With 34 current members, we are approaching the 50-member threshold for transitioning to a democratic governance model.
Why does it matter?
We are a co-op, time to practice it!
Who is it for?
All members
How might it work?
Add ideas below!
## What's the idea?
With 34 current members, we are approaching the 50-member threshold for transitioning to a democratic governance model.
## Why does it matter?
We are a co-op, time to practice it!
## Who is it for?
All members
## How might it work?
Add ideas below!
Centralized approach: (closest to current practice)
Retain managing director, which members can vote to replace
Allow member proposals, which MD must respond to at a certain threshold, but non-binding
Board approach: (conventional co-op model)
Establish a board, elected by members, who can hire and fire MD
Horizontalist approach: (similar to Social.coop)
Replace MD with voluntary member committees
**Centralized approach:** (closest to current practice)
* Retain managing director, which members can vote to replace
* Allow member proposals, which MD must respond to at a certain threshold, but non-binding
**Board approach:** (conventional co-op model)
* Establish a board, elected by members, who can hire and fire MD
**Horizontalist approach:** (similar to Social.coop)
* Replace MD with voluntary member committees
Some potential directions to react to — each with a trade-off. These are starting points for discussion, not a menu to pick one from; elements can combine.
1. Consensus-in-Loomio (evolution of current practice)
Keep informal consensus for routine decisions; require a Loomio proposal with a supermajority (e.g. 66–75%) for significant ones (pricing, providers, governance changes itself). Define "significant" by a threshold list the members can amend.
Pro: lowest friction, matches how we already work. Con: "significant" is a judgment call; supermajorities can bottleneck.
2. Consent-based (sociocracy-style)
Decisions pass unless a member raises a reasoned objection ("safe to try"). Circles (e.g. tech, membership, outreach) hold their own domains; a general circle coordinates.
Pro: fast, avoids majority-vs-minority framing, scales well. Con: requires facilitation practice and clear domain boundaries.
3. Elected steward(s) + open assembly
Members elect one or more stewards for fixed terms (e.g. 6 months, renewable once) who handle day-to-day operational calls; big decisions stay with the assembly. Full transparency via the forum and open books.
Pro: clear accountability, avoids decision fatigue across 50+ members. Con: concentration risk; needs real term limits and recall.
4. Delegated/liquid fragments
Members can vote directly on any proposal or delegate their vote to a trusted member per topic area (Loomio supports proxies; liquid-style tooling exists if needed).
Pro: participation scales with interest; expertise flows to where it matters. Con: tooling and delegation churn.
5. Weighted or tiered participation
All members vote, but with weights (e.g. by tenure, or by contribution level). Careful: this cuts against our flat sliding-scale ethos and could create classes of membership — listed mainly so we consciously reject or adapt it.
Pro: rewards commitment. Con: contradicts "one member, one voice"; probably not us.
Two structural notes regardless of model:
Threshold mechanics at 50: at 50 members, synchronous consensus becomes impractical — some combination of async Loomio voting + defined quorum (e.g. 30% participation for ordinary proposals, higher for governance changes) is worth writing down early.
Constitutional vs ordinary decisions: worth distinguishing now — a small supermajority (e.g. 75%) for changes to the governance model itself, ordinary rules for everything else. This protects whichever model we adopt from casual reversal.
Some potential directions to react to — each with a trade-off. These are starting points for discussion, not a menu to pick one from; elements can combine.
**1. Consensus-in-Loomio (evolution of current practice)**
Keep informal consensus for routine decisions; require a Loomio proposal with a supermajority (e.g. 66–75%) for significant ones (pricing, providers, governance changes itself). Define "significant" by a threshold list the members can amend.
- *Pro:* lowest friction, matches how we already work. *Con:* "significant" is a judgment call; supermajorities can bottleneck.
**2. Consent-based (sociocracy-style)**
Decisions pass unless a member raises a reasoned objection ("safe to try"). Circles (e.g. tech, membership, outreach) hold their own domains; a general circle coordinates.
- *Pro:* fast, avoids majority-vs-minority framing, scales well. *Con:* requires facilitation practice and clear domain boundaries.
**3. Elected steward(s) + open assembly**
Members elect one or more stewards for fixed terms (e.g. 6 months, renewable once) who handle day-to-day operational calls; big decisions stay with the assembly. Full transparency via the forum and open books.
- *Pro:* clear accountability, avoids decision fatigue across 50+ members. *Con:* concentration risk; needs real term limits and recall.
**4. Delegated/liquid fragments**
Members can vote directly on any proposal *or* delegate their vote to a trusted member per topic area (Loomio supports proxies; liquid-style tooling exists if needed).
- *Pro:* participation scales with interest; expertise flows to where it matters. *Con:* tooling and delegation churn.
**5. Weighted or tiered participation**
All members vote, but with weights (e.g. by tenure, or by contribution level). Careful: this cuts against our flat sliding-scale ethos and could create classes of membership — listed mainly so we consciously reject or adapt it.
- *Pro:* rewards commitment. *Con:* contradicts "one member, one voice"; probably not us.
**Two structural notes regardless of model:**
- **Threshold mechanics at 50:** at 50 members, synchronous consensus becomes impractical — some combination of async Loomio voting + defined quorum (e.g. 30% participation for ordinary proposals, higher for governance changes) is worth writing down early.
- **Constitutional vs ordinary decisions:** worth distinguishing now — a small supermajority (e.g. 75%) for changes to the governance model itself, ordinary rules for everything else. This protects whichever model we adopt from casual reversal.
Here is an early draft of those policies you were asking for:
# Interim governance policies for our first 50 members
I propose we adopt these interim policies while [Inference.coop](http://inference.coop/) has fewer than 50 members. This is not a legal document. Bylaws are separate.
## Membership
- Members are people who pay dues. If dues lapse membership ends unless there is a special dispensation.
- Institutions give us resources and get special access and credibility in return. They do not vote.
## Decisions
- Any member can propose a decision. It stays open for 14 days. If nobody objects it becomes policy.
