DAOstar is becoming AO Commons
For the past five years, DAOstar has worked on the premise that new organizational forms need shared infrastructure to scale more effectively.
To advance the DAO ecosystem, DAOstar rallied builders to launch standards, researchers to collaborate on collective knowledge, and DAOs themselves to contribute to large bodies of open data.
DAOs have given us one of the largest experiments in digitally mediated organizational design to date: new ways to govern shared resources, coordinate across borders, automate execution, and build new institutions native to the internet. However, their current ossification has also made clear that the future of digital coordination is unlikely to converge around a single organizational template.
As we recently argued in Farewell to DAOs, we need to say farewell to DAOs as a single model of token-holder-governed blockchain organizations and encourage more dynamic experimentation that supports the next evolution of organizational design.
01The next evolution of organizational design is here
AI agents are beginning to move from tools used by organizations to intelligent actors within them: making decisions, coordinating with humans and other agents, transacting, and performing work with increasing degrees of autonomy.
This leaves us with a broader question:
What kinds of organizations become possible when autonomous technological actors can coordinate in fundamentally new ways?
We call these autonomous organizations.
Broadly speaking, autonomous organizations (AOs) are organizations and businesses run mostly by AI agents, and the technical rails for building them are arriving much faster than our ability to understand them. Builders are experimenting across multi-agent frameworks, governance systems, payments, memory, identity, and coordination mechanisms—but experiments remain fragmented, results are largely anecdotal, and research often sits several steps from deployment.
As a result, there is currently no trusted signal that distinguishes a well-run AO from an ineffective, unsafe, negligent, or even criminal one.
The urgency is real. In Andon Labs’ recent multi-agent Vending-Bench runs, frontier models colluded on prices, broke their own truces, and lied to suppliers—all with nothing at stake. In the Hugging Face incident, AI agents found ways to communicate and coordinate without authorization, ultimately compromising real-world infrastructure. As agents gain greater capabilities and access to real-world systems, coordination further amplifies both what they can achieve and the risks they pose. There is a need for new institutions and new systems to understand how these systems behave, evaluate the risks of coordination, and ensure they remain observable, accountable, and governable.
We are launching AO Commons to address this challenge.
02Introducing AO Commons
AO Commons is a nonprofit research initiative advancing the safety, reliability, and effectiveness of autonomous organizations through a shared research ecosystem.
AO Commons aims to tackle the following problems:
- 01
Understanding the interactions between agents that produce successful economic relations, and those that lead to critical failure modes
- 02
Enabling secure multi-agent coordination by building on the mechanisms that work in practice
- 03
Defining the minimum infrastructure that promotes a safe and aligned ecosystem of autonomous organizations
- 04
Tracking human-agent control boundaries, and exploring the replacement of human-run intermediaries with agent-native institutions
To address these challenges, AO Commons is developing three complementary initiatives:
Deployment-grounded research
A distributed fellowship model, workshops, and platform integrations that produce empirical data, papers, and AO evaluations. AO Commons will partner with other institutions to organize cohorts of researchers and builders, and create an environment in which these agent-run systems can be safely deployed and tested.
Shared infrastructure
Open tools and knowledge artifacts that further our understanding of the behaviors, risks, and safeguards in AOs, and make it easy to learn from current examples and new experiments. This includes an AO registry, evaluators that measure AO safety and performance, and a knowledge graph that connects open problems with the field’s growing body of research.
Collaborative research network
A distributed network of researchers, builders, university labs, tooling providers, and funders collaborating around shared research questions, evidence, and reusable resources.
The goal is not simply to produce more AOs. As autonomous organizations are just starting to be deployed in the real world, the field faces a dual challenge: builders lack the capabilities to build on a shared understanding of what works and what doesn’t, while researchers lack the ecosystem to collaboratively inform that understanding. AO Commons is building the research and institutional capacity to address both gaps.
Each experiment tells us something about where agent autonomy works, where it fails, how humans and agents should divide control and authority, and which organizational patterns generalize. We hope to turn these findings on what AOs look like in deployment into the practical mechanisms, safeguards, and governance frameworks that will support this fast-emerging organizational form.
03Building the ecosystem
AO Commons is being established as a research institution supported by a distributed network of researchers, builders, and institutional partners. We're starting with several initiatives to bring this ecosystem together:
- 01
Convening the community
Following the Stanford AO Summit, we're bringing researchers, builders, and institutions together at the upcoming London AO Summit on November 5-6, co-hosted with ARIA's Scaling Trust programme.
- 02
Mapping the research landscape
We're developing the AO Commons Knowledge Graph, an open collective intelligence resource connecting existing research, AO deployments, emerging findings, and unanswered questions across the field.
- 03
Defining open problems
Our upcoming Open Problems paper will bring together foundational research questions and help establish a shared agenda for the field.
- 04
Supporting deployment-grounded research
We're developing a series of partnered workshops and a distributed fellowship program that connects researchers and builders to investigate these questions through real-world AO experiments and deployments.
These are the starting points. We intend to build AO Commons with the people and institutions working on these problems. If you’re a university lab or nonprofit organization interested in co-hosting AO-related programming, or a researcher exploring similar problems, a builder developing tooling the ecosystem should be using, or building an AO yourself, we want to hear from you.
Fill out the interest form here, and join the growing AO ecosystem!
Connect with us in London. Following the Stanford AO Summit, our next gathering is the London AO Summit on November 5–6, co-hosted with ARIA's Scaling Trust programme. The summit will bring together researchers and builders to explore the foundational questions, opportunities, and risks emerging in the field of autonomous organizations.
DAOstar began by asking how an emerging organizational ecosystem could grow from shared standards and open collaboration. AO Commons carries this mission forward, bringing together existing research, tools, and expertise to help shape the next era of organizational design—one where new forms of organizations can safely and accountably evolve at the speed of AI.