through April 2026 under the working name STRATA. Domain-availability research found the name STRATA heavily occupied across adjacent African mining sectors: Strata Worldwide (mine safety / collision avoidance, 210 employees in South Africa), Strata Mining Services (equipment rental, 75% Black-owned, Level 2 B-BBEE), Strata Energy Minerals Resources (diversified African mining operator at strata-africa.com), and the technical term 'strata control' (underground excavation safety). The word was doing brand work for many operators, none of them mineral intelligence platforms. A specialist reader hitting 'strata' in Google had 3-4 cognitive steps before reaching mineral intelligence. Every STRATA-anchored domain worth buying — strata.com, strataintelligence.com, strata-intelligence.com, strata.africa, strata.io, strataintel.com — was either taken by an unrelated operator or priced at premium. Domain availability pattern was primary evidence: the clean options were descriptive portmanteau names, not brand names.",
now: "Platform renamed to Afrimintel. Domain afrimintel.com registered at Cloudflare Registrar 22 April 2026, expires 22 April 2029. Email infrastructure: Cloudflare Email Routing configured, nikesh@afrimintel.com → forwards to personal Gmail, verified working (external test from nik@mauritel.mu delivered 22 April 2026). Branding transitioned across all user-visible surfaces: topbar logo (STR[A]TA → AFRI[M]INTEL), page titles, meta descriptions, onboarding copy, ASK overlay, scoring labels (STRATA Score → Afrimintel Score), AI system prompts (all LLM responses now self-identify as Afrimintel), methodology page, about page, roadmap page, audit log header/footer, Critical Minerals regulatory clock panel, partnership email templates (BGS, Fraser, AfDB, USGS — all rewritten), contact surfaces, Netlify function CORS origins (primary: afrimintel.com; strata1.netlify.app retained on allowlist for transition continuity). Version reset: Afrimintel v1.0 (April 2026). Historical record preserved: audit log entries dated before this release retain STRATA references as historical documentation; internal data fields (methodology notes, calibration report IDs like STRATA-CAL-2026-001, deposit record metadata) retain STRATA terminology because renaming them would misrepresent the historical chronology. Methodology page opens with explicit brand heritage note: 'Afrimintel was developed through April 2026 under the earlier working name STRATA.' Email address migrated from npukuk@gmail.com to nikesh@afrimintel.com across all user-facing surfaces. Platform hosting unchanged (Netlify); afrimintel.com DNS pointing to existing deploy.",
source: "Session decision 22 April 2026. Brand architecture: Option B (brand transition with continuity acknowledgement) selected over Option A (full rebrand with history erased) and Option C (dual-brand). Version numbering: Option A (fresh v1.0) selected over preserving v6.4.3 lineage in public marker. Institutional voice pass applied alongside rename. Rwanda Ministry of Mines brief remains outstanding as the final non-negotiable step before public LinkedIn launch; rename does not change that diplomatic protocol."
},
{
date: "2026-04-22",
title: "Africa map preview embedded on Intelligence pane landing (v6.4.3)",
type: "ux",
scope: "Intelligence pane — spatial anchor",
was: "Explorer role landing pane (Intelligence) led with Deep Dive dropdowns, stats row (13/18/40/19), and Province Ranking cards. The Africa map — the strongest visual credibility anchor STRATA carries — was one sidebar click away on the Map pane. A first-time visitor arriving from a specialist channel (LinkedIn launch, partner introduction) saw workflow tooling before they saw the spatial context that makes STRATA's province thesis legible. The map was treated as a workflow utility rather than a credibility signal.",
now: "320px non-interactive preview map embedded between the stats row and Province Rankings on the Intelligence pane. Renders all 106 deposits using the same two-layer convention as the full Map (spatial-reference dim / intelligence-grade bright, commodity colour-coded). Scroll, drag, and zoom disabled on the preview — one glance for spatial context, click anywhere on it (or 'Open full map →' link) to navigate to the full Map pane for interaction. Framing line below: '106 deposits curated from USGS MRDS and CGMW (6,000+ mine sites indexed continent-wide). 18 decision-grade. 88 spatial-reference.' Restores the v1.x-era reference to the ~6,000 continent-wide mine sites that was dropped during the role-based architecture rebuild, framed as context for the 106 decision-relevant records STRATA has actually consolidated. Scoped to Explorer role only for now — other roles land on workflow-specific panes where a map would be a detour, not an anchor.",
source: "User observation 22 April 2026 that the map was a visual tool worth surfacing more prominently before the LinkedIn launch. Option A (embedded preview) selected over sidebar reorder, static SVG, or role-aware landing to preserve workflow-first landing behaviour while restoring spatial-first visual signal."