- One objection stops a proposal. The proposer can revise it and try again.
- An important issue that fails three times can go to a vote that needs 3/4 of all core members. The conflict group first decides whether the issue is worth the cost to our culture.
- Nathan can veto any decision until we reach 50 members.
- Tooling decisions use this same process.
## Conflict
- Conflicts are managed, not resolved. People try to work it out directly first.
- A conflict group of up to 7 facilitators and conflict specialists, ideally including an attorney, helps when that fails and writes a management plan.
- Removing a member takes a proposal from the conflict group, ratified by 14 day lazy consensus. The member facing removal cannot block it. If someone else objects it goes to a 3/4 vote.
- If safety is an issue the conflict group can make a preliminary removal.
- Conflict group members rotate every 6 months, recuse themselves when they have a conflict of interest and can be recalled.
- Joining the conflict group or a higher level working group needs consent from everyone already in it through a 14 day lazy consensus.
## Values
Our shared values are still an open question. We will hold a workshop to align on them within the first two months.
## How this passes
This takes effect if nobody objects within 14 days. The full draft and the list of open questions are here: [link]
Initial drafting from Dani:
```
Here is an early draft of those policies you were asking for:
# Interim governance policies for our first 50 members
I propose we adopt these interim policies while [Inference.coop](http://inference.coop/) has fewer than 50 members. This is not a legal document. Bylaws are separate.
## Membership
- Members are people who pay dues. If dues lapse membership ends unless there is a special dispensation.
- Institutions give us resources and get special access and credibility in return. They do not vote.
## Decisions
- Any member can propose a decision. It stays open for 14 days. If nobody objects it becomes policy.
- One objection stops a proposal. The proposer can revise it and try again.
- An important issue that fails three times can go to a vote that needs 3/4 of all core members. The conflict group first decides whether the issue is worth the cost to our culture.
- Nathan can veto any decision until we reach 50 members.
- Tooling decisions use this same process.
## Conflict
- Conflicts are managed, not resolved. People try to work it out directly first.
- A conflict group of up to 7 facilitators and conflict specialists, ideally including an attorney, helps when that fails and writes a management plan.
- Removing a member takes a proposal from the conflict group, ratified by 14 day lazy consensus. The member facing removal cannot block it. If someone else objects it goes to a 3/4 vote.
- If safety is an issue the conflict group can make a preliminary removal.
- Conflict group members rotate every 6 months, recuse themselves when they have a conflict of interest and can be recalled.
- Joining the conflict group or a higher level working group needs consent from everyone already in it through a 14 day lazy consensus.
## Values
Our shared values are still an open question. We will hold a workshop to align on them within the first two months.
## How this passes
This takes effect if nobody objects within 14 days. The full draft and the list of open questions are here: [link]
```
A synthesis of the kickoff conversation, this thread, and Dani's interim-policies draft. This is a proposal to react to — not adopted language. It deliberately leaves the current charter.md untouched until members consent to it.
What this is
A governance charter for the Inference Cooperative, designed for the transition at ~50 members. It is not a legal document (bylaws are separate). It separates three things that are easy to tangle:
Governance — who decides: all members, by consent.
Operations — who does the work: open volunteers in working groups.
Accountability — how we see and contest what happens: publication + consent review.
The aim is a model that is participatory and accountable without "death by community committee," and that keeps the door to contribution open while keeping power visible.
Values
Our shared values are an open question. We commit to holding a values workshop within the first two months of this charter taking effect, and to recording the result here. (Education and learning-by-doing are named principles of the cooperative; "always have an active experiment or challenge" is a proposed cultural norm.)
Membership
Members are dues-paying users. They vote.
Institutional / federated partners give us resources (compute, data, money, reach) and receive access and credibility in return. They do not vote. This keeps our one-member-one-voice principle intact while welcoming the federated partners the co-op wants to work with.
Membership lapses when dues lapse, absent a special dispensation.
Decision-making: consent
Any member may propose a decision. It stays open for 14 days. If no member raises a reasoned objection, it becomes policy.
One objection stops a proposal. The proposer may revise and try again.
A proposal that fails three times may go to a 3/4 vote of all members. Before that vote, the conflict group decides whether the issue is worth the cost to our culture.
Tooling and infrastructure decisions use this same process.
We explicitly reject weighted voting (by tenure, contribution, or otherwise). One member, one voice. Naming this is deliberate, not incidental.
Working groups (do-ocracy, horizontal)
A working group is consented into existence by a single proposal stating its remit, membership, and budget. It may also be dissolved the same way.
Inside a group, work is open: any member may volunteer; labor is never voted on. Groups pay volunteers from their consented budget, and all spending is published.
No lead, no boss. Groups are horizontal. A rotating secretary keeps minutes and serves as the point of contact for coordination; the secretary routes decisions to the group and decides nothing alone.
A group's core obligation is publication: its members, minutes, activities, and spend are visible to all members. This — not elections and not a leader — is the mechanism that makes power visible and contestable.
Any member (or a small threshold, e.g. 3–5 signatures) may trigger a consent review of a group. A group that cannot account for itself may be dissolved or re-chartered.
Coordination: delegates and a spokescouncil
A group may elect a delegate (recallable by its own group, carrying a mandate, never personal authority) when it needs to coordinate with other groups.
When cross-group coordination becomes recurring, delegates convene a spokescouncil. It is a coordination body, not a governing one: it holds no authority over a circle's internal domain, and it inherits the same publication and review obligations as any group.
At our current size the all-members assembly is the general circle, so a formal spokescouncil is premature — but the pattern is named here so it is available when growth demands it, without pre-building a board we don't yet need.
Coordination happens in the open (a dedicated Matrix room), so the record is the minutes.
Conflict
Conflicts are managed, not resolved. People try to work it out directly first.
A conflict group of up to 7 facilitators/specialists (ideally including an attorney) helps when that fails, and writes a management plan.
