Afrimintel · Independence Policy · Full audit log

Corrections Register

Every correction Afrimintel makes is logged append-only — additively, never by silent overwrite — with a date and, where one exists, the authoritative document that prompted it. This register is the public, decision-maker-facing view of those entries, so the Correction-Velocity SLA can be checked rather than merely asserted.

58logged corrections
45with a named authoritative source
2026-05-09earliest logged
2026-06-22most recent
How to read this. Entries are drawn from the append-only audit log (regenerated at each deploy; current build v2.137.48). The Quality Standard sets the turnaround target — material errors corrected within 7 business days, non-material within 30 — and the independence firewall is visible here in action: COI-hygiene corrections sit alongside data and methodology fixes. By type: correction 45 · fix 9 · recheck-correction 2 · hostile-audit-remediation 1 · data-correction 1. This register reports what was corrected and when; it does not assert a per-entry detection-to-fix interval, which the audit log does not record.
DateVersionTypeWhat changedSource marker
2026-06-22v2.28.29fixSingle-sourced the cross-asset bankability table from the feed. /bankability/ hardcoded all eight assets' NPV figures, duplicating /api/v1/bankability.json and permitting silent numeric drift. …AL-2026-06-22-V2_28_29-BANKABILITY-SINGLE-SOURCE
2026-06-20v2.28.22fixVersion-display consistency fix. The v2.28.21 deploy correctly set afm-build-id to v2.28.21 but the live footer kept showing v2.28.20 - root cause: the version stamp is a data-afm-version span whose inline fall …AL-2026-06-20-V2_28_22-VERSION-DISPLAY-FIX
2026-06-20v2.28.21fixCOI hygiene - genericised the Rwanda 3T operator out of historical audit-log narrative, matching the convention already applied to recent entries. …AL-2026-06-20-V2_28_21-RWANDA-OPERATOR-COI-SCRUB
2026-06-20v2.28.20fixResponsive restyle of the two new cross-asset tables (ownership, bankability-at-8%). On narrow screens each row now stacks into a card with the column label above each value, so the Basis/caveat column — which …AL-2026-06-20-V2_28_20-E5
2026-06-17v2.28.5recheck-correctionKabanga G4 re-firmed to Met(design): deeper FS TRS §1.8 fetch confirms GISTM relevant-parts-met + ITRB in the PRIMARY filing (supersedes v2.28.4 downgrade); G5 stays Partial; wrong-asset candidate source reject …AL-2026-06-17-V2_28_5-E1
2026-06-17v2.28.4recheck-correctionRecheck caught Kabanga G4/G5 over-attribution: GISTM-alignment was secondary-sourced not primary-confirmed; G4/G5 Met->Partial; FS-TRS mis-citation removedAL-2026-06-17-V2_28_4-E1
2026-06-11v2.12.0fixv2.12.0 - 17W DCF engine repair ships, plus the province-IC verification closes. ENGINE: the five features from the 13 May validation report section 4 are implemented in the platform calcDCF and a python mirror …—
2026-06-11v2.8.2correctionv2.8.2 - Operator post-deploy screenshot of /insurance/ exposed a stale-band claim the v2.5.0 correction itself propagated: the live Manono panel correctly renders DRC at 45 MEDIUM from the chain, while the met …—
2026-06-11v2.8.1correctionv2.8.1 - Pre-deploy deep-structure audit on the accumulated v2.5.0-v2.8.0 bundle (operator-invoked protocol), six blocking findings, all fixed before deploy. …—
2026-06-11v2.6.0correctionv2.6.0 - F-EITI-8 RESOLVED (ruling: Option C graded, Claude-recommended / Nikesh-ratified 11 June). The composite EITI lookup was missing seven register members - including via a name-mismatch bug (map keyed Co …—
2026-06-11v2.5.1correctionv2.5.1 - Operator post-deploy screenshot of countries.json caught a third surviving surface of the retired editorial-override claim: the feed note still read 'Editorial override applies where composite understa …—
2026-06-10v2.4.3correctionv2.4.3 - Operator post-deploy check caught a second surface carrying the corrected 16-country claim: the eiti-detail.json resource note (a hand-written string in the emitter) still read 'the 16 African implemen …—
