A new wave of governance experiments is emerging across the Web3 ecosystem.
Interestingly, it does not seem to be moving toward “more governance”, but precisely the opposite. I would say that the essence of the new wave is to reduce the scope of governance by embedding decisions directly into the workflow. Embedded governance.
For years, DAOs have often treated governance as the main layer: proposals, forums, votes, delegates, committees, councils, and treasury processes. These mechanisms matter, but they create overhead quickly. The more decisions move through explicit governance, the more time communities spend deciding what should happen instead of making things happen.
The Minimum Viable Institution starts from a different premise: What would be the minimum governance infrastructure an institution would need to function and sustain itself indefinitely within a complex system?
If governance can be built into the workflow itself through clear constraints, explicit agreements, role-based authority, verification loops, and institutional memory, then we can streamline the initial stages of coordination in order to focus on what we set out to build in the first place, when we decided to come together.
Instead of asking delegates or tokenholders to vote on every operational question, the system defines who has authority over what, under which conditions, and through which feedback loops.
Decentralization lives not only in voting rights, but in the distribution of authority across domains, the transparency of constraints, and the recurring practices that keep the system accountable: check-ins, double links, retrospectives, audits, peer review, and learning loops.
The model is role-based rather than actor-based. A person may hold different roles at different moments, such as Member, Delivery Steward, Contributor, or Validator. What matters is contextual authority. Each role carries specific permissions, constraints, and conflict-of-interest rules.
Each role contributes from its strongest position: Members steward capital and intent, Delivery Stewards translate intent into executable work, Contributors create value, Validators verify outcomes, and the AI Layer preserves what the institution learns over time.
The Minimum Viable Institution combines several primitives that function independently:
PEN: permanent capital and yield allocation
Octant: yield-based public goods funding
POIDH: outcome bounties and claim verification
Holacracy-inspired roles: domain-based authority and accountabilities
Bread Cooperative: solidarity fund for mutual risk-sharing
Social Oracles: qualified human verification
AI agents: memory, context, and learning support
This model proposes linking them to explore one goal:
Build a minimal institution capable of converting capital, trust, and distributed work into verifiable outcomes, while staying lightweight and hard to capture.
I think DAOs could be the flagship use case for blockchain technology.
In my opinion, the strongest case has always been the ability for communities to coordinate around shared resources and do things together. But coordination requires more than voting. It requires capital, execution, verification, reward allocation, memory, and trust.
Treasuries, grants, bounties, retroactive funding each solve part of the coordination problem, but none of them alone completes the loop from capital to execution to learning.
The Minimum Viable Institution connects these primitives into a coherent cycle:
Intent → Funding → Stewardship → Bounties → Execution → Verification → Retroactive Recognition → AI-Assisted Learning → Next Intent
This is governance as a workflow.
🍏 To make this concrete, imagine a local impact DAO called Park Commons DAO: a lightweight institution created to help a neighborhood keep their public park clean, safe, and active.
The goal is modest and easy to understand: coordinate recurring local funding, small outcome-based bounties, verified execution, and community learning around a shared public space.
This is not about replacing the municipality or building a large organization. It is about helping residents, local businesses, parents, park users, and nearby organizations coordinate around visible, local, verifiable improvements.
The community does not need to vote on every cleanup, bench repair, workshop, or sign installation. Instead, members set a quarterly intent, stewards translate it into clear bounties, contributors execute, validators verify, and the system learns from what happened.
Each role becomes active where its authority is relevant.
Fewer decisions require explicit governance because more decisions are embedded into roles, constraints, tools, and recurring workflow stages. The model is hard to capture by design because authority is distributed and bounded by default.
Members steward intent and funding distribution, but do not directly manage execution.
Delivery Stewards shape and publish bounties, but operate within fixed budgets, public criteria, and time-limited mandates.
Contributors execute and submit claims, but must meet predefined success criteria.
Validators assess outcomes, but are selected by sortition through qualified pools and conflict-aware processes.
The AI Layer preserves context, but does not decide.
Capture-resistance comes from making the critical transitions visible: intent setting, bounty publication, execution, verification, settlement, retroactive recognition, and learning.
The first layer is a covenant: a shared agreement that defines the purpose of the institution, the rights and responsibilities of members, the roles through which work happens, and the rules for allocating, executing, verifying, and remembering local improvements.
The covenant also serves as the institution's constitutional substrate. Beyond defining participation, it establishes the basic guarantees that make coordination legitimate over time. Different implementations may choose different mechanisms, ranging from mediation and peer review panels to formal appeals processes, graduated sanctions, temporary suspension of privileges, or ultimately voluntary exit. The goal is not to eliminate conflict, but to ensure that disagreement, abuse, and institutional stress can be processed without undermining trust in the system itself.
The covenant may also define mutual support mechanisms for exceptional circumstances. Some communities may choose to maintain a small Solidarity Fund, with predefined eligibility criteria and review processes, to help contributors navigate hardships that materially affect their ability to fulfill commitments.