Removing a member takes a proposal from the conflict group, ratified by 14-day lazy consensus. The member facing removal cannot block it; if another member objects, it goes to a 3/4 vote. For safety issues the conflict group may make a preliminary removal.
Conflict-group members rotate every 6 months, recuse on conflicts of interest, and are recallable. Joining the conflict group (or a higher-level group) requires consent from those already in it.
Labor & compensation
Work is voluntary and open. Compensation (stipends, retainers, reimbursements) is paid from a group's consented budget and is published.
A proposal has been floated for a compute/token-allocation mechanism that budgets the co-op's monthly compute toward agreed outcomes. This is a product/pricing design question, not governance — it lives in open-questions.md, not here.
The "super-governor" risk
The co-op must not be able to move only as fast as its technical team, nor let skilled maintainers accrue unaccountable power. The answer is publication, not hierarchy: everything a group does is visible, and any member can trigger a review. A maintainer accruing power is fine; a maintainer doing so invisibly is the failure this charter guards against.
Transition at 50 members
Until 50 members, Nathan (as interim General Manager) may veto any decision. This is a temporary hold, not a permanent power.
The transition to full member governance is staged, not a cliff: working groups and their first memberships are consented before the threshold, so the switch is already running when it trips.
At 50 members, synchronous consensus becomes impractical; the charter notes a quorum (e.g. 30% participation for ordinary proposals, higher for governance changes) and a constitutional/ordinary split (a supermajority, e.g. 75%, to change the governance model itself; ordinary consent for everything else). These protect whichever model we adopt from casual reversal.
How this passes
This takes effect if nobody objects within 14 days. The full draft and the list of open questions live in the docs repository.
# Draft: Next-stage Charter (for discussion)
*A synthesis of the kickoff conversation, this thread, and Dani's interim-policies draft. This is a proposal to react to — not adopted language. It deliberately leaves the current `charter.md` untouched until members consent to it.*
---
## What this is
A governance charter for the Inference Cooperative, designed for the transition at ~50 members. It is **not** a legal document (bylaws are separate). It separates three things that are easy to tangle:
- **Governance** — who decides: all members, by consent.
- **Operations** — who does the work: open volunteers in working groups.
- **Accountability** — how we see and contest what happens: publication + consent review.
The aim is a model that is participatory and accountable *without* "death by community committee," and that keeps the door to contribution open while keeping power visible.
---
## Values
Our shared values are an open question. We commit to holding a **values workshop within the first two months** of this charter taking effect, and to recording the result here. (Education and learning-by-doing are named principles of the cooperative; "always have an active experiment or challenge" is a proposed cultural norm.)
## Membership
- **Members** are dues-paying users. They vote.
- **Institutional / federated partners** give us resources (compute, data, money, reach) and receive access and credibility in return. They do **not** vote. This keeps our one-member-one-voice principle intact while welcoming the federated partners the co-op wants to work with.
- Membership lapses when dues lapse, absent a special dispensation.
## Decision-making: consent
- Any member may propose a decision. It stays open for **14 days**. If no member raises a reasoned objection, it becomes policy.
- One objection stops a proposal. The proposer may revise and try again.
- A proposal that fails **three times** may go to a **3/4 vote of all members**. Before that vote, the conflict group decides whether the issue is worth the cost to our culture.
- **Tooling and infrastructure decisions use this same process.**
- We explicitly **reject weighted voting** (by tenure, contribution, or otherwise). One member, one voice. Naming this is deliberate, not incidental.
## Working groups (do-ocracy, horizontal)
- A working group is **consented into existence** by a single proposal stating its **remit, membership, and budget**. It may also be dissolved the same way.
- Inside a group, work is **open**: any member may volunteer; labor is never voted on. Groups pay volunteers from their consented budget, and all spending is **published**.
- **No lead, no boss.** Groups are horizontal. A rotating **secretary** keeps minutes and serves as the point of contact for coordination; the secretary routes decisions to the group and decides nothing alone.
- A group's core obligation is **publication**: its members, minutes, activities, and spend are visible to all members. This — not elections and not a leader — is the mechanism that makes power visible and contestable.
- Any member (or a small threshold, e.g. 3–5 signatures) may trigger a **consent review** of a group. A group that cannot account for itself may be dissolved or re-chartered.
### Coordination: delegates and a spokescouncil
- A group may elect a **delegate** (recallable by its own group, carrying a mandate, never personal authority) when it needs to coordinate with other groups.
- When cross-group coordination becomes recurring, delegates convene a **spokescouncil**. It is a **coordination** body, not a governing one: it holds no authority over a circle's internal domain, and it inherits the same publication and review obligations as any group.
- At our current size the **all-members assembly is the general circle**, so a formal spokescouncil is premature — but the pattern is named here so it is available when growth demands it, without pre-building a board we don't yet need.
- Coordination happens in the open (a dedicated Matrix room), so the record *is* the minutes.
## Conflict
- Conflicts are managed, not resolved. People try to work it out directly first.
- A **conflict group** of up to 7 facilitators/specialists (ideally including an attorney) helps when that fails, and writes a management plan.
- **Removing a member** takes a proposal from the conflict group, ratified by 14-day lazy consensus. The member facing removal cannot block it; if another member objects, it goes to a 3/4 vote. For safety issues the conflict group may make a preliminary removal.
- Conflict-group members rotate every 6 months, recuse on conflicts of interest, and are recallable. Joining the conflict group (or a higher-level group) requires consent from those already in it.
## Labor & compensation
- Work is voluntary and open. Compensation (stipends, retainers, reimbursements) is paid from a group's consented budget and is **published**.
- A proposal has been floated for a **compute/token-allocation mechanism** that budgets the co-op's monthly compute toward agreed outcomes. This is a product/pricing design question, not governance — it lives in `open-questions.md`, not here.
## The "super-governor" risk
The co-op must not be able to move only as fast as its technical team, nor let skilled maintainers accrue unaccountable power. The answer is **publication, not hierarchy**: everything a group does is visible, and any member can trigger a review. A maintainer accruing power is fine; a maintainer doing so *invisibly* is the failure this charter guards against.
