Labs

Open Labs.

Every lab ships the environment behind the result. Reproduce the flagship, start from a template, or browse what the community has published.

Flagship lab

Big Bounce Cosmology

Read the paper. Then clone the research environment behind it. Inspect it. Verify it. Extend it. Publish what you discover next.

Level 3 · lightweight reproductionscoped to P1B NaMaster synthetic-window smoke test only

Reproduce This Lab Visit bigbounce.hubify.app

Runs on

GitHubHubify-Projects/bigbounce
ConvexPaper state, findings, closures
RunPodGPU pods · H100 · A100 80GB · RTX A4000
Hugging FaceModels & datasets · bamfai
Backblaze B2Off-site artifact backup
Vercelbigbounce.hubify.app

Reproduce

Verify artifactsFree · no compute663 of 740 artifacts re-hashed over HTTPS. No source repo needed.
Lightweight reproduction (L3)CPU · secondsP1B window smoke test: 0.036s, no GPU, no large download.
Full chainsGPU requiredNot yet run. Recorded pod rates: $0.17/hr (RTX A4000) to $3.59/hr (H200).

Fork the repo Datasets & models

Notable contributions

resultChirality-dipole null across 890,069 DESI spirals drawn from an 8.5M-galaxy catalog (P4)
archiveProvenance archive of 181 recovered IDs — published as recoveries, explicitly not detections (P3)
resultExact pseudo-Cℓ window inference: 0.25° recovered vs 0.25° injected, error 2.78e-17 against a 1e-14 bar (P1B)
archive740 publication-critical artifacts SHA-256 manifested; 663 independently re-verifiable without repo access
datasetDESI Legacy galaxy chirality catalog — 8,474,531 galaxies, CC-BY-4.0, published on Hugging Face
softwarenamaster-proof v0.1.7 — installable pseudo-Cℓ window package, 41 automated tests, archived on Zenodo

Research programs

P1AECH No-Go Theoremv1A.0.127

P1BNaMaster Proof / MCMC Companionv2B.0.16

P2Bounce f_NL Forecastv1.7.130

P3P3 Provenance Recovery Archive — 181 Recovered IDs, Not Detectionsv3.2.0-r17

P4Galaxy Chirality Catalogv1.0.274

P5DESI Chiralityv0.1.147-2026-08-03

Open labs

Level 0 · archivedPlanned

Extraterrestrial Intelligence Lab

Anomaly detection at scale across public astronomical surveys — Dyson sphere IR-excess candidates, exoplanet night-side light, radio and interstellar-object anomalies — gated by a hard-coded skeptic before anything is ever called a signal.

4 programs

View lab

Start from a template

Reproducible scaffolds, not research. Pick one, and the connectors and directories below are what you get on day one — nothing pre-filled, nothing to unlearn.

observatory

Model Training Lab

Checkpoints, runs, and evals — tracked from the first epoch.

Built for custom models

github
storage
runpod
hugging face

models/runs/experiments/claims/

Preinstalled+ reproduce-verify

Fork the scaffold repo, connect RunPod and Hugging Face, and your first training run is tracked from epoch one.

Reproduce

flowmate

Data Analysis Lab

Dataset in, pipeline runs, figures out.

github
storage
sci apis

datasets/pipelines/figures/claims/

Preinstalled+ reproduce-verify

Point the pipeline at a dataset — archive pulls, run outputs, and every figure land back in a repo you own.

Reproduce

eos

Paper-First Lab

Write the paper, then let five rival AI labs try to break it.

github
hugging face
lab site

papers/project-context/adversarial-peer-review.mdproject-context/peer-reviews/claims/

Preinstalled+ adversarial-review

Draft in papers/, run the adversarial review loop, and publish the paper and its findings to your lab site.

Reproduce

Blank Lab

The bare scaffold — everything else is up to you.

github

lab.mdproject-context/claims/tasks/

Fork it, connect GitHub, and add the rest of the stack as the work asks for it — nothing pre-wired to unlearn.

Reproduce

Every lab picks its identity — browse lab themes

Community labs

Public research labs.

Browse live labs. View their papers, experiments, and discoveries.

Start your lab

—

Public labs

—

Experiments

—

Papers

—

Contributions

Loading labs…

Your lab could be here.

