From AI answers to finished expert work

One Person Lab App is the lab’s published desktop workbench for serious, long-running knowledge work. It gives users one place to start from a real goal, attach a workspace, enter the right Foundry line, follow progress, review evidence, and collect deliverables such as manuscripts, proposals, slide decks, books, reviews, and project files.

The technology is built around two reinforcing advantages.

1. A foundation for high-value knowledge delivery

OPL organizes open-ended expert judgment into visible stages with real artifacts, evidence, review, recovery, and clear human decisions. The shared Framework provides the reliable runtime and delivery model; the App makes that work legible and actionable; each Foundry Agent retains the standards and judgment of its domain.

MAS, MAG, RCA, and OBF are working examples across research, grants, visual communication, and long-form publishing. They demonstrate that the same foundation can support different forms of high-value knowledge delivery without flattening their professional standards into one generic workflow.

2. A meta-agent and self-evolving agent foundry

OPL Meta Agent (OMA) is the agent that develops agents. It turns a create, takeover, or improvement objective into a reviewable Agent blueprint, evaluation specification, and evidence-bound evolution proposal. The OPL Foundry Kernel then materializes candidates, runs evaluations, versions accepted results, and controls activation or rollback.

This separation of design judgment from controlled evaluation creates a compounding platform: evidence from existing agents can improve the next design, while reusable infrastructure lowers the time and marginal cost of creating each new specialist.

Five standard agents

The official OPL profile currently provides five standard agent entry points. They are defaults, not a hard-coded ceiling: the App discovers compatible agent packages dynamically.

Agent Role What it delivers
OPL Meta Agent (OMA) Meta-agent and Agent Foundry Designs, takes over, diagnoses, and evolves other OPL-compatible agents.
Med Auto Science (MAS) Research agent Evidence organization, analysis, manuscript preparation, review, and auditable study progression.
Med Auto Grant (MAG) Grant agent Grant direction setting, proposal development, simulated review, and revision.
RedCube AI (RCA) Visual-deliverable agent Presentations, reports, defenses, narrative design, rendering, review, and export.
OPL Book Forge (OBF) Book agent Planning, drafting, reviewing, and delivering books and long-form manuscripts.

Whitepapers

These four public whitepapers provide the shortest path from the overall idea to the product, cloud architecture, and a fully developed domain-agent example.

Whitepaper Focus Read online
One Person Lab Whitepaper Why complex knowledge work needs expert stages, evidence, review, and recoverable delivery. HTML
OPL App Whitepaper How the local-first workbench turns goals, materials, progress, artifacts, and decisions into one continuous user experience. HTML
OPL Cloud Whitepaper How the same work can extend to online workspaces, governed resources, collaboration, and Agent services. HTML
MAS Agent Whitepaper How the OPL model is applied to real medical-research work and publication-oriented delivery. HTML

Install One Person Lab App

Use the current Latest Stable release rather than a historical release or version-pinned asset. The maintained guides follow the current release channel:

OPL Flow and OPL Fleet

OPL Flow is the Codex experience baseline and durable work-coordination layer. It keeps ownership, recovery, repository integration, and verifiable closeout coherent when work spans multiple tasks, conversations, and repositories.

OPL Fleet extends that coordination across machines. It is an Agent-native distributed execution and continuity system for durable objectives, compatible workspaces, fresh node admission, protected capacity, task dispatch, checkpoints, and owner-safe continuation. It complements Codex, SSH, GitHub runners, HPC, cloud, and container systems rather than replacing their strongest execution primitives.

Flow and Fleet are already used in regular and power-user workflows where one person is coordinating several AI tasks or machines at once. They are not required for basic App use; users can adopt them when the scale and continuity of the work justify the additional control layer.