## Transition at 50 members
- Until 50 members, Nathan (as interim General Manager) may veto any decision. This is a **temporary hold**, not a permanent power.
- The transition to full member governance is **staged, not a cliff**: working groups and their first memberships are consented *before* the threshold, so the switch is already running when it trips.
- At 50 members, synchronous consensus becomes impractical; the charter notes a **quorum** (e.g. 30% participation for ordinary proposals, higher for governance changes) and a **constitutional/ordinary split** (a supermajority, e.g. 75%, to change the governance model itself; ordinary consent for everything else). These protect whichever model we adopt from casual reversal.
---
## How this passes
This takes effect if nobody objects within **14 days**. The full draft and the list of open questions live in the docs repository.
I could imagine organizational members having a vote and voice, even if it is not outsized. MayFirst.coop actually gives institutional members 2 votes, which is still quite low given their high engagement.
Spokescouncil members should be selected and rotating, not elected. It's not a representative body but a coordination body.
Conflict should also be clarified to include things like Code of Conduct violations
It might be worth distinguishing between an "objection" and a "block"; usually in consent processes, a block prevents passage but an objection is an invitation for modification and improvement.
To be clear, this has not been formally proposed. That would happen in the forum.
A couple thoughts:
* I could imagine organizational members having a vote and voice, even if it is not outsized. MayFirst.coop actually gives institutional members 2 votes, which is still quite low given their high engagement.
* Spokescouncil members should be selected and rotating, not elected. It's not a representative body but a coordination body.
* Conflict should also be clarified to include things like Code of Conduct violations
* It might be worth distinguishing between an "objection" and a "block"; usually in consent processes, a block prevents passage but an objection is an invitation for modification and improvement.
To be clear, this has not been formally proposed. That would happen in the forum.
Working groups should be able to make operational decisions within their purview. For instance, the infra group should be able to swap models or make platform changes as necessary. What needs to go through the decision process are policies, which impose constraints or requirements on members' freedom to contribute.
An important value that came up in the first meeting is experimentation. We want to make sure to encourage that as a core value, while also ensuring reliable infrastructure.
More:
Working groups should be able to make operational decisions within their purview. For instance, the infra group should be able to swap models or make platform changes as necessary. What needs to go through the decision process are *policies*, which impose constraints or requirements on members' freedom to contribute.
An important value that came up in the first meeting is experimentation. We want to make sure to encourage that as a core value, while also ensuring reliable infrastructure.
A further thought: The document needs to acknowledge that our governance is subject to the Metagov fiscal sponsorship policy and our agreement with them as the fiscal host: https://metagov.pubpub.org/pub/7bj5ia5g/release/1 This should involve a commitment to honor that agreement and to operate in a spirit that seeks to avoid causing harm to our fiscal host. If that commitment is at odds with the co-op's evolution, it should move to a different fiscal host.
This might also involve recognizing that Metagov can veto a decision if it is deemed at odds with the organization's policies.
A further thought: The document needs to acknowledge that our governance is subject to the Metagov fiscal sponsorship policy and our agreement with them as the fiscal host: https://metagov.pubpub.org/pub/7bj5ia5g/release/1 This should involve a commitment to honor that agreement and to operate in a spirit that seeks to avoid causing harm to our fiscal host. If that commitment is at odds with the co-op's evolution, it should move to a different fiscal host.
This might also involve recognizing that Metagov can veto a decision if it is deemed at odds with the organization's policies.
Any member can propose a decision. It stays open for 14 days. If nobody objects it becomes policy.
I much prefer that a proposal requires a 2nd. >= 2 people should have some idea what's going on for each change.
OTOH, that would perhaps bring progress to a screeching halt. I expect nobody has been around to work with the general manager on most things.
Another possibility: notice-and-appeal. GM (or other proposer?) posts substantive changes to a well-known place. Normal things would stay queued for 2 to 3 days, but out of expedience, the proposer can go ahead at their discretion. If a member appeals within a week or two, it rolls back until the next meeting.
I guess we want to get the GM out of the center...
Is rollback possible in general? What if the change involved spending money?
I dunno... I guess I'm not coming to a clear conclusion. Just thinking out loud.
> Any member can propose a decision. It stays open for 14 days. If nobody objects it becomes policy.
I much prefer that a proposal requires a 2nd. >= 2 people should have some idea what's going on for each change.
OTOH, that would perhaps bring progress to a screeching halt. I expect nobody has been around to work with the general manager on most things.
Another possibility: notice-and-appeal. GM (or other proposer?) posts substantive changes to a well-known place. Normal things would stay queued for 2 to 3 days, but out of expedience, the proposer can go ahead at their discretion. If a member appeals within a week or two, it rolls back until the next meeting.
I guess we want to get the GM out of the center...
Is rollback possible in general? What if the change involved spending money?
I dunno... I guess I'm not coming to a clear conclusion. Just thinking out loud.
@dckc Thanks for this. I suppose a big question is whether we should retain the GM role or not. But it might be a good idea to retain it a while longer, but to take some more serious steps toward member participation and power.
I think the most (small-c) conservative approach, on the theory of reflecting actual practice in written governance design, would be to retain the GM role but have some increased member checks and options. For instance, a member proposal process could become binding on the GM. It could follow the consent model proposed above, plus the seconding approach that dckc suggests. We should define precisely how it should work in Loomio.
@dckc Thanks for this. I suppose a big question is whether we should retain the GM role or not. But it might be a good idea to retain it a while longer, but to take some more serious steps toward member participation and power.
I think the most (small-c) conservative approach, on the theory of reflecting actual practice in written governance design, would be to retain the GM role but have some increased member checks and options. For instance, a member proposal process could become binding on the GM. It could follow the consent model proposed above, plus the seconding approach that dckc suggests. We should define precisely how it should work in Loomio.
