The book
Internet-native coordination
How to DAO is a Penguin Random House publication on how decentralized organizations coordinate and govern.
Autonomous organization consulting
Your agents are running. Your organization isn't built for them yet.
We help executive teams redesign decision rights, oversight, and organizational context so AI agents can do useful work without making accountability harder.
Self-serve starting point
See where your operating model is ready—and where it is still implicit.
Answer twelve questions about decision rights, organizational context, oversight, and workflow readiness. Your answers and score stay in this browser unless you choose to send the result with an inquiry.
Take the readiness diagnosticThe engagement path
Make the operating model explicit before scaling agents.
Start with a bounded organizational question, turn it into a design leadership can evaluate, and test one workflow only when the conditions are right.
01 · Diagnostic
See the current system
Map workflows, context flows, decision rights, agent-ready opportunities, and governance risks.
Explore the diagnostic →02 · Design
Define the operating model
Set ownership, agent boundaries, human review, escalation paths, and rollout sequence.
Explore operating-model design →03 · Follow-on
Test a contained pilot
Use agreed success criteria to decide whether one bounded workflow should stop, change, or expand.
Explore contained pilots →Why this practice
A published foundation, a named principal, and public research.
The principal
Jan “Puncar” Brezina
Puncar co-authored the book and leads the practice. He also builds Corgtex, the toolchain used in contained pilot work.
Meet the principal →The research
Claims linked to sources
Our research note examines published operating-model changes and the governance questions they create.
Read the research →Proof posture: no client names or case studies are currently published here. Public company examples are illustrations, not client claims or endorsements.
The shift in public
Companies built on programmable coordination are documenting the change.
Block and Coinbase are public illustrations. They are not How to DAO clients, case studies, or endorsements.
Block
“Hierarchy was just an information router.”
Dorsey and Botha argue that hierarchy solved an information-flow constraint. Their alternative uses a company world model, an intelligence layer, DRIs, and player-coaches. Agents handle more coordination while people remain responsible for the work.
Read Block's thesis →Coinbase
“No pure managers. AI-native pods.”
Armstrong's memo describes fewer layers, no managers who only manage, AI-native pods, and individuals running fleets of agents. Coinbase engineering also documents the infrastructure, observability, and auditability required.
Read Armstrong's memo →Latest article
The one-person company is not the future.
A short opinion piece on one-person companies, AI agents, and why human connection still matters in AI-native organizations.
Read the article →Start with one operating problem.
Bring the workflow, decision, or accountability question that becomes harder as agents take on more work.
FAQ
Straight answers
What does How to DAO actually do?
We help executive teams design AI-native operating models — the governance, decision rights, and coordination structures that make agents productive rather than just present.
What is Corgtex?
Corgtex is the platform we built and use to do the work — mapping organizations, testing agent workflows, and running pilots. You hire us, not software.
Are Block and Coinbase your clients?
No. They are public companies documenting this shift in real time. We reference their work because it is the clearest proof that this reorganization is happening now.
How do engagements start?
With a working session about one operating problem. If there is a fit, the first formal step is usually a scoped diagnostic with an agreed fee and decision-ready outputs.
Who does the work?
Work is led by Jan “Puncar” Brezina. Any additional contributors and their roles are agreed before they join an engagement.
What if we are not ready for a pilot?
A pilot is never the first step by default. The diagnostic and operating-model work can stand on their own and may show that a pilot should wait.