2026-06-10v2.4.2correctionv2.4.2 - Hostile contrarian audit on the full 10 June chain (v2.1.1 through v2.4.1): four probes, two defects fixed, one pass, one housekeeping. …—
2026-06-10v2.4.1correctionv2.4.1 - Hallucination-layer catch on v2.4.0, same day: the claim of THE 16 African implementing countries was wrong - the authoritative EITI register (eiti.org/countries, retrieved 2026-06-10) lists 28. …—
2026-06-10v2.3.2correctionv2.3.2 - Probe results: premise corrected, two live defects caught and fixed. (1) PREMISE CORRECTION: the v2.3.0 entry claimed Netlify Drop runs no dependency install; the staged live probe falsified this - the …—
2026-06-10v2.1.3fixv2.1.3 - Roadmap header logo still read STRATA - the old working name, live on a public page since the v1.0.9 rebrand. …—
2026-06-10v2.1.2fixv2.1.2 - Post-deploy verification of v2.1.1 caught what the v2.1.1 sweep missed. Three defects found reading the live about page: (1) a second check-count instance - 'pre-deploy pipeline (23 automated checks)' …—
2026-06-10v2.1.1fixv2.1.1 - Pre-deploy check-count drift purge across four static pages. about.html, /continuity/, /methodology/afdb-mandate-fit/, and the Kabanga pre-FID case study carried '23-check pre-deploy pipeline' claims ( …—
2026-06-10v2.1.0fixv2.1.0 - Commercial gating: summary-public / depth-paid (operator ruling: people must pay). The v2.0.0 full-read-layer-public architecture overshot - nothing load-bearing remained behind the paywall. …—
2026-06-06v1.8.7.4correctionv1.8.7.4 - Repointed the request-access capture function from Resend to Brevo, per operator confirmation that Brevo is the configured provider. …Operator confirmation of Brevo as provider (6 June 2026) + v1.8.7.3 live-test fallback screenshot. Function rewrite this session. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-06v1.8.7.3correctionv1.8.7.3 - Replaced the failing Netlify Forms capture with a dedicated serverless function so the institutional-access form actually delivers. …Operator live-test screenshots (6 June 2026) showing the fallback message = no capture. Function build and form rewire this session. Resend identified from the project _email.js module. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-06v1.8.7.2correctionv1.8.7.2 - Fixed the institutional-access form dumping submitters on the custom 404 page. Symptom (operator screenshot): submitting the v1.8.7.1 Netlify form returned the 'Page not found' 404 instead of a confi …Operator screenshot of the post-submit 404 (6 June 2026). Diagnosis and fix this session. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.7.1correctionv1.8.7.1 - Landing page (v1.8.7.0) corrected on five points from a hallucination check and hostile contrarian audit. …Hallucination check + hostile contrarian audit of v1.8.7.0 (this session); Independence Policy wording verified against methodology/independence. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.9correctionv1.8.6.9 - Stripped a site-wide stray '-e' artifact from 57 HTML pages. A build script dated 2026-05-24 appended a literal '-e ' line after </html> on nearly every page (homepage, methodology, about, press, all …Finding from the v1.8.6.8 post-deploy verification screenshots (operator, 5 June 2026). Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.7correctionv1.8.6.7 - Corrected the IG primary-source enumeration in the v1.8.6.6 platform.html noscript fallback (hallucination-check finding). …Finding from the v1.8.6.6 hallucination check, cross-checked against the about.html canonical IG definition. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.5correctionv1.8.6.5 - Closes the last two count-surfacing items so per-tier counts are internal across EVERY public surface. …AI-footer finding from the v1.8.6.4 hostile audit; PDF correction targets the live data-layer reality and the v1.8.6.3 editorial ruling, content reproduced verbatim from the shipped v1.8.6.4 PDF. