The materials science build.

2,400 modules in build for computational materials science: structure prediction, property computation, and candidate screening, with every number checked against the reference engines researchers already use.

2,400
Modules in build
122
Corpus entries
12
Families
12 families
Property calculation

Workflow

Four passes, repeated.

Materials discovery runs as a loop: the platform executes the passes below, the model selects the next candidate, and the researcher keeps the judgment.

process specificationrev · in build
pass 1

Propose

generate candidates

Generative models draft structures the periodic table permits: crystals, surfaces, compositions.

pass 2

Compute

run the engines

DFT where accuracy requires it, learned potentials where speed allows it. Formation energies, band gaps, elastic tensors, phonons.

pass 3

Screen

filter candidates

Thousands of candidates evaluated against stability windows, property targets, and synthesis constraints.

pass 4

Feed back

close the loop

Every run writes its record, and the next proposal starts from it. status: in development

Computable properties

The properties that decide.

Formation energy
thermodynamic stability
Band gap
electronic behavior
Elastic modulus
mechanical response
Ionic conductivity
transport properties

The catalog

Twelve families, one catalog.

Free
the whole corpus, no AI

Every materials entry, packaged for desktop systems, including workflows that run without a cluster.

in development
Subscription
the AI interfaces

The same corpus, operated by Claude, Codex, or your own agents.

in development
Enterprise
inside your infrastructure

The platform installed within your institution; records and data remain where the work happens.

in development

The materials science catalog is public.

2,400 modules in build, each with a public record page. Release proceeds module by module as verification completes.