In a PEN-like model, a person or organization may buy a seat to become a member. That seat gives them participation rights in capital allocation decisions. The important feature is that membership is tied to stewardship, not speculation.
The capital layer follows the Octant logic: preserve the principal and allocate only the yield. This gives the organization a long time horizon. Instead of spending down a treasury, the network becomes a capital steward.
🍏 For Park Commons DAO, members could be residents, local businesses, parents, civil society groups, park users, or supporters.
The covenant is not just a values statement. It is the institution’s participation agreement. It clarifies:
who can become a Member
what rights a seat grants
what responsibilities Members accept
how quarterly Intents are set
how Delivery Stewards, Contributors, and Validators operate
how conflicts of interest are handled
how claims are verified
how retroactive recognition is distributed
how records, evidence, and decisions are preserved
how decisions are challenged
how conflicts are resolved
what kind of difficult situation triggers the Solidarity Fund
how the covenant itself can be amended
In this sense, the covenant acts as the root coordination object of the institution. It links membership, capital, roles, permissions, verification, and learning into one shared agreement. This gives the DAO a social and procedural foundation before capital starts moving.
At the beginning of each quarter, in the closing/opening of the cycle, the system runs a lightweight Governance Review as a structured AI-Assisted learning ritual.
The purpose of the Governance Review is not only to document activity, but to evaluate whether the institution's assumptions about coordination, funding, and execution proved correct. In this sense, each Intent can be understood as a hypothesis about how resources, effort, and attention can best advance the community's purpose.
It reviews:
Which assumptions proved wrong?
What outcomes were achieved?
Which bounties worked?
Which Contributors demonstrated reliable execution?
Which verification criteria were ambiguous?
Which coordination failures repeated?
What should be updated in the protocols, bounty templates, or eligibility criteria?
Where is the system becoming too dependent on specific people?
Then, based on the needs, tensions, shortcomings, and lessons that have emerged, Members establish the specific Requirements (similar to Key Results) that must be met through an Intent.
An Intent is the strategic direction for this funding cycle. It is similar to a slate, mission, mandate, or seasonal priority.
🍏 For Park Commons DAO, a quarterly Governance Review might show that the playground is cleaner than before, but elders still lack shaded seating and several benches remain damaged.
A possible quarterly Intent could be:
Improve the usability of the playground for families and elders.
Requirements behind the Intent:
Elders need safe and shaded seating.
Weekend users need extra waste disposal.
Volunteers need clearer time slots .
The community needs a visible record of what has been done.
This matters because it connects action to need. A bounty should not only say “repair a bench.” It should show which requirement it helps satisfy.
Members also define a simple funding split for the quarterly funding pool.
For example:
20% Delivery Stewardship
60% Outcome Bounties
15% Retroactive Recognition
5% Solidarity Fund
This replaces many separate budget decisions with one strategic allocation. The question Members answer is: Which funding distribution best helps us achieve this Intent?
Members stay focused on stewardship and allocation. Operational decisions move into the workflow.
Delivery Stewards are the lightweight coordination role.
They are time-bound stewards with bounded authority by default. Their role is to curate bounty proposals, package them clearly, align them with the quarterly Intent, connect them to Requirements, define acceptance criteria, support Contributors with context, reduce duplication, and help the system maintain coherence.
The role combines elements of a delivery manager, operations support role, domain lead, and expert council, but with authority bounded by public criteria, fixed budgets, and time-limited mandates.
Delivery Stewards can edit the flow, but they do not own it.
Anyone can propose a bounty. Delivery Stewards help decide which bounties are publishable under the current Intent and fixed bounty budget.
🍏 In Park Commons DAO, neighbors may propose “clean the park”. Delivery Stewards translate that energy into clear, verifiable bounties: playground cleanup, bench repair, waste signage, or shade mapping.
A proposer brings initiative. A steward brings alignment. The workflow remains open, while the published bounty set remains coherent.
Delivery Stewards are selected from a Qualified Steward Pool: a transparent list of people or teams eligible to serve in the role.
Eligibility can be earned through onboarding, demonstrated execution, domain credentials, prior verified contributions, training, community-recognized expertise, absence of relevant conflicts of interest, etc.
The system should make visible who is eligible, why they are eligible, when eligibility expires or needs renewal, and what conflicts of interest apply.
Once the pool exists, Delivery Stewards can be selected through rotation, sortition, or a simple eligibility-weighted process. The goal is to preserve quality filters while reducing informal proximity to power.
Once the Intent and funding split are defined, the bounty layer translates strategy into executable outcomes. This is where POIDH-style mechanisms become valuable.
A bounty is a concrete outcome with:
Intent Link
Requirement Link
Success Criteria
Budget
Deadline
Verification Method
Claim Process
A bounty should make execution legible.
🍏 For Park Commons DAO, examples could be:
Bounty: Playground Cleanup Day
Requirement Link: Families need cleaner play areas.