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.4correctionv1.8.6.4 - CHECK-22 description corrected on quality-standard/index.html (hallucination-check finding). The v1.8.6.3 pass had reworded the CHECK-22 line to claim it 'enforces that the surfaced intelligence-grad …Finding from the v1.8.6.3 hallucination check (this session), cross-checked against track-record/index.html as the authoritative in-bundle description of CHECK 22. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.2correctionv1.8.6.2 - Count-drift close-out on live public surfaces. Brings stale hardcoded deposit and transaction counts into line with the live data-layer source of truth (18 intelligence / 9 operator-attestation / 57 …Drift identified by operator review (why-18, press) plus the full-surface count-drift sweep (methodology block + line 264, insurance). Authoritative target counts are the live computed data-layer tiers. Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.6.1correctionv1.8.6.1 - Corrections to the v1.8.6 Aterian Rwanda discovery record from a hallucination check and a hostile contrarian audit, applied before any deploy. …Corrections originate from the v1.8.6 hallucination check and hostile contrarian audit (this session). Underlying asset facts still rest on FurtherAfrica 16 Jul 2025 and Aterian RNS 19 Feb 2026 (operator/spec-supplied; network-on back-check recommended before any gated public surfacing). Build and corrections: Claude-drafted, Nikesh-directed. Editorial sign-off: Nikesh Patel.
2026-06-05v1.8.4hostile-audit-remediationv1.8.4 - Hostile contrarian audit of the v1.8.3 reconciliation caught a duplicate-inflated total and a one-surface-only fix. v1.8.3 was HELD pre-deploy (never shipped) and is superseded by this build. …Hostile contrarian audit run against the shipped artifact, not the working copy. Duplicate detection by name-base normalisation plus coordinate proximity over the runtime AFRICA_DEPOSITS array. No external authority introduced; no new data values asserted; the three coordinate conflicts are left Pending rather than resolved by guess. Counterparty Extension maintained - no body named as a partner. Audit, dedup ruling, and narrative: Claude-drafted, Nikesh-directed and reviewed.
2026-06-03v1.8.1data-correctionv1.8.1 - Verification sweep batch 1 (price bands): copper, nickel, gold spot prices flipped Pending to Sourced via World Bank Pink Sheet x IMF/LBMA dual-sourcing; gold corrected ~2x.World Bank Commodity Markets 'Pink Sheet' (released 4 May 2026, data to Apr 2026); IMF Primary Commodity Prices (PCPS, via FRED with IMF attribution); LBMA Gold Price PM (ICE Benchmark Administration). All cited as public-source authorities under their published terms - Counterparty Extension maintained, no partnership implied. DCF regression note: dcfUpdateDefaults() auto-fills spot prices from SUPPLY_DEMAND, so DCF default runs now reflect corrected prices; the Kamoa DCF test-battery NPV anchors are tied to their own stated copper assumptions and should be re-checked against the updated defaults.
2026-06-03v1.6.24correctionv1.6.24 - Public-surface audit: closed 7 orphan internal documents reachable by direct URL; hardened robots.txt.Public-surface audit run 2026-06-03: bundle file inventory + link-reachability analysis; afrimintel-public-surface-audit-spec-v0_1.md.
2026-06-03v1.6.23correctionv1.6.23 - Forced (410!) gravestone redirects so removed working files stop being served even when physically present in an overlay-built deploy.Netlify deploy permalink 6a1fd8853370e96e3e6ac03a--strata1.netlify.app returning 200 on a path absent from the canonical bundle; Netlify _redirects force-flag (!) shadowing behavior.
2026-06-02v1.6.22correctionv1.6.22 - no-store Cache-Control on /audit-log/* to evict a stuck Netlify Edge cache entry.Live header: cache-status Netlify Edge hit persisting on a path absent from the published deploy.