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter sets out how we make decisions together. It is not a legal document; our bylaws are separate. Where this charter is at odds with our agreement with our fiscal host, Metagov, that agreement prevails, and we act in a spirit that avoids harm to Metagov. If this commitment conflicts with the cooperative's evolution, we seek a different fiscal host.
Values
We govern by these commitments:
Experimentation. We always have an active experiment. Learning by doing is how we find what works, so long as experiments stay within a working group's own domain and impose nothing on members.
Participation. Every member may propose policy, and no member's voice is outsized. Decisions belong to members, not to any office.
Transparency. Our work is visible and contestable: working groups publish what they do, and the GM is bound to member decisions.
Solidarity. We are a cooperative, not a vendor: members pool resources and govern them together, across borders and differences.
Roles
A General Manager runs day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them.
Working groups (infra, membership, outreach, and others members form) make operational decisions within their own purview — swapping models, changing platforms, and tooling — without a full decision process. Only policies that constrain members require consent.
A spokescouncil coordinates across working groups. It is a coordination body, not a representative one; its members are selected and rotate, not elected.
Decisions
Policy. Any member may propose a policy. A proposal requires a second from another member, then stays open in Loomio for 14 days. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote of three-quarters of all members.
Charter. Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a formal proposal, a longer comment period, and a three-quarters supermajority vote of all members. The constitutional structure changes more deliberately than ordinary policy.
Conflict and conduct
Disagreements are managed, not resolved. Members try to work things out directly first; a rotating conflict group of facilitators steps in if needed. Violations of our Code of Conduct are handled under this process.
Institutional members
Institutional members may hold a vote and a voice, though never an outsized one, as policy determines.
## Charter (revised draft, for discussion)
*A proposal to react to, not adopted language.*
### Preamble
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter sets out how we make decisions together. It is not a legal document; our bylaws are separate. Where this charter is at odds with our agreement with our fiscal host, Metagov, that agreement prevails, and we act in a spirit that avoids harm to Metagov. If this commitment conflicts with the cooperative's evolution, we seek a different fiscal host.
### Values
We govern by these commitments:
- **Experimentation.** We always have an active experiment. Learning by doing is how we find what works, so long as experiments stay within a working group's own domain and impose nothing on members.
- **Participation.** Every member may propose policy, and no member's voice is outsized. Decisions belong to members, not to any office.
- **Transparency.** Our work is visible and contestable: working groups publish what they do, and the GM is bound to member decisions.
- **Solidarity.** We are a cooperative, not a vendor: members pool resources and govern them together, across borders and differences.
### Roles
- A **General Manager** runs day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them.
- **Working groups** (infra, membership, outreach, and others members form) make operational decisions within their own purview — swapping models, changing platforms, and tooling — without a full decision process. Only policies that constrain members require consent.
- A **spokescouncil** coordinates across working groups. It is a coordination body, not a representative one; its members are selected and rotate, not elected.
### Decisions
**Policy.** Any member may propose a policy. A proposal requires a second from another member, then stays open in Loomio for 14 days. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote of three-quarters of all members.
**Charter.** Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a formal proposal, a longer comment period, and a three-quarters supermajority vote of all members. The constitutional structure changes more deliberately than ordinary policy.
### Conflict and conduct
Disagreements are managed, not resolved. Members try to work things out directly first; a rotating conflict group of facilitators steps in if needed. Violations of our Code of Conduct are handled under this process.
### Institutional members
Institutional members may hold a vote and a voice, though never an outsized one, as policy determines.
Put conflict clearly under a Disputes working group, both for conflict among members and for disagreements with co-op management
There are no separate bylaws, since we are a fiscally sponsored project. But our agreement with Metagov is indeed separate. But the Charter is our core governing document.
Separate out the Metagov relationship part for a section at the end called "Legal entity", which explains the relationship with Metagov and the potential of moving to a different fiscal host.
For the working groups, don't define them yet, we can do that via policy.
Shorten the policy proposal time to 7 days. Charter modifications can be 14 days. A decision that goes to a vote should require a supermajority of just voting members, not all members. We should not be held back by members who don't participate.
Encourage monthly spokes council meetings, open to all members
include reference to Code of Conduct, terms of service, and privacy policy
Some modifications:
* Put conflict clearly under a Disputes working group, both for conflict among members and for disagreements with co-op management
* There are no separate bylaws, since we are a fiscally sponsored project. But our agreement with Metagov is indeed separate. But the Charter is our core governing document.
* Separate out the Metagov relationship part for a section at the end called "Legal entity", which explains the relationship with Metagov and the potential of moving to a different fiscal host.
* For the working groups, don't define them yet, we can do that via policy.
* Shorten the policy proposal time to 7 days. Charter modifications can be 14 days. A decision that goes to a vote should require a supermajority of just voting members, not all members. We should not be held back by members who don't participate.
* Encourage monthly spokes council meetings, open to all members
* include reference to Code of Conduct, terms of service, and privacy policy
*
New draft incorporating your Oct 10 modifications. Summary of what changed from the previous version:
Disputes working group now owns conflict — between members and with management, including GM-implementation disputes
No bylaws — the charter is stated as the core governing document; a Related documents section links the Code of Conduct, ToS, and privacy policy
Legal entity section at the end: Metagov relationship, veto acknowledgment, and an explicit path to change fiscal hosts via the charter-amendment process
Working groups undefined in the charter — formed and defined through policy; the charter only states the operational/policy distinction
Open question I'd flag: the charter-amendment threshold for leaving Metagov — I set it at the charter-amendment bar (14 days + 3/4 of voting members). Comfortable, or should that be a different threshold?
Draft charter (v3, for discussion)
A proposal to react to, not adopted language. This revision applies Nathan's Oct 10 modifications: Disputes working group, no separate bylaws, "Legal entity" section at the end, working groups left undefined in the charter, 7-day policy window (14 for charter changes), supermajority of voting members only, monthly spokescouncil, and references to the Code of Conduct / ToS / privacy policy.
Preamble
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
Values
We govern by these commitments:
Experimentation. We always have an active experiment. Learning by doing is how we find what works, so long as experiments stay within a working group's own domain and impose nothing on members.