Success Criteria: before/after photos from fixed points, trash collected and disposed in an approved location, completion by Sunday 6pm.
Bounty: Repair and Paint Two Benches
Requirement Link: Elders need safe seating.
Success Criteria: before/after photos, materials list, confirmation that benches are stable and safe to use.
Bounty: Map Shade and Seating Needs
Requirement Link: The community needs better information for future improvements.
Success Criteria: map with observation points, photos, and three priority recommendations for the next cycle.
This creates an operational bridge between capital allocation and real-world action. The bounty layer answers: What specific outcome would prove that this Intent is moving forward?
Contributors are the people who execute. They can be individuals, collectives, service providers, working groups, or temporary teams.
The system should allow permissionless initiative while preserving trust standards. For low-risk bounties, anyone may claim. For higher-risk bounties, Contributors may need minimum credentials, identity checks, prior execution history, or a small stake.
The more risky the work, the stronger the filter. This keeps the institution lightweight while protecting the work that actually matters.
Not all allocated bounty funding must be spent every cycle. If stewards determine that the available bounty proposals do not adequately serve the Intent, unallocated funds may remain undeployed, and roll these unused resources into future funding cycles. This creates a bias toward funding quality over funding velocity and reduces pressure to allocate capital simply because it is available.
Execution needs verification.
For onchain work, smart contracts can verify many things directly. For offchain work, human judgment is often required. This is where a Social Oracle becomes useful.
A Social Oracle is a small verification panel selected from a Qualified Verification Pool. Its job is to answer: Did the claim satisfy the success criteria?
Verification eligibility is different from delivery eligibility. A Delivery Steward needs coordination capacity. A Validator needs proximity, context, independence, or relevant expertise.
🍏 For Park Commons DAO, a Validator might be eligible because they live nearby, attended the cleanup, regularly use the park, hold a local attestation, have relevant expertise, and did not contribute to the bounty being verified.
The panel can be selected through anonymized sortition. This makes it harder for Contributors to influence specific Validators before review.
The Validator role is narrow. Validators do not decide whether the bounty was strategically valuable. They verify whether the agreed outcome was achieved.
The same verification infrastructure can also support non-bounty processes defined by the covenant, including eligibility reviews for mutual support mechanisms such as a Solidarity Fund.
When a claim passes verification, payment is released. Programmable escrow, pull payments, claim records, and automatic settlement reduce the need for trusted intermediaries. If the agreed criteria are verified, settlement follows. Crypto rails for coordination.
Bounties are useful for expected work. Retroactive recognition is useful for unexpected value.
Every quarter, some valuable contributions will fall outside formal bounties: coordination work, care work, sensemaking, strategic introductions, conflict resolution, documentation, process improvements, or support that helped others succeed.
🍏 In Park Commons DAO, retroactive recognition might reward someone who coordinated with the municipality to unlock waste pickup, a parent group that brought volunteers, a local café that donated water, a resident who documented the process, or someone who improved the cleanup template for future cycles.
A lightweight retro process can use nominations, contribution tags, evidence, peer signals, and validator review. The question is: What value helped advance the Intent that the original bounty structure did not anticipate?
This is how the system recognizes care work, coordination work, sensemaking, and emergent contribution.
The AI-assisted institutional memory preserves context across cycles. It maintains artifacts, protocols, attestations, contributor histories, verification rationales, funding decisions, retroactive recognition explanations, and lessons learned.
Each Intent and bounty can be treated as a coordination object: a living record with state, participants, commitments, evidence, validation, amendments, and outcomes.
AI agents should not govern the system. They should help the system remember.
Their role is closer to institutional librarianship:
summarize cycles
surface repeated coordination failures
connect current decisions to past outcomes
support steward onboarding
help maintain protocols
preserve rationale behind decisions
prepare quarterly learning reviews
🍏 In Park Commons DAO, the AI Layer might summarize that playground cleanup worked well, bench repair was underbudgeted, validator criteria for signage were ambiguous, and shade mapping should inform the next quarter.
This allows Delivery Stewards to rotate while the system keeps continuity.
The AI Layer should also track whether recurrent knowledge is migrating from people into shared artifacts, protocols, and tools. A lightweight institution becomes more resilient when it reduces dependency on specific individuals without erasing their contributions. A key review question: Is the institution making the community more capable of coordinating by itself over time?
AI improves output quality because it helps Members, Contributors, Stewards, and Validators avoid repeating mistakes, recognize trustworthy execution, and understand what actually worked.
DAOs need to become better at turning shared resources into shared action.
That means improving the whole coordination loop:
Capital formation → Intent setting → Execution → Verification → Reward allocation → Memory
The current wave of lightweight mechanisms is promising because it treats governance as design constraints, not endless deliberation. The hypothesis here:
DAOs become more effective when fewer decisions require explicit governance, and the authority is embedded into clear workflows with strong defaults.
I would expect the result to be a more streamlined coordination process, greater transparency, and fewer opportunities for informal power to accumulate quietly.
Just a matter of getting it up and running.