},
{
date: "2026-04-22",
title: "Nav overflow render bug fixed — 'More' items were hidden even when expanded (v6.4.2)",
type: "integrity",
scope: "Navigation — sidebar render",
was: "The buildNav() render function set inline style='' (empty) on overflow nav buttons when showMore=true, expecting that to display them. But the CSS rule .nav-hidden { display: none; } applies unconditionally when no inline style overrides. Result: on page load with showMore=true in localStorage, the '▲ Less' label rendered (indicating expanded state) but all 10 overflow items remained hidden. Click-toggle via toggleNavMore() worked at runtime because it explicitly set style.display='flex', but any re-render (navigation, role switch, refresh) reverted items to hidden. Pre-existing bug. Critical & Battery Minerals was in overflow position 17 for Explorer, meaning it was invisible even though v6.4.1 added it to the menu correctly.",
now: "buildNav() render line changed: style=(showMore ? 'display:flex' : 'display:none'). Inline style now explicitly beats the .nav-hidden CSS rule when showMore=true. Sidebar now correctly shows all overflow items when in expanded state, across page loads, navigation, and role switches.",
source: "User bug report 22 April 2026 — 'Critical Minerals menu not visible after v6.4.1 deploy'. Traced through NAV_LABELS → PANELS_CONFIG → buildNav() → CSS .nav-hidden. Render/CSS conflict identified. Single-line fix."
},
{
date: "2026-04-22",
title: "Critical & Battery Minerals sub-vertical reconnected to role menus (v6.4.1)",
type: "integrity",
scope: "Navigation — role menu entries",
was: "The Critical & Battery Minerals pane (p-critical-minerals) was built and fully wired — the pane HTML existed, the NAV_LABEL was defined, the goTo() handler rendered scenario / Africa supply / battery chemistry / processing gap / geopolitics / province profiles on navigation — but the 'critical-minerals' key was never added to any of the five role menu objects in NAV_LABELS. The sub-vertical was therefore orphaned: accessible in code, unreachable through UI. Discovered during v6.4 deploy verification when the user could not locate the pane from any role sidebar. The v6.4 Regulatory Clock panel sat inside this orphaned pane, meaning no reader would have reached it through normal navigation.",
now: "'critical-minerals':'⚡ Critical Minerals' added to four of five role menus: Explorer (after Risks, before NI43-101), DFI (after Supply/Demand, before Operators), Government (after EITI, before Risk), Junior Mining (after Provinces, before Raise Capital). Excluded from Hardware Partner role — Hardware role is scoped to survey gear and terrain, not commodity-strategic positioning. Pane now reachable from every role where commodity-strategic context informs the workflow.",
source: "Pre-existing bug surfaced during v6.4 public deploy. Not a v6.4 regression — orphaned state existed before v6.4 was cut. Fixed 22 April 2026."
},
{
date: "2026-04-22",
title: "Precision fix — passport vs certification wording on Regulatory Clock panel (v6.4)",
type: "integrity",
scope: "Critical Minerals — Where STRATA fits",
was: "Regulatory Clock panel 'Where STRATA fits' paragraph read: 'STRATA does not issue passports or certifications — those sit with accredited bodies (IRMA, SGS, Bureau Veritas, regulated chain-of-custody systems).' This conflated two distinct regulatory instruments: battery passports (issued by economic operators under EU Reg 2023/1542, audited by notified bodies) and mine-site assurance (issued by independent bodies like IRMA, SGS, Bureau Veritas). Grouping them under a single sentence was directionally correct but imprecise — a regulatory specialist would flag.",
now: "Paragraph rewritten to separate the two layers explicitly: 'Battery passports are issued by economic operators under EU rules, audited by notified bodies. Mine-site assurance sits with independent bodies like IRMA, SGS, Bureau Veritas, and regulated chain-of-custody systems.' Closing phrase also extended from 'the passport layer applies to a specific shipment' to 'the passport or certification layer applies to a specific shipment or site' — keeping both regulatory surfaces distinct.",
source: "Internal hallucination check 22 April 2026 caught conflation before public LinkedIn launch."
