About ORG-2
Models set the ceiling. Organization decides how close you get.
ORG-2 is named after a conviction: the AI race is not merely a technological competition — it's a question of organizational sociology. How groups divide labor, build trust, create accountability, and decide together. Agents don't dissolve those questions. They make them urgent, because now some of the actors aren't human.
The thesis
The teams that ship the best AI products are not the ones with the best models. They are the ones that figured out how to work with AI as a team — how to delegate, how to review, how to keep shared context, and how to keep humans in the loop on the decisions that matter.
Accountability — who did what, on whose authority, to what end — is precisely what organizational sociology studies. An agent you cannot audit is an agent you cannot trust. So we build the record: what was asked, what the agent understood, what it did, and whether it delivered.
The origin
It started as an experiment. Our own agents outran our tools — work scattered across Cursor, Claude Code, Codex, GitHub, and Linear, none of which saw each other, and pull requests became unreviewable: too much code, arriving too fast for anyone to keep track of. So we built our own harness and shared agent memory to survive our own workflow.
Then it stopped looking like a tooling problem. Today ORG-2 is built inside ORG-2 — every session recorded, replayed, and reviewed in the product itself, including the fleets of subagents that modularize its own codebase.
Founders
Trained as an organizational sociologist at Swarthmore and Harvard, studying systems that tried to make work legible at scale — scientific management, innovation theories, control theory in the Soviet Bloc, the transnational history of databases. At 20, working alone, built an AI customer-service backed flight OTA during COVID that did $500k in bookings in 30 days and got 300+ stranded students home.
@harry-he →Six years in security and data-provenance organizations at Microsoft and Duolingo, building audit trails and systems of record for large enterprises — the same primitives ORG-2 rests on. He knows what organizations accept as proof, and how they buy it.
@neonforge98 →Who we build with
ORG-2 is open source, and the record is shaped by the people inside it — engineers at top AI labs, indie builders, and teams we haven't met yet. Anyone can join: the repo is open, the replays are shared, and the best ideas in the product started as someone's session.