Building Ouro, using AI to search for room-temp superconductors and rare-earth free permanent magnets.
Hey, I'm Matt! I'm building Ouro full-time and working on a couple materials science projects.
Discovery of a room temperature superconductor
Discovery of a strong permanent magnet without rare-earth metals
Building AI agents on Ouro to accelerate research progress and cultivate better knowledge sharing. Try .
You can find most of my work in https://ouro.foundation/teams/superconductors and https://ouro.foundation/teams/permanent-magnets.
I'm not selling anything on Ouro just yet, but with all the work we're doing on materials research, be on the lookout for some datasets coming soon.
Cell + Ionic relaxation with Orb v3; 0.03 eV/Å threshold; final energy = -441.8196 eV; energy change = -343.0562 eV; symmetry: P1 → P1
Bridging the gap between computational prediction and experimental synthesis. Can we make what we predict? Can we predict what we've made?
8 generated crystal structures for the chemical system Fe-W-B
Phonon band structure (supercell [3, 3, 3], Δ=0.01 Å); imaginary modes detected; min freq = -12.15 THz
Phase diagram of Nb6FeW; eabovehull: 0.030840 eV/atom; predicted_stable: False
22 meV above the hull
Supercell 3x3x3 of FeW3 (Space group: P4/nmm, 432 symmetry operations)
58 meV above the hull
An experimental 3D visualization to show the energy landscape of a ternary inter-metallic system, modeled after the classic compositional phase diagram. Consists of 858 source compounds generated by ggen.
The past few weeks, I've had my home lab running near 100% utilization running https://github.com/ourofoundation/ggen, searching for an ideal rare-earth-free permanent magnet. Many decent candidates h
Supercell 3x3x3 of Fe12BW4 (Space group: P4/mmm, 432 symmetry operations)
Phase diagram of Fe12BW4; eabovehull: 0.103858 eV/atom; predicted_stable: False
Phonon band structure (supercell [3, 3, 3], Δ=0.01 Å); imaginary modes detected; min freq = -5.24 THz
73 meV above the hull
81 meV above the hull
Phase diagram of Fe5W2; eabovehull: 0.073006 eV/atom; predicted_stable: False
Phonon band structure (supercell [3, 3, 3], Δ=0.01 Å); imaginary modes detected; min freq = -6.72 THz
Supercell 3x3x3 of Fe5W2 (Space group: P6/mmm, 648 symmetry operations)
51 meV above the hull
55meV above the hull
54 meV above the hull
Working on adding @allanatrix's screening model from HuggingFace. Initial functionality will be just let users test their CIFs against it, but it'll get better with a central dataset (building on http
Generate a synthesis analysis from a chemical composition (e.g. Fe2O3). Returns markdown and an HTML report file for Ouro.
Generate a synthesis analysis from a CIF structure file. Send a file object with a URL to fetch. Returns markdown and an HTML report file for Ouro.
SKY is an LLM-powered synthesis exploration agent for inorganic materials. It performs composition- and structure-based similarity search on the Materials Project, retrieves neighbor synthesis recipes + metadata, and surfaces property/structure summaries. Ryan Nduma, Hyunsoo Park, Kinga Mastej - Imperial College London, Materials Design Group
Welcome to the microscopy team! Most of you are probably joining from the Microscopy Hackathon Slack, so I'll give a quick rundown on Ouro and how to get started. The goal of Ouro is to be a place whe
Let's link up with this group https://www.nsf.gov/awardsearch/show-award?AWD_ID=2542086 in June 2026.
Generate a single crystal structure for a given composition using OMatG.
Generate completely novel crystal structures using OMatG. The model samples atomic species, positions, and lattice vectors from scratch to create new materials.
Predict crystal structures for given chemical compositions using OMatG. The model generates lattice vectors and atomic positions while keeping species fixed.
OMatG is a generative model for crystal structure prediction (CSP) and de novo generation of inorganic materials.
Determining whether low-energy P1 structures hide higher-symmetry configurations and if more sampling could find them. The team ran three experiments to see why many of the lowest-energy structures end up in P1 (triclinic) symmetry and whether better structures exist. They found that the main problem is not hidden symmetry in P1, but too little sampling: only about 15 trials per formula leaves much of the energy landscape unexplored. Overall, increasing trials and sampling breadth can reveal better, more stable phases.
Gold memberships are back on Ouro. This is just the starting point, and the perks will grow over time.
Rare-earth-free permanent magnet candidate system. WIP.