},
{
date: "2026-04-22",
title: "Stale forward reference corrected on Methodology page (v6.4)",
type: "integrity",
scope: "Methodology — record tier classification timing",
was: "Methodology page stated that full record-level tier classification was 'scheduled for v6.4 (May 2026)'. v6.4 has now shipped (April 2026) without the sub-tier classification work. Original forward reference is no longer consistent with reality.",
now: "Reference changed to 'scheduled for a future release; timing will be confirmed in the Roadmap'. The sub-tier classification work remains an open commitment; the specific version and date will be re-stated when the work is scheduled.",
source: "Caught during v6.4 content-addition pre-deploy review."
},
{
date: "2026-04-22",
title: "Trust-gap framing added to Methodology page (v6.4)",
type: "methodology",
scope: "Methodology page — upstream framing",
was: "Methodology page described how STRATA was built (source hierarchy, epistemic tagging, audit cadence) but did not state the upstream problem the methodology was designed to solve. A reader could examine the disciplines without understanding why they exist, or what trust gap is being closed. Absence made STRATA's methodological discipline look like ornament rather than load-bearing.",
now: "New section 'The problem STRATA is built to solve' prepended before the 'What STRATA is' section. Names the African mineral verification discount explicitly — thinner verification chains, fewer NI 43-101 / JORC sign-offs, opacity on methodology — and states STRATA's methodological (not cryptographic) closure: source hierarchy, epistemic tagging, public audit log, three-tier deposit classification, Known Gaps and Verdict Reversal. Positions STRATA upstream of the certification layer rather than competing with it. Methodology page document version bumped to 1.3 (22 April 2026).",
source: "Session synthesis 22 April 2026. Framing informed by external perspective reviewed same day; cryptographic / Verifiable-Credential / eIDAS 2.0 pathway explicitly rejected as category contamination and retained only the trust-gap framing as directionally correct."
},
{
date: "2026-04-22",
title: "Regulatory clock panel added to Critical & Battery Minerals sub-vertical (v6.4)",
type: "methodology",
scope: "Critical Minerals — regulatory context",
was: "Critical & Battery Minerals sub-vertical covered IEA demand scenarios, Africa supply positions, battery chemistry decoding, processing gap, geopolitics, and province profiles. It did not anchor these to the dated regulatory events now driving capital-allocation decisions on African critical minerals — EU Battery Regulation, EU due diligence obligations, IRA §30D. Without the clock, the sub-vertical described the market without naming the regulatory timing that makes upstream intelligence commercially urgent.",
now: "New panel 'The Regulatory Clock — Why 2026-2027 Matters for African Critical Minerals' inserted between Africa Supply Positions and Battery Chemistry Decoder. Three dated regulatory triggers with [VERIFIED] tags and source citations: 18 Feb 2027 EU Battery Passport under Reg (EU) 2023/1542; 18 Aug 2027 EU due diligence obligations under Reg (EU) 2025/1561 (postponed from Aug 2025); IRA §30D critical-minerals sourcing threshold 80% in tax year 2027 (70% 2026, 90% 2028, 100% thereafter), FEOC restriction already in effect from 2025. 'Where STRATA fits' close explicitly disclaims passport/certification issuance — those functions sit with IRMA, SGS, Bureau Veritas, regulated chain-of-custody systems. STRATA is positioned as the upstream intelligence layer for the capital-allocation decision that precedes the passport layer.",
source: "EU Regulation (EU) 2023/1542; Regulation (EU) 2025/1561 (postponement); IRS Final Regulations on Clean Vehicle Credit Sections 25E and 30D; Federal Register Rule 2024-09094 (6 May 2024); CSIS analysis on IRA 30D critical mineral requirements and FTA partner sourcing."