2026-06-02v1.6.21correctionv1.6.21 - 410 gravestones to evict stale Netlify Edge cache serving removed working files.Live header inspection: cache-status Netlify Edge hit on a path absent from the deploy.
2026-06-02v1.6.20correctionv1.6.20 - clean full-site redeploy to evict orphaned working files from production.Operator: rebuild clean rather than patch the deploy.
2026-06-01v1.6.19correctionv1.6.19 - methodology-canonical reconciliation: province scoring, tier count, exports, and bankability inputs aligned to the methodology page (governing authority for province IC/weighting) per operator rulings …Operator rulings 2026-06-01 (tier count to true 110, no padding; exports full reconciliation + re-sort, non-optional; bankability recompute, methodology rules; FIX5 sensitivity reframe; keep Fraser Pendings).
2026-06-01v1.6.18correctionv1.6.18 — public-bundle scrub: removed 10 internal working files from the publicly-served /audit-log/ path, clearing the 8 CRITICAL named-individual Counterparty-Extension leaks (v1.0.53 check).Operator: fix. Pre-deploy audit scripts/audit/pre-deploy-audit.js v1.0.53 named-individual Counterparty-Extension check; legacy scrub intent per v1_0_53-legacy-counterparty-scrub-spec.
2026-06-01v1.6.17correctionv1.6.17 — URGENT Loulo-Gounkoto cross-surface corrections ahead of Mining.com editorial scrutiny: Barrick rename/ticker, two wrong-year dispute entries, op-control date standardisation, neutral Bristow language …Operator urgent prompt (Mining.com editor-in-chief, Bristow-departure author, has platform access). Primary verification: SEC EDGAR 6-K filings (Sept 29 2025 leadership transition; Nov 10 2025 buyback; May 6/9 2025 name change), Barrick IR; settlement facts per Afronomicslaw + prior v1.6.x cycles.
2026-06-01v1.6.16correctionv1.6.16 — hostile contrarian audit on the Component E build (v1.6.13-15): three diligence-grade fixes (one inverted tier double-standard, one group-level padding, one scope-inflation in titles).Operator: hostile contrarian audit. Self-attack as a Cat-A/Cat-B diligence team; fixes verified against the same primary sources cited in v1.6.13-15.
2026-06-01v1.6.15correctionv1.6.15 — hallucination check on v1.6.13/v1.6.14 Component E build: two over-attribution precision fixes (no fabrications found).Operator: hallucination check. Re-verified against Lifezone IR/Sustainability Report + Crux (Kabanga); Barrick Sustainability/Afronomicslaw/Ecofin (Loulo); Zijin/Mongabay/BHRC/CFR (Manono); MIGA ESIA/DFC PIS/USTDA/Global Witness (Lobito).