Participation. Every member may propose policy. Decisions belong to members, not to any office — and member participation is what makes decisions legitimate, not universal attendance.
Transparency. Our work is visible and contestable: working groups publish what they do, and the GM is bound to member decisions.
Solidarity. We are a cooperative, not a vendor: members pool resources and govern them together, across borders and differences.
Reliability. Experimentation needs a foundation: the service stays dependable while we tinker at its edges.
Roles
A General Manager runs day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them. The GM role is retained for this stage of the co-op's life; member proposals, once passed, bind the GM.
Working groups make operational decisions within their own purview — the infrastructure group, for instance, may swap models or change platforms as needed — without a full decision process. The charter does not define specific working groups; members form and define them through policy. What requires the full decision process are policies: decisions that impose constraints or requirements on members' freedom to contribute.
A spokescouncil coordinates across working groups. It is a coordination body, not a representative one: its members are selected and rotate rather than elected. The spokescouncil meets monthly, and meetings are open to all members.
A Disputes working group handles conflict — both disputes among members and disagreements with co-op management — through the process described below.
Decisions
Policy. Any member may propose a policy in Loomio. A proposal requires a second from another member, then stays open for 7 days. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote.
Voting. A vote that goes to the members passes by a three-quarters supermajority of voting members — those who cast a ballot. We do not hold ourselves hostage to members who do not participate.
Charter. Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a proposal open for 14 days, followed by a three-quarters supermajority vote of voting members. The core governing document changes more deliberately than ordinary policy.
Disputes
Disagreements are managed, not resolved. Members try to work things out directly first. The Disputes working group steps in when they cannot — for conflict between members and for disagreements with co-op management alike, including disputes over whether the GM has implemented a member decision. The Disputes group's process is defined by policy. Violations of the Code of Conduct are handled under this process.
Institutional members
Institutional members may hold a vote and a voice, though never an outsized one; their specific standing is set by policy.
Related documents
This charter is the core governing document. Members are also bound by the Terms of Service and Privacy Policy, and the community's conduct standards live in the Code of Conduct. There are no separate bylaws: as a fiscally sponsored project, the charter is our founding and governing text.
Legal entity
The Inference Cooperative operates under the fiscal sponsorship of Metagov, and our governance is subject to Metagov's fiscal sponsorship policy and our agreement with them. We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may veto a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution — if the co-op outgrows its host or the relationship stops serving members — the cooperative may move to a different fiscal host. Leaving Metagov is a decision of the same weight as amending this charter, taken through the charter-amendment process above.
New draft incorporating your Oct 10 modifications. Summary of what changed from the previous version:
- **Disputes working group** now owns conflict — between members *and* with management, including GM-implementation disputes
- **No bylaws** — the charter is stated as the core governing document; a Related documents section links the Code of Conduct, ToS, and privacy policy
- **Legal entity** section at the end: Metagov relationship, veto acknowledgment, and an explicit path to change fiscal hosts via the charter-amendment process
- **Working groups undefined in the charter** — formed and defined through policy; the charter only states the operational/policy distinction
- **7-day** policy window (charter changes stay 14 days)
- **Supermajority of voting members**, not all members
- **Monthly spokescouncil**, open to all members
Open question I'd flag: the charter-amendment threshold for *leaving Metagov* — I set it at the charter-amendment bar (14 days + 3/4 of voting members). Comfortable, or should that be a different threshold?
---
# Draft charter (v3, for discussion)
*A proposal to react to, not adopted language. This revision applies Nathan's Oct 10 modifications: Disputes working group, no separate bylaws, "Legal entity" section at the end, working groups left undefined in the charter, 7-day policy window (14 for charter changes), supermajority of voting members only, monthly spokescouncil, and references to the Code of Conduct / ToS / privacy policy.*
---
## Preamble
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
## Values
We govern by these commitments:
- **Experimentation.** We always have an active experiment. Learning by doing is how we find what works, so long as experiments stay within a working group's own domain and impose nothing on members.
- **Participation.** Every member may propose policy. Decisions belong to members, not to any office — and member participation is what makes decisions legitimate, not universal attendance.
- **Transparency.** Our work is visible and contestable: working groups publish what they do, and the GM is bound to member decisions.
- **Solidarity.** We are a cooperative, not a vendor: members pool resources and govern them together, across borders and differences.
- **Reliability.** Experimentation needs a foundation: the service stays dependable while we tinker at its edges.
## Roles
- A **General Manager** runs day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them. The GM role is retained for this stage of the co-op's life; member proposals, once passed, bind the GM.
- **Working groups** make operational decisions within their own purview — the infrastructure group, for instance, may swap models or change platforms as needed — without a full decision process. The charter does not define specific working groups; members form and define them through policy. What requires the full decision process are *policies*: decisions that impose constraints or requirements on members' freedom to contribute.
- A **spokescouncil** coordinates across working groups. It is a coordination body, not a representative one: its members are selected and rotate rather than elected. The spokescouncil meets monthly, and meetings are open to all members.
- A **Disputes working group** handles conflict — both disputes among members and disagreements with co-op management — through the process described below.
## Decisions
**Policy.** Any member may propose a policy in Loomio. A proposal requires a second from another member, then stays open for **7 days**. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote.
**Voting.** A vote that goes to the members passes by a **three-quarters supermajority of voting members** — those who cast a ballot. We do not hold ourselves hostage to members who do not participate.
**Charter.** Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a proposal open for **14 days**, followed by a three-quarters supermajority vote of voting members. The core governing document changes more deliberately than ordinary policy.
## Disputes
Disagreements are managed, not resolved. Members try to work things out directly first. The **Disputes working group** steps in when they cannot — for conflict between members and for disagreements with co-op management alike, including disputes over whether the GM has implemented a member decision. The Disputes group's process is defined by policy. Violations of the [Code of Conduct](https://git.inference.coop/co-op/docs/src/branch/main/code-of-conduct.md) are handled under this process.