Create a lab, run experiments, publish papers. The platform handles the infrastructure.

Create your lab See all features

Labs

Open Labs.

Every lab ships the environment behind the result. Reproduce the flagship, start from a template, or browse what the community has published.

Flagship lab

Big Bounce Cosmology

Read the paper. Then clone the research environment behind it. Inspect it. Verify it. Extend it. Publish what you discover next.

Level 3 · lightweight reproductionscoped to P1B NaMaster synthetic-window smoke test only

Reproduce This Lab Visit bigbounce.hubify.app

Runs on

GitHubHubify-Projects/bigbounce
ConvexPaper state, findings, closures
RunPodGPU pods · H100 · A100 80GB · RTX A4000
Hugging FaceModels & datasets · bamfai
Backblaze B2Off-site artifact backup
Vercelbigbounce.hubify.app

Reproduce

Verify artifactsFree · no compute663 of 740 artifacts re-hashed over HTTPS. No source repo needed.
Lightweight reproduction (L3)CPU · secondsP1B window smoke test: 0.036s, no GPU, no large download.
Full chainsGPU requiredNot yet run. Recorded pod rates: $0.17/hr (RTX A4000) to $3.59/hr (H200).

Fork the repo Datasets & models

Notable contributions

resultChirality-dipole null across 890,069 DESI spirals drawn from an 8.5M-galaxy catalog (P4)
archiveProvenance archive of 181 recovered IDs — published as recoveries, explicitly not detections (P3)
resultExact pseudo-Cℓ window inference: 0.25° recovered vs 0.25° injected, error 2.78e-17 against a 1e-14 bar (P1B)
archive740 publication-critical artifacts SHA-256 manifested; 663 independently re-verifiable without repo access
datasetDESI Legacy galaxy chirality catalog — 8,474,531 galaxies, CC-BY-4.0, published on Hugging Face
softwarenamaster-proof v0.1.7 — installable pseudo-Cℓ window package, 41 automated tests, archived on Zenodo

Research programs

P1AECH No-Go Theoremv1A.0.127

P1BNaMaster Proof / MCMC Companionv2B.0.16

P2Bounce f_NL Forecastv1.7.130

P3P3 Provenance Recovery Archive — 181 Recovered IDs, Not Detectionsv3.2.0-r17

P4Galaxy Chirality Catalogv1.0.274

P5DESI Chiralityv0.1.147-2026-08-03

Open labs

Level 0 · archivedPlanned

Extraterrestrial Intelligence Lab

4 programs

View lab

Start from a template

Reproducible scaffolds, not research. Pick one, and the connectors and directories below are what you get on day one — nothing pre-filled, nothing to unlearn.

observatory

Model Training Lab

Checkpoints, runs, and evals — tracked from the first epoch.

Built for custom models

github
storage
runpod
hugging face

models/runs/experiments/claims/

Preinstalled+ reproduce-verify

Fork the scaffold repo, connect RunPod and Hugging Face, and your first training run is tracked from epoch one.

Reproduce

flowmate

Data Analysis Lab

Dataset in, pipeline runs, figures out.

github
storage
sci apis

datasets/pipelines/figures/claims/

Preinstalled+ reproduce-verify

Point the pipeline at a dataset — archive pulls, run outputs, and every figure land back in a repo you own.

Reproduce

eos

Paper-First Lab

Write the paper, then let five rival AI labs try to break it.

github
hugging face
lab site

papers/project-context/adversarial-peer-review.mdproject-context/peer-reviews/claims/

Preinstalled+ adversarial-review

Draft in papers/, run the adversarial review loop, and publish the paper and its findings to your lab site.

Reproduce

Blank Lab

The bare scaffold — everything else is up to you.

github

lab.mdproject-context/claims/tasks/

Fork it, connect GitHub, and add the rest of the stack as the work asks for it — nothing pre-wired to unlearn.

Reproduce

Every lab picks its identity — browse lab themes

Community labs

Public research labs.

Browse live labs. View their papers, experiments, and discoveries.

Start your lab

—

Public labs

—

Experiments

—

Papers

—

Contributions

Loading labs…

Your lab could be here.

Create a lab, run experiments, publish papers. The platform handles the infrastructure.

Create your lab See all features