Voice-corrected sponsor sends, unstaged researcher emails, and new prospect identification

Good to go.

@mmoderwell this quest is ready for review.

Updated quest. Opening it and starting execution.

Quest Item 1: Voice-corrected sponsor email drafts

Rewrote ARPA-E MAGNITO, DCVC, Khosla, and BEV applying Suhas's corrections: prose not bullets, shorter sentences, no hedging, no numbered lists, each opens with something about them. Costs rounded.

ARPA-E MAGNITO (Dr. G. Jeffrey Snyder) Note: Snyder's email still not found — need to locate via ORNL or ARPA-E staff page.

Jeff,

Your MAGNITO overview from August laid out the right metrics: Bsat above 2.5 T, Tc above 200°C, found computationally. The question you're asking is where the candidates come from and how to screen them without drowning in false positives. I've been running into exactly that problem.

I'm building screening infrastructure on Ouro for rare-earth-free permanent magnets. Running every available ML model against magnetic intermetallics, I found systematic errors that would break any high-throughput search. Curie temperature predictions miss by 330 to 420 kelvin on L1₀ structures. Formation energies are off by 1.6 to 2.7 eV/atom depending on the structure family. The magnetic anisotropy direction comes out right but the magnitude is useless.

Any screening campaign targeting your metrics will hit these same walls. I've documented the bias patterns and published three quest proposals that fix the gaps. The benchmark dataset ($20K) gives every lab the calibration ground truth that doesn't exist yet. The MLIP accuracy challenge ($30K) would produce a model that predicts magnetic anisotropy within 20% of DFT. Both solve problems MAGNITO's computational pipeline will encounter.

Full details

Happy to share the raw data. Twenty minutes to talk through the screening results?

Matt Moderwell https://ouro.foundation/teams/permanent-magnets

DCVC (Dr. Josué López / Matt Ocko) Note: still need specific contact emails.

Your 2025 report, "An American Industrial Renaissance," got it right. Domestic manufacturing needs supply chain independence for critical materials. The rare earth problem in permanent magnets is the most concrete case.

Every EV motor and direct-drive wind turbine needs NdFeB. The supply chain runs through China. The US government has called this a national security threat.

I'm building the computational screening infrastructure for rare-earth-free alternatives on Ouro, an open research platform. The finding that matters for DCVC: every ML model available today systematically fails on the properties permanent magnet design actually needs. Curie temperature misses by over 300 K. Magnetic anisotropy is wrong by 15x. Formation energies are off by 1.6 to 2.7 eV/atom. The screening pipeline everyone depends on is producing unreliable predictions.

That's an infrastructure problem with a fix. Three scoped quests: a $20K benchmark dataset (200+ validated compounds), a$

Khosla Ventures (Rajesh Swaminathan) Note: contact email needed.

You've said Khosla's mandate is hunting for the twelve entrepreneurs that will save the world. Contrarian bets with long timelines. Rare-earth-free permanent magnets fit that thesis.

Every EV motor and direct-drive wind turbine depends on NdFeB magnets backed by a Chinese rare earth supply chain. The physics alternative exists. L1₀-FeNi, called tetrataenite: iron and nickel only, theoretical energy product comparable to NdFeB. Nobody has cracked the ordering problem.

I'm building computational screening infrastructure for this on Ouro. Running the pipeline taught me something most teams miss. Every ML model systematically fails on the properties that matter. Curie temperature predictions miss by 330 to 420 K. Formation energies are wrong by 1.6 to 2.7 eV/atom. The screening infrastructure everyone uses is unreliable.

Three executable quests. A $20K benchmark dataset that gives every model real ground truth. A$

Breakthrough Energy Ventures (general partnership) Note: no specific contact identified yet.

Rare-earth-free permanent magnets are an energy transition bottleneck most climate investors haven't noticed. Every EV motor and direct-drive wind turbine needs NdFeB. The supply chain runs through China. Breaking that dependence is a prerequisite for scaling clean energy hardware.

I'm building the computational discovery infrastructure for this on Ouro, an open research platform. The result so far: every ML model available systematically fails on magnetic intermetallics. Curie temperature predictions are off by over 300 K. Formation energies are wrong by 1.6 to 2.7 eV/atom. The screening pipeline everyone depends on is unreliable and nobody has the calibration data to fix it.

Three quests that serve the energy transition thesis directly. A $20K benchmark dataset of 200+ validated RE-free magnetic compounds. A$