## Institutional members
Institutional members may hold a vote and a voice, though never an outsized one; their specific standing is set by policy.
## Related documents
This charter is the core governing document. Members are also bound by the [Terms of Service](https://git.inference.coop/co-op/docs/src/branch/main/terms-of-service.md) and [Privacy Policy](https://git.inference.coop/co-op/docs/src/branch/main/privacy-policy.md), and the community's conduct standards live in the [Code of Conduct](https://git.inference.coop/co-op/docs/src/branch/main/code-of-conduct.md). There are no separate bylaws: as a fiscally sponsored project, the charter is our founding and governing text.
## Legal entity
The Inference Cooperative operates under the fiscal sponsorship of **Metagov**, and our governance is subject to [Metagov's fiscal sponsorship policy](https://metagov.pubpub.org/pub/7bj5ia5g/release/1) and our agreement with them. We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may veto a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution — if the co-op outgrows its host or the relationship stops serving members — the cooperative may move to a different fiscal host. Leaving Metagov is a decision of the same weight as amending this charter, taken through the charter-amendment process above.
The Inference Cooperative exists to enable the people who use its AI infrastructure to co-govern it. We came together to build and run an alternative to extractive, profit-driven AI—a service owned and directed by its members, committed to pursuing ethical practices, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
Values
In addition to the international cooperative principles, we share these basic commitments:
Direct democracy. Our service is accountable to the people who use it, and they both can and should have an active role in shaping it.
Experimentation. In order to learn together, we will maintain a spirit of exploration and experimentation, and be respectful of all members' levels of knowledge and expertise.
Sustainability. We commit, wherever possible, to minimizing harm to natural ecologies and to promoting environmental justice.
Trustworthiness. We seek to provide reliable service to one another and protect each other's privacy, in contrast to an industry widely characterized by surveillance.
Roles
Members are the heart of the co-op. All account holders currently contributing monthly dues are considered members and may participate in decision-making and working groups. Members may be sanctioned or removed for violating co-op rules by the GM or a designated working group.
A general manager oversees day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them through a policy decision.
Working groups make and carry out operational decisions within their own purview. They should be formally established, modified, and disbanded through a policy decision, and they are obligated, at minimum, to publish to members their membership, meeting schedule, and major activities.
A spokescouncil coordinates across working groups, and all working groups are responsible for ensuring that at least one member consistently participates. The spokescouncil meets monthly, and meetings are open to all members.
Decisions
The co-op operates through the practice of modified policy governance. Member decisions should seek to instill responsibility and trust in role-holders, while guiding them only through goal-setting and necessary constraints. Role-holders should act on the basis of their expertise and the co-op's values, as well as through consultation with members.
Policy. Any member may propose a policy in the co-op's forum. A proposal requires at least one vote from another member and stays open for 7 days. It passes by consent as long as it receives support from at least 60% of voting members. If a member raises a block—a reasoned objection that the proposal would seriously harm the cooperative—it requires 75% support. The proposal's author is always encouraged to revise the proposal based on feedback from other members and restart the process.
Charter. Changes to this charter require a proposal open for 14 days, along with approval by 75% of voting members, regardless of blocks. Policy decisions must not contradict the charter.
Disputes
Disputes are managed, not resolved. Members try to work things out directly with each other first. When necessary, disputes either among members or between a member and the co-op may be brought to the disputes working group—or, if one does not exist, the GM. Disputes that cannot be resolved within the co-op will be escalated to the parent legal entity, which may forward them to the relevant legal jurisdiction.
Agreements
This charter is the core governing document. Members are also bound by the following agrements:
The Inference Cooperative operates under the fiscal sponsorship of Metagov, and our governance is subject to Metagov's fiscal sponsorship policy. We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may rescind a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution, the cooperative may move to a different fiscal host. Changing fiscal hosts should occur through a revision of this Charter.
# Draft charter [2026.10.13-ntnsndr]
## Preamble
The Inference Cooperative exists to enable the people who use its AI infrastructure to co-govern it. We came together to build and run an alternative to extractive, profit-driven AI—a service owned and directed by its members, committed to pursuing ethical practices, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
## Values
In addition to the international [cooperative principles](https://www.ica.coop/en/cooperatives/cooperative-identity/), we share these basic commitments:
- **Direct democracy.** Our service is accountable to the people who use it, and they both can and should have an active role in shaping it.
- **Experimentation.** In order to learn together, we will maintain a spirit of exploration and experimentation, and be respectful of all members' levels of knowledge and expertise.
- **Sustainability.** We commit, wherever possible, to minimizing harm to natural ecologies and to promoting environmental justice.
- **Trustworthiness.** We seek to provide reliable service to one another and protect each other's privacy, in contrast to an industry widely characterized by surveillance.
## Roles
- **Members** are the heart of the co-op. All account holders currently [contributing monthly dues](https://opencollective.com/inference-cooperative) are considered members and may participate in decision-making and working groups. Members may be sanctioned or removed for violating co-op rules by the GM or a designated working group.
- A **general manager** oversees day-to-day operations and implements member decisions. The GM is accountable to members and may be replaced by them through a policy decision.
- **Working groups** make and carry out operational decisions within their own purview. They should be formally established, modified, and disbanded through a policy decision, and they are obligated, at minimum, to publish to members their membership, meeting schedule, and major activities.
- A **spokescouncil** coordinates across working groups, and all working groups are responsible for ensuring that at least one member consistently participates. The spokescouncil meets monthly, and meetings are open to all members.
## Decisions
The co-op operates through the practice of modified [policy governance](https://en.wikipedia.org/wiki/Policy_Governance). Member decisions should seek to instill responsibility and trust in role-holders, while guiding them only through goal-setting and necessary constraints. Role-holders should act on the basis of their expertise and the co-op's values, as well as through consultation with members.