},
{
date: "2026-04-21",
title: "Content-drift audit — 13 provinces, 43 edits across Gaps + Reversal panels (v6.3.43)",
type: "methodology",
scope: "PROVINCE_GAPS + PROVINCE_REVERSAL",
was: "v6.3.42 shipped Known Gaps and Verdict Reversal panels 21 April 2026. Within 72 hours, targeted internal audit against live 2025-2026 events surfaced material drift across all 13 provinces.",
now: "v6.3.43 applies 38 discrete edits: 3 factual rewrites where content was overtaken by events, 16 severity/framing updates, 4 new reversal conditions. Drivers include: Feb 2026 Xcalibur $297.8M DRC airborne survey launch; March 2026 Kamoa-Kakula reserve/guidance downgrade post May 2025 seismic; March 2026 Atlantic Lithium Ewoyaa parliamentary ratification; November 2025 Barrick-Mali settlement; Nov 2024 Centamin→AngloGold acquisition; Dec 2024 Somaïr nationalisation; July 2024 SAMIM withdrawal from Cabo Delgado; Eskom 300+ consecutive days without loadshedding (March 2026); Ethiopia-Eritrea tensions (Feb 2026 ICG 'powder keg' assessment). Observation: African mining intelligence ages on weekly-to-monthly cycles. Public audit log is a differentiator against static-at-ship competitors — every dated correction is the evidence.",
source: "Internal content audit 21 April 2026 against Ecofin Agency, Reuters, Mining Weekly, Bloomberg, Fitch/BMI, Ivanhoe Mines, Lifezone Metals, Rio Tinto, Syrah Resources, AngloGold Ashanti, Orano disclosures and ICG briefings."
},
{
date: "2026-04-21",
title: "Kalahari Platform — content audit (v6.3.43)",
type: "data",
scope: "Kalahari Platform — Gaps + Reversal",
was: "G1 sand-cover fill scoped only to 1km reconnaissance line spacing, no deadlock-of-data framing, no discovery-grade tier. G2 Namibia IAI severity MEDIUM; no mention of SWAPO-NDP local-participation rhetoric. G3 title conflated MMG (Khoemacau) with Sandfire (Motheo). R2 failure mode unspecified.",
now: "G1 split into Tier-1 reconnaissance (1km, $12-18M) + Tier-2 discovery-grade (200m) with deadlock-of-data framing. G2 severity HIGH; Nandi-Ndaitwah government (March 2025) + Mining Act amendment signal. G3 title corrected to MMG only; Sandfire note added as distinct entity operating Motheo. R2 rewritten with saline-aquifer conductivity interference as the specific failure mode. New R5 added: Namibian mandatory local-participation threshold >15%.",
source: "Internal audit vs Sandfire/MMG public disclosures and Namibian SWAPO-NDP government statements."
},
{
date: "2026-04-21",
title: "Lufilian Arc — content audit (v6.3.43)",
type: "data",
scope: "Lufilian Arc — Gaps + Reversal + Summary",
was: "G2 fill claimed 'no government or bilateral programme currently funded' — false for the Zambia side. G3 severity MEDIUM, fill claimed 3D seismic would solve depth problem. R4 LFP condition framed with undefined 'permanent shift' threshold. Summary framed DRC as only binding constraint.",
now: "G2 fill updated: Zambia national aerial mapping programme acknowledged; Ivanhoe Western Forelands budget referenced [UNVERIFIED ~$50M/yr]. G3 severity HIGH; fill references Kamoa-Kakula May 2025 seismic and March 2026 reserve downgrade (420kt→290-330kt 2026 output, 500kt steady-state pushed to 2028) as proof that depth is geotechnical not data-gap; combined 3D seismic + geomechanical programme specified. R4 tightened: Co price <$25k/t for 18 months AND LFP >70% AND DRC quotas lifted; note added that current DRC quotas have driven Co above $44k/t, counter-firing. New R5 added: Makoko District Q2 2026 grade gate. Summary rewritten to capture DRC-Zambia divergence.",
source: "Ivanhoe Mines webinar April 2026 (Kamoa-Kakula reserve/guidance update); Ecofin Agency 31 March 2026 on 2026 guidance; internal sources on Zambia aerial mapping programme."