2026-06-01v1.6.7.3correctionv1.6.7.3 — Hostile-contrarian audit. Corrected two false Fraser claims in the PROV province narratives that the v1.6.x Fraser-2025 reconciliation missed; flagged three findings that need an operator decision.Hostile-contrarian audit of the shipped v1.6.7.2 zip. DRC Fraser 2024 IAI ~48.52 (not 18.4) per Engineering News / Mining Weekly (2023-edition reporting). Namibia 2025 rank 51/68 per TanzaniaInvest, ZambiaInvest, African Mining Market (Mar 2026). AI-assistance attribution: Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-06-01v1.6.7.2correctionv1.6.7.2 — Hallucination check. Independently re-verified the 2025 Fraser ranks I had asserted, and resolved a same-record Sourced-vs-Derived contradiction the new fraser_state field exposed on three profiles.Hallucination check of the shipped v1.6.7.1 zip. Rank re-verification: ZambiaInvest, TanzaniaInvest, African Mining Market (all Mar 2026). Namibia ~56: African Mining Market quoting Chamber of Mines of Namibia (Mar 2026). AI-assistance attribution: Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-06-01v1.6.7.1correctionv1.6.7.1 — Fraser provenance render honesty. Four country profiles were rendering Derived IAI estimates as 'Sourced / verification-grade' on acquirer-facing surfaces; fixed at the root.Hostile-contrarian audit of the shipped v1.6.7 zip. Zambia second-source: African Leadership Magazine 'What Is Africa's Blueprint For Mineral-Led Growth?' (2 Mar 2026), 2025 Fraser IAI Botswana ~86 / Zambia ~73; ZambiaInvest (Mar 2026) 72.84 with YoY 64.23->70.02->72.84. 2025 ranks confirmed multi-source: SA 57/68, DRC 50/68, Namibia 51/68. AI-assistance attribution: Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.7correctionv1.6.7 — Closed the four items the hostile contrarian audit flagged for decision (H2, H3, H9, H10). All four were genuine and are now fixed rather than carried as known gaps.Operator instruction to fix everything flagged. SA rank-derivation reasoning from the Sourced 57th/68 rank; DRC blended-IAI ~58 from 50th/68 rank consistency with methodology; Tanzania overclaim removed pending direct PDF. AI-assistance attribution: fixes Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.6correctionv1.6.6 — Hostile contrarian audit (operator-requested). Adopted an adversarial posture (skeptical-acquirer / competitor-trying-to-discredit). …Operator-requested hostile contrarian audit. Adversarial internal grep (methodology-vs-code formula diff; low_confidence render-layer presence; RGI map provenance; manual-override detection; ownership-claim spot-checks). Rossing ownership verified against Rio Tinto plc SEC Form 6-K FY2019 (sale to CNUC completed 16 July 2019). AI-assistance attribution: audit + fixes Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.5correctionv1.6.5 — Second hallucination check (re-check of the v1.6.4 fix). Caught a residual the v1.6.4 reconciliation itself introduced: the PROV array IC values were updated for the SA/Ghana provinces but only one of …Operator-requested re-check. PROV-vs-CSV IC cross-surface diff (robust bracket-matched PROV parse + regex extraction of CSV rows). Confirms the discipline that a fix must be re-audited on the same full-surface basis as the original. AI-assistance attribution: re-check + propagation fix Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.4correctionv1.6.4 — Hallucination check (operator-requested). No fabrications found, but the check caught a live internal contradiction: the v1.6.0-1.6.3 Fraser 2025 updates were applied to COUNTRY_PROFILES + the risk-com …Operator-requested hallucination check. Full-bundle adversarial grep across all js/html/csv surfaces (not Claude-touched files only) for: permanent exclusions, counterparty/partnership overclaims, placeholders, version consistency, and every Fraser numeric introduced this session. Fraser 2025 values cross-checked against the published 2026 coverage already cited in v1.6.2-v1.6.3 (TanzaniaInvest, Ecofin, News Ghana, The South African, North Africa Post, African Leadership Magazine, MiningFocus Africa). AI-assistance attribution: check + reconciliation Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.2correctionv1.6.2 — Fraser 2025 verification across all nine Country Pitch Builder profiles, plus a second false-EITI map (missed in v1.6.1) corrected. …Operator directed a Fraser verification across all country pitch builders. Latest-edition status and African IAI scores/ranks verified against multiple primary-coverage sources for the Fraser 2025 survey (TanzaniaInvest, Ecofin Agency, News Ghana, The South African, North Africa Post, African Leadership Magazine, MiningFocus Africa); the Fraser PDF itself was bot-blocked, so exact decimals for DRC/Namibia/Zimbabwe remain pending a direct PDF read and are flagged rather than inferred. AI-assistance attribution: Fraser verification, corrections and the second-EITI-map fix Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-31v1.6.1correctionv1.6.1 — Material data correction: false EITI membership claims removed for three countries, plus a stale Fraser score and a fabricated-fiscal-terms engine leak fixed. …Operator directed a deep accuracy sweep across all nine Country Pitch Builder country profiles and the shared EITI / Fraser / fiscal datasets the engine reads. EITI membership verified against the EITI implementing-countries roster and the EITI \'Countries\' outreach page (eiti.org). Fraser correction is an internal-consistency alignment to the dataset\'s own stated current value; the 72.84 figure itself remains on the standing Fraser-PDF back-check list. Full primary-source re-verification of every numeric field in every country profile remains the deeper track. AI-assistance attribution: the adversarial audit, the EITI cross-check, the corrections and the engine hardening were Claude-executed under operator direction. Editorial sign-off: Nikesh Patel (pending review).
