One Person Lab
A local AI workbench for starting expert work, activating domain agents, tracking progress, and collecting research, grant, and presentation deliverables from one place.
Overview
One Person Lab is the lab’s local AI workbench for serious long-running expert work. It gives users one place to start a task, attach a real workspace, activate the right domain agent, keep progress visible, and collect deliverables such as manuscripts, proposal drafts, slide decks, review materials, and project files.
The current public entry is intentionally practical: users can get the macOS App from the GitHub Releases page, or run a one-command terminal installer. The workbench prepares the local environment, connects Codex-facing skills and modules, and keeps the specialized product families visible without forcing every task through the same workflow.
Current domain readiness
| Product family | Current agent | Readiness |
|---|---|---|
| Research Foundry | Med Auto Science (MAS) | Comparatively mature for medical research workflows, evidence organization, manuscript-facing delivery, and auditable study progression. |
| Presentation Foundry | RedCube AI (RCA) | Comparatively mature for presentation and visual-deliverable workflows, including slide decks, scripts, and review loops. |
| Grant Foundry | Med Auto Grant (MAG) | Technical prototype for grant authoring and revision workflows; useful for exploration, but not yet at the maturity level of MAS or RCA. |
Install
- macOS App: use the GitHub Releases page for the current package.
- Terminal path: run the one-command installer.
The project is now suitable for internal beta testing. We welcome researchers and students to try the workbench, especially for MAS-backed medical research and RCA-backed presentation work, while treating MAG as an early-stage technical prototype.