},
{
date: "2026-04-21",
title: "Congo Craton — content audit (v6.3.43)",
type: "integrity",
scope: "Congo Craton — Gaps + Reversal + Summary + Framing",
was: "G1 fill claimed 'near-zero modern aeromagnetic, gravity, or EM coverage' with paused-2023 bilateral survey needing resumption — factually overtaken. G3 MEDIUM severity; GRACE/GOCE as adequate proxy. G4 lumped Cameroon/CAR/Gabon under single 'minimal modern data' framing. R3 compound condition mixed CAR failure with Cameroon/Gabon Fraser decline. R4 framing did not reference rainforest logistics. Bangui Anomaly entirely absent.",
now: "G1 rewritten for factually-correct state: Xcalibur Smart Mapping $297.8M programme launched Feb 2026, flying 2.7M line-km at 250m spacing across Kasai, Kwango, Kongo Central and Katanga (>700,000 km² combined); severity downgraded HIGH→MEDIUM; gap reframed from data absence to release timing. G3 severity HIGH; GRACE/GOCE limitation specified; >2km cover threshold economics named. G4 split into three country trajectories (Cameroon rising, CAR promotional, Gabon audit-risk); Bangui Anomaly named as the single most consequential unsurveyed target. R3 split into R3a (Gabon contract audit, Nguema Dec 2025) and R3b (CAR institutional failure). R4 sharpened: 'Witwatersrand-depth economics in rainforest logistics'. New R5 added: A&S Resources CAR iron ore 36-month NI 43-101/JORC validation checkpoint. Framing now explicitly time-bounded — gap-bonus will re-rate after Q4 2026 Xcalibur first releases.",
source: "Xcalibur Smart Mapping press release 6 Feb 2026; Ecofin Agency 24 Feb 2026 on DRC programme second phase; A&S Resources/CFI.co Feb 2026; Fortescue/Ivindo Iron disclosures; Cameroon 2026-2030 mining strategy."
},
{
date: "2026-04-21",
title: "West Africa Birimian — content audit (v6.3.43)",
type: "data",
scope: "West Africa Birimian — Gaps + Reversal",
was: "G4 Ghana Li pegmatite at MEDIUM severity, flagged Ewoyaa as 'only documented example' pending systematic survey. R2 condition written as binary — Barrick Loulo-Gounkoto permanent loss vs negotiated resolution.",
now: "G4 severity LOW: Atlantic Lithium Ewoyaa mining lease ratified by Ghana Parliament 20 March 2026 with sliding royalty (5% below $1,500/t, 12% above $3,200/t); first Li mining lease in Ghana; Ewoyaa DFS 36.8 Mt @ 1.24% Li2O. Province-wide pegmatite belt systematic assessment still pending. R2 reframed: Nov 2025 Barrick-Mali settlement established negotiated-resolution precedent; 10-year licence extension; 2026 guidance 260-290 koz attributable (vs 723 koz in 2024). Forward condition tests whether that precedent holds, not whether a binary resolution is reached.",
source: "Atlantic Lithium parliamentary ratification announcements 20 March 2026; Barrick Mining settlement disclosures Nov 2025 and Feb 2026 licence extension."