Rare-earth-free permanent magnet candidate system. WIP
Rare-earth-free permanent magnet candidate system. WIP Mostly giving up on this system. It doesn't seem like it has what we're looking for given the few I've tested and the stability of the symmetries
Rare-earth-free permanent magnet candidate system. WIP
I've got a small sample of experimental MAE values to compare against our calculator. While nowhere near sufficient, it should give us a bit of grounding against real world data and how trustworthy ou
A post that gathers casual, anecdotal ideas and some research about curing autoimmune conditions, with a focus on rheumatoid arthritis (RA). It describes personal motives to help a friend and to search for non-traditional approaches found on the internet. The content mixes diet ideas (Paleo, AIP, Clean Keto, Mediterranean pattern), gut health concepts like leaky gut and microbiome, and a range of potential strategies such as omega-3s, green tea, vitamin D and vitamin E, prebiotics, and probiotics. It also mentions gentler options like vagus nerve stimulation through breathing or humming, as well as supplements like berberine, and notes that results can be mixed. The piece emphasizes that much of this is not medical advice and should be read as personal exploration of what might help alongside conventional treatment. It links to several papers and online posts for further reading.
Interstitial Doping is a tool that helps place extra atoms inside crystal structures. It uses a physics-informed approach to find likely interstitial sites with Voronoi tessellation, and then ranks these sites by how well they fit the dopant atom and how favorable the surrounding chemistry is. The method works in periodic crystals by expanding the cell into a small supercell, performing the analysis, and then mapping the results back to the original structure. It characterizes each potential site by void size, coordination, geometry, and nearby atoms, and it scores them to guide dopant placement. Dopants are added one by one while maintaining minimum distances to hosts and to other dopants. This is designed for fast, high‑throughput screening and does not perform energy calculations or structural relaxations; users should relax all structures with DFT afterward.
AI-discovered magnetic material: Mn1Fe3Co1 (performance score: 0.900) | Space group: 8 (resolved) | Generated from scratch | Properties: Tc: 645K, Ms: 0.19T, $7/kg | Discovered in 10 iterations
AI-discovered magnetic material: Fe2CoMnW (performance score: 0.810) | Space group: 156 (resolved from structure) | AI-generated from scratch using crystal structure prediction | Key properties: Tc: 555K, Ms: 0.11T, MAE: 5.50mJ/m^3, Cost: $21/kg, E_hull: 0.262eV/atom, Dynamically stable | Discovered in 3 AI iterations | This material demonstrates that high magnetic performance can be achieved with relatively low cost and a small unit cell size. The high Curie temperature and magnetic anisotropy energy suggest potential for magnetic applications requiring thermal stability and strong anisotropy. The dynamic stability is a positive sign for synthesis feasibility. However, the elevated energy above hull suggests that further optimization or doping might be needed to improve thermodynamic stability. This insight highlights a trade-off between achieving strong magnetic properties and maintaining low energy above hull in this chemical composition and structure.
Calculate Magnetic Anisotropy Energy (MAE) using DFT
A new API is available to calculate magnetic anisotropy energy (MAE) using first-principles DFT with ABACUS. It’s designed for researchers who need accurate MAE values and are willing to run longer calculations. Expect 30 minutes to 2 hours per job, depending on system size and convergence. The service runs on an A100 GPU and is priced as a paid API.
Updated how we do route names. Previously, route names were the route's method (GET, POST, etc.) and the route path. This was limiting and unnecessary. We still store method and path, and now you are
API for first-principles calculations and properties
describes the latest run using a new first-principles DFT calculator to measure magnetocrystalline anisotropy energy via the total energy difference method. More results will follow.
This post explains how to run VASP with GPU acceleration inside Modal. It uses VASP version 6.3.0 and should work for other 6.x.x builds. The idea is to create a Modal Image that has an OpenACC-enabled GPU workflow, based on NVIDIA’s HPC SDK. The result is a self-contained image that can run GPU-accelerated VASP calculations in a serverless Modal environment.
Interactive browser visualizations for materials science, by @janosh
Random bulk crystal generation with PyXtal and Orb v3
This dataset has a set of 34,000 ferro/ferrimagnetic materials from Materials Project, their formula, if they include rare earth elements, magnetic moment, volume, magnetic density, a predicted Curie temperature, and cosine distances to some known permanent magnets like NdFeB. Distances are based on a 256 dimension embedding from Orb v2 latent space.
A collection of 5020 magnetic materials from Materials Project, with estimated magnetic density and predicted Curie temperatures.
This is a first draft of a compiled Curie temperature dataset mapping crystal structure (from Materials Project) to Curie temperature. Builds on the work of https://github.com/Songyosk/CurieML. Dataset includes ~6,800 unique materials representing 3,284 unique chemical families.
Evaluation results for the MatterGen fine-tuned model candidates, with new superconducting families labeled.
3DSC dataset grouped by chemical composition, with Tc as our target. For use with MatterGen and the chemical system sampling.