2026-05-13correctionv1.0.51 — CRITICAL Fraser year-labeling error correction: Tanzania value 46.38 (deployed v1.0.48.1 labeled 'Fraser 2024') was actually Fraser 2023; actual Fraser 2025 Tanzania = 68.04 (VERIFIED). …Fraser Institute Annual Survey of Mining Companies 2025 (released February 2026, downloadable PDF at https://www.fraserinstitute.org/sites/default/files/2026-02/annual-survey-of-mining-companies-2025.pdf — 256 responses; 68 jurisdictions); TanzaniaInvest March 2026 reporting 'Tanzania Ranks 4th in Africa for Mining Investment as World-Class Mineral Potential Outshines Policy Challenges' (Tanzania 2025 = 68.04 rank 34/68; trajectory 46.38 in 2023 → 62.75 in 2024 → 68.04 in 2025); ZambiaInvest March 2026 reporting (Zambia 2025 = 72.84 rank 25/68); Fraser Institute Annual Survey of Mining Companies 2024 PDF (released July 2025, Table 2 Policy Perception Index multi-year breakdown showing Tanzania 62.75/46.38/52.90/45.76/42.08 across years 2024-2020); Engineering News 16 May 2024 (Tanzania 46.38 Fraser 2023 reporting; Mozambique 31.90 Fraser 2023 reporting). Cross-verification: Fraser 2025 PDF Tanzania ranked 34/68; Fraser 2024 PDF Tanzania ranked 37/82; Fraser 2023 PDF Tanzania ranked 53/86 — three distinct surveys with three distinct scores.
2026-05-13correctionv1.0.50 — CRITICAL DCF tool ('deal evaluator') three-bug fix: (1) province-empty-on-load alert; (2) Cobalt→Copper substring mismatch in SUPPLY_DEMAND lookup; (3) substring-include lookup brittleness; plus Compo …DCF tool bug findings: senior external advisory feedback received 13 May 2026 noted 'I tried the deal evaluator for several provinces, metals and asks, and it did not work. Worth having a look to see if the tool is functional.' Reproduction via code-trace: dcf-province innerHTML population in fillProvSels() shows empty default option; onchange handler only fires on user-initiated change not on innerHTML population (documented browser behavior); dcf-price has no value attribute, placeholder text only. SUPPLY_DEMAND.find behavior with includes-substring matching reproduced via runtime simulation. Component E extension framework references verified: TNFD via tnfd.global (final recommendations September 2023; 500+ organisations committed to adopting by 2025; TNFD's 2025 MoU with the IFRS Foundation aligning with the ISSB framework; LEAP-approach: Locate/Evaluate/Assess/Prepare); SAVi via iisd.org/savi (developed by IISD, Canadian NGO; combines Vensim system-dynamics with Corality Smart project-finance models; published applications include transport, energy, water, and natural infrastructure but not mining; Kagera River SAVi work in northern Tanzania was a hydropower-and-natural-capital application, not a mining application — provides IISD with regional familiarity but does not constitute a mining-methodology precedent); IFC Performance Standards are 8 standards governing private-sector environmental and social sustainability (well-established framework); World Bank ESF + AfDB ISS are the parallel multilateral DFI safeguards frameworks.