},
{
date: "2026-04-21",
title: "Kaapvaal Craton — content audit (v6.3.43)",
type: "methodology",
scope: "Kaapvaal Craton — Reversal",
was: "R3 threshold 'coordinated PGE production cuts beyond current 2024-2025 curtailments' did not specify baseline — Valterra demerger, Sibanye/Impala/Anglo restructurings and 4,000-7,000 SA jobs cut were not acknowledged as the baseline. R4 Eskom condition read as continuation of 2022-2023 crisis — inverted: SA had no loadshedding for 300+ consecutive days as of March 2026.",
now: "R3 baseline explicitly anchored to 2024-2025 survival-phase (Valterra demerger completed 2025; Anglo American Platinum 10% forecast production decline for 2025-2026; coordinated restructuring). Condition now fires on cuts 'significantly beyond' that baseline, not on any coordinated cut. R4 reframed: loadshedding regression from current stability rather than continuation of prior crisis. 300-consecutive-day baseline (March 2026) noted; firing now requires Stage 4+ return for 90+ days — a genuine regression, not normal state.",
source: "Eskom press releases (13 March 2026 300-day milestone); S&P Global PGM deep-dive February 2025; Valterra demerger 2025; SA Minerals Council 4,000-7,000 job cuts briefing February 2024."
},
{
date: "2026-04-21",
title: "Mozambique Belt — content audit (v6.3.43)",
type: "data",
scope: "Mozambique Belt — Gaps + Reversal",
was: "G2 fill referenced SADC force deployment 2021+ as reducing conflict. R1 condition 'Cabo Delgado escalates beyond 2021-2023 AND SADC force withdraws' — both parts were overtaken: SAMIM fully withdrew July 2024; insurgency activity was at most-active level since Palma 2021 in H1 2024. R3 attributed potential Balama suspension to Chinese synthetic-graphite competition — incorrect: the 2024-2025 Balama suspension was civil unrest following disputed Oct 2024 elections.",
now: "G2 fill updated: SAMIM withdrew July 2024; Rwandan Defence Force remains and has expanded post-withdrawal; AU technical assessment mission deployed July 2025. R1 rewritten: forward trigger is Rwandan force drawdown without successor guarantor AND insurgency at Palma intensity. R2 framing note: NdPr at ~$99k/t Jan 2026, BMI forecast 2026 average $90k/t in second-year-deficit market — $50k/t threshold is ~50% below current levels, condition fully dormant. R3 rewritten: Balama 2024-2025 suspension cause corrected to civil unrest post-October 2024 elections; forward trigger is third extended suspension period as structural signal.",
source: "SADC official SAMIM withdrawal announcement (July 2024); Syrah Resources force majeure declaration Dec 2024 and production restart announcements May-June 2025; BMI NdPr oxide forecast February 2026."
},
{
date: "2026-04-21",
title: "East African Rift — content audit (v6.3.43)",
type: "data",
scope: "East African Rift — Gaps + Reversal",
was: "G5 Kabanga extension noted 'Lifezone focus is Tanzania side only' without FS status. R1 referenced Feasibility Study reconciliation as trigger, but FS status was not specified.",
now: "G5 updated: Lifezone Metals filed Definitive Feasibility Study Technical Report Summary 18 July 2025 — Tanzania-side underground mine + concentrator, first Mineral Reserves declared. Tanzania side materially de-risked. Binding gap now is DRC/Burundi extension. R1 advanced: post-DFS trigger is either >15% downward resource reconciliation in a subsequent update OR construction/financing milestones slipping >24 months beyond currently guided schedule.",
source: "Lifezone Metals 18 July 2025 Feasibility Study Technical Report Summary; operational updates through Q1 2026."
},
{
date: "2026-04-21",
title: "West African Craton — content audit (v6.3.43)",
type: "data",
scope: "West African Craton — Gaps + Reversal",
was: "G4 Simandou downstream integration referenced 'Rio Tinto/Winning Consortium timelines publicly stated' without confirming first ore status. R1 condition 'Guinea Simandou timeline slips by >24 months from stated COD' was backward-looking. R4 Fe threshold $90/t without current-price context.",
now: "G4 updated: first shipment from Simandou left Guinea December 2025; mining formally launched 11 November 2025; Q4 2025 mine gate 2.3 Mt; Rio Tinto 2026 guidance 5-10 Mt/yr; RBC assesses 48-60 month ramp. Execution risk now rail + port throughput, not permitting. R1 advanced: post-first-ore trigger is 2026 guidance missed by >40% OR nameplate slippage beyond RBC 2029-2030 window. R4 framing note: Fe at ~$107/t April 2026; World Bank forecasts decline 2026-27 partly from Simandou supply; monitor but dormant.",
source: "Rio Tinto Q1 2026 operations update; Ecofin Agency 19 Jan 2026; S&P Global iron ore commentary January 2026; World Bank Commodity Markets Outlook October 2025."
