🪨 Mineral AI — Top 12
Daxos batch DD · 2026-07-14 · 100 scanned → 45 rated → 12 finalists → ROUND-2 ADVERSARIAL DD COMPLETE (4 agents, ~400 searches: LinkedIn, forums, patents, registries, physics) · 11 of 12 downgraded, 1 held
Round-2 revised board — outreach order
1
VOLUNA HELD ✓ only survivor
The Fernandini mining-dynasty link is REAL (Cerro de Pasco family; 8.1% of Sociedad Minera El Brocal). CTO is a genuine NASA Kennedy physicist (UW-Madison fusion PhD). Round-2 catch: airborne neutron assay at survey altitude is not physically credible; the real product is drone-delivered land-and-dwell near-surface geochem, still differentiated. Call first. Ask: which Fernandini, written physics specs, CMD IP split.
2
Hypermine TIME-SENSITIVE
The mystery major is ANGLO AMERICAN: Minas-Rio field demos → funded in-pit Sensing Pod PoC (published 7/2/26). Patents verified (CA/CL filings), chairman is ex-BofA nat-resources ML head. Catches: seed-PR CEO quietly gone within 12 months; "100x accuracy" has zero published accuracy-vs-assay data; MineSense/NextOre ahead on true bulk grading. Reach out before the Anglo PoC converts and prices the round.
3
Founders fully verify: Stanford geophysics PhDs, product implements their own published drill-VOI method. 6 named paying customers. Catches: all customers are TSXV/CSE micro-caps; the "repeat customer" is one renamed asset (Giant Mining → Copper One); "7/7 drill hits" = anomaly intersections, not grades. Likely raising within ~12 months. Ask: revenue split, share-payment exposure.
4
Real traction found in round 2: Blue Jay Gold 40,000m PAID deployment (6/30/26), Eldorado Québec Phase 2, ~8 Quebec clients in year one. Catches: GeologicAI raised $44M with BHP AND Rio for the identical mobile-scan model; no patents; ~half the capital stack is Quebec subsidy. Ask: paid-vs-pilot split, why-you-vs-GeologicAI.
5
Science is REAL: Nature Comms Earth & Env 2023 buried-kimberlite validation beat geochemistry; UBC Crowe Lab (2026 Killam Prize). Catches: validated on kimberlite only (n≈2), copper — BHP's actual ask — unproven; sequencing service not SaaS; all-academic team, no commercial CEO; co-developer died 2023. Re-rate trigger: a blind copper hit in Xplor.
Downgraded — watchlist
6
It's a pivoted carbon-removal company ("BlueShift Carbon" at Greentown; first revenue = CDR credits; licensed UMich patents are CO2-capture chemistry). Co-founder/CSO already left (now at Rewind). Boston Harbor pilot milestone unmet. Phoenix Tailings ($116M) is 10 miles away. ConocoPhillips equity is real. One call answers everything: pilot status, license field-of-use, the CSO exit.
7
Founder PhD is VU Amsterdam, not MIT; the borehole muon pilot was Schlumberger work/IP. "BHP Invent on cap table" = multi-sponsor cohort, not verified equity, no BHP deployment. Their own paper: 2 weeks of integration for one 2-D map on a 6.65m test heap — commercial heaps are 30-130m. Ideon is 5x funded and already selling. Owns US patent 11,971,373 + NSF SBIR I/II.
8
NOT vaporware — laser-plasma muon beams were just third-party validated (BELLA detection, Nature Physics 2025) and they hired B.S. Rao, author of the founding 2018 compact-muon-source paper. But every result so far needed a petawatt facility laser: realistic product is 6-10 years and €30-80M away. Põder still part-time at DESY; DESY IP terms unverified; two LinkedIn claims (€2.5M pre-seed, €1M contract) contradicted by every credible source. Re-rate on a written DESY license.
9
Credential rebrand. The UChicago PhD is real but in exoplanet climate modeling — zero exploration geophysics; "ex-NASA" = a 2015 undergrad internship. No uranium claims staked anywhere; hiring their first-ever geologist. Athabasca costs ~$400K/hole and Fleet Space is already deployed there with IsoEnergy. Integrity flag on the bio inflation.
10
Three pivots in three years; CEO seat quietly swapped; "HQ" is a residential address; 20-person "team" is 6 US + scattered offshore part-timers; the claimed "big contract with one of the largest miners" is unsized and unverifiable. Raising now — only re-engage if contract sizes go on paper.
11
The "$505K raise" is BHP Xplor's equity-free grant. "Ex-Rio Tinto founders" = an open-innovation program link, not employment. The corpus is free government data (WAMEX) and GeoGPT does it free. 5+ years old, no institutional round, one testimonial, and a client logo went into administration. Watch for a BHP earn-in at cohort end (~late 2026).
12
No field trial has happened — the first (2026) is gold-from-tailings, not the copper thesis. "Rio + BHP on cap table" = accelerator programs, not conviction equity. CEO is a sitting part-time UWA professor; no commercial operator on the team. Jetti is the cautionary tale ($2.5B valuation, Freeport quietly dropped it). Re-open on trial results + confirmed close.
Round-2 lessons (why 11 of 12 dropped)
Four systemic round-1 errors, now fixed in process: (1) grants and accelerator programs were counted as raises and cap-table conviction (BHP Xplor is equity-free; Founders Factory is a program); (2) founder pedigrees were taken from scan blurbs (a VU-Amsterdam PhD listed as MIT, an exoplanet PhD sold as exploration geophysics, "ex-Rio" that was a program membership); (3) pilot and traction language was unverified (milestones unmet, one renamed customer counted twice); (4) physics claims went unexamined (drone neutron assay at altitude, 4D muon imaging cadence, portable muon beams).
What survived: VOLUNA held because its strongest claims got STRONGER under scrutiny (the mining-dynasty family link verified, backers better than listed) even as the physics got haircut. Hypermine and LithologIQ gained real traction evidence (Anglo American PoC; a 40,000m paid deployment) even as their scores dropped on governance and competition.
Net for Daxos: five calls worth making, in order: VOLUNA, Hypermine (time-sensitive, pre-Anglo-conversion), ExploreTech, LithologIQ, Discovery Genomics. Nothing in this batch is a "wire it now" — every one has a specific question that gates a check, listed on its card.
Daxos Capital internal · generated 2026-07-14 · round-1 log: 2026-07-14-minerals-ai-45co-batch-dd.md · round-2 log: 2026-07-14-minerals-ai-round2-adversarial-dd.md