2026-05-13correctionv1.0.48.2 — Hostile contrarian audit post-v1.0.48.1: SIGNIFICANT Kamoa Element A 2023 IDP Cu price correction ($3.10/lb → $3.70/lb for two rows; pre-existing error in v1.0.47-deployed panel exposed by back-chec …Hostile contrarian audit executed on v1.0.48.1 build pre-deploy. Web verifications anchoring corrections: Ivanhoe Mines 'Ivanhoe Mines announces outstanding economic results of the independent Integrated Development Plan (IDP) for the tier one Kamoa-Kakula Copper Project' 8 September 2020 (verified 2020 DFS/PFS/PEA all used $3.10/lb long-term Cu price consensus); Mining Weekly 'Kamoa-Kakula Copper Complex, Democratic Republic of Congo – update' April 2025 (verified 2023 IDP PFS used $3.70/lb long-term Cu price; same long-term price applied to 2023 IDP PEA). Code-trace verification of js/app.js dcfCountryRiskDiscountRate() function confirmed dynamic PROV.ic read with Math.round rounding to 1 decimal — Element D 12.5% auto-fill claim verified. Lifezone Form 6-K June 2025 IA TRS confirmed refinery sequencing 5 years after mine commencement.
2026-05-13correctionv1.0.48.1 — Hostile hallucination-check post-deploy-prep: CRITICAL data-layer error (Tanzania Fraser IAI value) and 2 MEDIUM panel-content errors corrected; Kabanga panel Element A expanded with Initial Assessm …Hostile hallucination-check executed on v1.0.48 build pre-deploy. Web verifications anchoring corrections: Engineering News 'Fraser Institute survey shows marked declines in African countries' investment attractiveness' 16 May 2024 (Tanzania 2024 IAI 46.38 down from 52.90, Mozambique 31.90 down from 34.96); Fraser Institute Annual Survey 2024 PDF (82 jurisdictions; insufficient-response jurisdictions noted); critical-minerals-news.com April 2026 nickel-price page (LME Class 1 $17,171/t spot; $20-22k/t Western sulphide incentive level); IEA Global Critical Minerals Outlook 2025 + carboncredits.com summary citing 'IEA project nickel demand more than doubling by 2035 under current pledges, potentially tripling in net-zero scenarios'; ING Think Nickel outlook Dec 2025 ($15,250/t 2026 forecast; 261 kt 2026 surplus); S&P Global per carboncredits.com (156 kt 2026 surplus); Lifezone Metals Form 6-K June 2025 IA TRS ($2.37bn NPV @ 8% / 22.9% IRR / 22-yr mine plan / vertically integrated scope per SEC filing); Lifezone Metals Form 6-K 18 July 2025 FS TRS ($1.58bn NPV @ 8% / 23.3% IRR / 18-yr LoM / Phase 1 mine+concentrator scope).
2026-05-13correctionv1.0.47.3 — Post-contrarian-audit comprehensive fix: all 4 SIGNIFICANT + 4 MEDIUM + 3 MINOR fixable findings applied across Bankability panel, methodology page, and js/data.js Simandou ownership precisionOperator instruction post-contrarian-audit 13 May 2026: 'fix everything fixable and build file for deploy.' v1.0.47.3 applies all SIGNIFICANT and MEDIUM and MINOR findings from the prior contrarian audit that were operator-decision-deferred rather than mandatory-fix. Ownership verification: Wikipedia Simandou mine article (Mar 2026 revision); Business & Human Rights Resource Centre 'Simfer S.A.' company page citing 45.05%/39.95%/15% split; Rio Tinto 'Government of Guinea, WCS and Rio Tinto Simfer incorporate La Compagnie du TransGuinéen' press release 27 July 2022 giving Simfer Jersey 53/47 internal split + CIOH composition (Chinalco 75% / Baowu 20% / CRCC 2.5% / China Harbour Engineering 2.5%); Reuters 18 January 2026 + energynews.oedigital.com 18 January 2026 for Baowu raise from 49% to 51%; Global Neighbours Nov 2025 cover-story Simandou for original WCS shareholding history (Winning International 45% / China Hongqiao 35% / UMSI 20% pre-Baowu); Mining Technology + Mining Weekly + FQM Q3 2025 results + Bloomberg 19 Aug 2025 for FQM Kansanshi commissioning dates (carried forward from v1.0.47.2 source citation).