},
{
date: "2026-04-21",
title: "Arabian-Nubian Shield — content audit (v6.3.43)",
type: "integrity",
scope: "Arabian-Nubian Shield — Reversal",
was: "R1 named 'Centamin Sukari' as the operator subject to care-and-maintenance trigger. Centamin entity is obsolete: AngloGold Ashanti completed $2.5B takeover November 2024; Centamin delisted from LSE/TSX. 2025 Sukari production 500 koz, +4% YoY — actively counter-firing.",
now: "R1 entity updated: 'AngloGold Ashanti Sukari (acquired from Centamin Nov 2024 in $2.5B takeover; Centamin now delisted)'. Condition expanded to include export permit suspension and licence revocation as firing paths. 2025 production 500 koz (+4% YoY) noted as current counter-firing signal.",
source: "AngloGold Ashanti acquisition completion July 2025; Arab Finance/Zawya Sukari 2025 production disclosure February 2026."
},
{
date: "2026-04-21",
title: "Afar Depression — content audit (v6.3.43)",
type: "data",
scope: "Afar Depression — Reversal",
was: "R2 condition limited to 'renewed Tigray conflict or Amhara escalation at 2022-2023 intensity'. R3 Eritrea-Ethiopia condition referenced Djibouti corridor closure without specifying threshold between troop buildup and actual conflict.",
now: "R2 expanded: now includes direct Ethiopia-Eritrea cross-border conflict as a distinct trigger; Amhara Fano insurgency baseline specified; ICG Feb 2026 'powder keg' assessment referenced. R3 tightened: condition requires military forces to clash (not just troop buildup) AND Djibouti corridor interdicted; Feb-April 2026 context noted — Ethiopian formal demand for Eritrean troop withdrawal, both sides massing forces near border, risk elevated but below firing threshold.",
source: "International Crisis Group Briefing 210 'Ethiopia, Eritrea and Tigray: A Powder Keg in the Horn of Africa' 18 Feb 2026; Al Jazeera Ethiopia FM letter to Eritrea 8 Feb 2026; Foreign Policy Jan 2026 2026 conflict risk assessment."
},
{
date: "2026-04-21",
title: "Saharan Metacraton — content audit (v6.3.43)",
type: "integrity",
scope: "Saharan Metacraton — Reversal",
was: "R3 listed 'Orano Arlit U, Tasiast Au-Mauritania, Gara Djebilet Fe Algeria' as operator-precedent proof points whose sustained interruption would erode the floor. Orano Arlit (SOMAÏR) condition has fully fired: Niger nationalised Somaïr December 2024; Imouraren permit revoked post-2023 coup; Nov 2025 Niger independent uranium sales despite ICSID injunction. Entry did not reflect this.",
now: "R3 rewritten: explicitly records Orano operator-precedent eliminated (Somaïr nationalised Dec 2024, Imouraren revoked post-coup, Nov 2025 Niger independent uranium sales). Forward trigger advances to Tasiast (Mauritania, Kinross) or Gara Djebilet (Algeria, Sonarem) facing sustained interruption — the remaining operator-precedent proof points.",
source: "World Nuclear News coverage of Somaïr nationalisation Dec 2024-2025; ICSID arbitration filings and September 2025 tribunal decision; Enerdata Niger nationalisation brief; Africa Business Insight Tiani televised address 17 Feb 2026."
