Infrastructure for
AI-native biology

Tested, registered analysis libraries that AI agents look up and call.

Explore the verses View on GitHub

mouse embryo · E8.5 · 3D spatial16,989 cells

omicverseBiologyExplore →

generative design · denoise → relax → hull2 candidates

matverseMaterialsExplore →

socialverseSocial scienceExplore →

many-body / spin glass

import physverse as pv

pv.utils.registry_lookup("spin glass ground state")

H = pv.quantum.heisenberg(pv.quantum.lattice_from_bonds("hamiltonian.json"))

rbm = pv.quantum.init_rbm(H.n_sites, 2 * H.n_sites)

res = pv.quantum.vmc_optimize(rbm, H, metric_samples=16384)

m = pv.opt.read_ising_json("instance_0.json")

cb = pv.opt.cluster_lower_bound(m, cells)

gs = pv.opt.parallel_tempering(m, ladder, target=cb.bound)

fit = pv.pde.select_model(lib.Theta, ut, lib.names)

physversePhysicsPrivate preview

exoplanets / transit search

import astroverse as astv

astv.utils.registry_lookup("transit search") # find, do not guess

grid = astv.ts.transit_period_grid(t, minimum_period=1.0, maximum_period=10.0)

res = astv.ts.bls(t, flat, dy, grid)

vet = astv.ts.vet_transit(t, flat, dy, res.best_period, res.best_t0, res.best_duration)

d = astv.cosmo.theory_3x2pt(params, lay, zs, nzs, zl, nzl)

chain = astv.infer.run_emcee(like.log_prob, p0, n_steps=5000)

astv.infer.convergence_report(chain) # R-hat, ESS, tau

astroverseAstronomyPrivate preview

One substrate

Cells, crystals, light curves. Same object.

Every verse keeps its data on AnnData. Observations, annotations, descriptors and results travel together and write to h5ad or zarr, so nothing gets lost when you merge, filter or featurise twice.

Sciences

Data object

h5ad · zarr

Open formats

cells · mouse embryo E8.5one AnnData

.Xmeasurements

.obscells · materials · sources · units

.vargenes · elements · bands · items

.obsmembeddings · structures · light curves

.unsprovenance · run records

Capability lives in the library

Agents already read papers and write code well. What they lack is a trusted, queryable toolbox for each science.

registry_lookup("transit search")

Registered for lookup

Every public function says what it does and what it needs, so an agent asks before it writes code.

assert close(ours, reference)

Tested against a reference

Each function is checked against an independent implementation or a known result.

tutorials/*.ipynb executed

Executed in a tutorial

Every function appears in a tutorial that actually runs.

held-out task → new function

Grown from failures

When an agent fails a held-out research task, the missing method goes into the library.

Built in the open, one science at a time

github.com/scienceinfra

Infrastructure forAI-native biology