2026-05-13correctionv1.0.47.1 — Hallucination-check fix: Bankability v0.1 panel NPV compression claim softened from overstated 38-48% to defensible 25-35% with profile-dependence framingHallucination check applied 13 May 2026 post-v1.0.47 first deploy. Arithmetic verification: for uniform cash flows over mine life n at discount rates r1=8% vs r2=12.2%, NPV ratio = annuity_factor(r2,n) / annuity_factor(r1,n). Results: n=15: 21.3% compression; n=20: 24.9%; n=21: 25.5%; n=25: 27.5%; n=33: 30.4%. Front-loaded CFs (which Kakula DFS exhibits given 77% IRR + 3.5yr payback) compress less than uniform-CF estimate. Cross-scenario realistic range: 22-38%. Pre-staged panel claim 35-45% was at high end of plausible range; v1.0.47 first-pass bumped to 38-48% which exceeded plausible range. Honest claim 25-35% with profile-dependence framing.
2026-05-09correctionv1.0.45 — Hostile-audit response: time-cost overclaim stripped from four dossiers; DCF cycle-1 Findings A and D input-set disambiguated; dossier disclaimer sections consolidated; Editorial Judgment discipline a …Hostile-audit pass against v1.0.45 deliverables conducted at editorial request. Vulnerabilities surfaced through five-layer analytic frame: (a) numeric verifiability under cold scrutiny, (b) decision-shape utility under skeptical reading, (c) Counterparty Extension under stress-testing, (d) self-presentation tells, (e) overall coherence as institutional product. Of fifteen vulnerabilities catalogued, three classified critical (time-cost overclaim across multiple dossiers; same overclaim as self-presentation tell; audit-log strategic-narrative language) and six classified significant. This entry corrects the three critical and two of the significant items (DCF input-set disambiguation; Quality Standard qualitative-claim discipline gap). Remaining significant items deferred: dossier disclaimer-vs-decision-aid framing (partially addressed by consolidation; further restructure deferred to W3); /why-18 page 'we chose 18, we could scale to 1,800' framing (deferred); Counterparty Extension scope question on third-party naming in published research (deferred — this is a discipline-architecture question requiring Quality Standard amendment, not a build fix). Material items (logistics-cost figure not Sourced; Angola country composite INFERRED; pricing-page Counterparty promise vs dossier-page practice; deploy cadence as a tell) acknowledged and deferred to W3. Items 5.2 (presentation-vs-operation gap) and 5.3 (no demonstrated institutional uptake) accepted as structural, not buildable.
2026-05-09correctionv1.0.45 — Hallucination-check corrections: Counterparty Extension violation strip + pricing-tier reconciliation to canonical 25 April 2026 structure + Lobito numeric/Derived-flag correctionsHallucination check executed at request as final discipline pass on v1.0.45 five-fix release before declaring complete. Counterparty Extension violations identified by full-build grep on the third-party name and associated programme/credential markers. Pricing inconsistency identified by cross-referencing new pricing page against canonical 25 April 2026 structure recorded in conversation history (afrimintel-pricing-page-copy.md artefact + Acquirer Brief rewrite same date). Lobito numeric/Derived-flag errors identified by re-reading the dossier funder table and matching against Harvard International Review Sept 2024 primary source which names $455M Angola + $100M DRC but not the residual figure. The discipline pattern: each correction is a primary-source verification chain follow-through, not an editorial preference adjustment.
Generated from the append-only audit log; full entries (with regression checks and editorial sign-off) at /audit-log/. Independence framework at /methodology/independence/. · Afrimintel v2.137.48