},
{
date: "2026-04-21",
title: "Cape Fold Belt — content audit (v6.3.43)",
type: "data",
scope: "Cape Fold Belt — Gaps + Reversal",
was: "G2 Namibia Fraser IAI contribution noted 'direct verification would tighten composite' without referencing Kalahari G2 linkage. R4 Zn/Pb threshold 'below current levels' lacked market context.",
now: "G2 linked to Kalahari Platform G2 (shared Namibian SWAPO-NDP local-participation dynamics). R4 framing note: Zn AISC projected to decline 2026 on byproduct credits + TC shifts; Pb AISC similarly modestly down; prices currently stable; condition dormant.",
source: "S&P Global Market Intelligence mine cost outlook 2026 January/February 2026; Teck Resources 2026 guidance."
},
{
date: "2026-04-21",
title: "West Congolian Belt — content audit (v6.3.43)",
type: "data",
scope: "West Congolian Belt — Gaps + Reversal",
was: "G1 regional geophysical coverage referenced only future multi-country bilateral programme. R4 condition 'Dikulushi or Kipushi operations suspended' did not reflect Kipushi's actual operational status.",
now: "G1 updated: Xcalibur $297.8M DRC programme covers the Kongo Central portion of the belt, results expected Q4 2026 onwards; Angola and Republic of Congo segments remain unaddressed; multi-country programme still needed for non-DRC portions. R4 updated: Kipushi restarted June 2024, 2025 production 203,168 t zinc (guidance met), 2026 guidance 240,000-290,000 t; condition dormant, ramp-up on track; forward trigger is second unplanned suspension.",
source: "Xcalibur Smart Mapping programme scope Feb 2026; Ivanhoe Mines 2025 production results and 2026 guidance released 15 Jan 2026."
},
{
date: "2026-04-21",
title: "Verdict-reversal mechanic added to every province dossier",
type: "methodology",
scope: "Province detail view",
was: "Province scores (e.g. Kalahari Platform 9.2, Saharan Metacraton 6.2) were published as static verdicts. A reader could see the methodology behind the score but not the specific conditions under which STRATA would change its mind.",
now: "Every province dossier now carries a 'WHAT WOULD CHANGE THIS VERDICT' panel. Each panel opens with a one-paragraph framing (what the score actually depends on) and lists 4 named, observable reversal conditions per province (52 conditions total across 13 provinces). Conditions are specific — named operators, named price thresholds, named governance survey outcomes — not generic risks. A reader or diligence analyst can now check the platform's verdicts against their own scenarios.",
source: "Three independent reviewer passes (April 2026) converged on this mechanic as the single strongest differentiator between a 'research report' and a 'decision platform'. This is the implementation."
},
{
date: "2026-04-21",
title: "Known Gaps panel added to every province dossier",
type: "methodology",
scope: "Province detail view",
was: "Gaps and UNKNOWN-tagged fields existed across province records but were not consolidated in the dossier. A reader had to parse the methodology text to identify what STRATA does not know.",
now: "Every province dossier now carries a dedicated KNOWN GAPS panel listing between 3 and 5 structural gaps per province (45 gaps total across 13 provinces). Each entry states the gap, its severity (HIGH / MEDIUM / LOW-MEDIUM / LOW), and what specifically would fill it (named survey programme, bilateral aid vehicle, data source, or political condition). Surfaces rigour that was previously invisible to readers who do not read methodology text.",
source: "Third-party reviewer pass April 2026 recommended converting epistemic tags from 'weakness signals' into 'roadmap signals'. This is the implementation."
},
{
date: "2026-04-21",
title: "Stale dataset version displayed on every page",
type: "version",
scope: "Topbar marker",
was: "DATASET v1.1 displayed in platform topbar across every page.",
now: "DATASET v6.3 (Apr 2026), matching the actual release branch.",
source: "Flagged via third-party platform review that described STRATA as 'v1.1 (April 2026)' — the reviewer was reading the stale label."
},
{
date: "2026-04-21",
title: "Orphaned HTML comment markers rendering as visible text",
type: "ux",
scope: "platform.html structure",
was: "Ten developer comment section headers (e.g. ═════════ PROVINCE BRIEF MODAL ═══════════════════════ -->) rendered as visible ASCII text on every page. Desktop users saw them as decorative-looking dividers; mobile users saw them wrap and stack.",
now: "All ten comment blocks repaired with opening