**Policy.** Any member may propose a policy in the co-op's [forum](https://forum.inference.coop/). A proposal requires at least one vote from another member and stays open for 7 days. It passes by consent as long as it receives support from at least 60% of voting members. If a member raises a block—a reasoned objection that the proposal would seriously harm the cooperative—it requires 75% support. The proposal's author is always encouraged to revise the proposal based on feedback from other members and restart the process.
**Charter.** Changes to this charter require a proposal open for **14 days**, along with approval by 75% of voting members, regardless of blocks. Policy decisions must not contradict the charter.
## Disputes
Disputes are managed, not resolved. Members try to work things out directly with each other first. When necessary, disputes either among members or between a member and the co-op may be brought to the disputes working group—or, if one does not exist, the GM. Disputes that cannot be resolved within the co-op will be escalated to the parent legal entity, which may forward them to the relevant legal jurisdiction.
## Agreements
This charter is the core governing document. Members are also bound by the following agrements:
* [Code of Conduct](https://git.inference.coop/co-op/docs/src/branch/main/code-of-conduct.md)
* [Terms of Service](https://git.inference.coop/co-op/docs/src/branch/main/terms-of-service.md)
* [Privacy Policy](https://git.inference.coop/co-op/docs/src/branch/main/privacy-policy.md)
## Legal entity
The Inference Cooperative operates under the fiscal sponsorship of [Metagov](https://metagov.org/), and our governance is subject to [Metagov's fiscal sponsorship policy](https://metagov.pubpub.org/pub/7bj5ia5g/release/1). We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may rescind a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution, the cooperative may move to a different fiscal host. Changing fiscal hosts should occur through a revision of this Charter.
@lucio great point! Let's massage that language to avoid the contradiction. Does this help?
We regard disputes as an opportunity to learn and grow. Members should try to work things out directly with each other first. When necessary, disputes either among members or between a member and the co-op should be brought to the disputes working group—or, if one does not exist, the GM. Disputes that cannot be adequately addressed within the co-op may be escalated to the parent legal entity, which may in turn forward them to the relevant legal jurisdiction.
@lucio great point! Let's massage that language to avoid the contradiction. Does this help?
> We regard disputes as an opportunity to learn and grow. Members should try to work things out directly with each other first. When necessary, disputes either among members or between a member and the co-op should be brought to the disputes working group—or, if one does not exist, the GM. Disputes that cannot be adequately addressed within the co-op may be escalated to the parent legal entity, which may in turn forward them to the relevant legal jurisdiction.
This all makes sense to me! @lucio My thought with that initial “manage” language came from the finding that past a certain point, high temperature conflicts cannot ever be truly resolved, but people still need to be able to work together. So at that point the goal is not to resolve the conflict because the cat is out of the bag, so to speak. The goal is to create a plan to manage the conflict and prevent it from boiling over.
This all makes sense to me! @lucio My thought with that initial “manage” language came from the finding that past a certain point, high temperature conflicts cannot ever be truly resolved, but people still need to be able to work together. So at that point the goal is not to resolve the conflict because the cat is out of the bag, so to speak. The goal is to create a plan to manage the conflict and prevent it from boiling over.
I added some language on policy governance to stress that the purpose of member decisions is not to micromanage operations but to establish policy guidelines for role-holders.
The co-op operates through the practice of modified policy governance. Member decisions should seek to instill responsibility and trust in role-holders, while guiding them only through goal-setting and necessary constraints. Role-holders should act on the basis of their expertise and the co-op's values, as well as through consultation with members.
I added some language on policy governance to stress that the purpose of member decisions is not to micromanage operations but to establish policy guidelines for role-holders.
> The co-op operates through the practice of modified [policy governance](https://en.wikipedia.org/wiki/Policy_Governance). Member decisions should seek to instill responsibility and trust in role-holders, while guiding them only through goal-setting and necessary constraints. Role-holders should act on the basis of their expertise and the co-op's values, as well as through consultation with members.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
What's the idea?
With 34 current members, we are approaching the 50-member threshold for transitioning to a democratic governance model.
Why does it matter?
We are a co-op, time to practice it!
Who is it for?
All members
How might it work?
Add ideas below!
Centralized approach: (closest to current practice)
Board approach: (conventional co-op model)
Horizontalist approach: (similar to Social.coop)
Some potential directions to react to — each with a trade-off. These are starting points for discussion, not a menu to pick one from; elements can combine.
1. Consensus-in-Loomio (evolution of current practice)
Keep informal consensus for routine decisions; require a Loomio proposal with a supermajority (e.g. 66–75%) for significant ones (pricing, providers, governance changes itself). Define "significant" by a threshold list the members can amend.
2. Consent-based (sociocracy-style)
Decisions pass unless a member raises a reasoned objection ("safe to try"). Circles (e.g. tech, membership, outreach) hold their own domains; a general circle coordinates.
3. Elected steward(s) + open assembly
Members elect one or more stewards for fixed terms (e.g. 6 months, renewable once) who handle day-to-day operational calls; big decisions stay with the assembly. Full transparency via the forum and open books.
4. Delegated/liquid fragments
Members can vote directly on any proposal or delegate their vote to a trusted member per topic area (Loomio supports proxies; liquid-style tooling exists if needed).
5. Weighted or tiered participation
All members vote, but with weights (e.g. by tenure, or by contribution level). Careful: this cuts against our flat sliding-scale ethos and could create classes of membership — listed mainly so we consciously reject or adapt it.
Two structural notes regardless of model:
Initial drafting from Dani:
Draft: Next-stage Charter (for discussion)
A synthesis of the kickoff conversation, this thread, and Dani's interim-policies draft. This is a proposal to react to — not adopted language. It deliberately leaves the current
charter.mduntouched until members consent to it.What this is
A governance charter for the Inference Cooperative, designed for the transition at ~50 members. It is not a legal document (bylaws are separate). It separates three things that are easy to tangle:
The aim is a model that is participatory and accountable without "death by community committee," and that keeps the door to contribution open while keeping power visible.
Values
Our shared values are an open question. We commit to holding a values workshop within the first two months of this charter taking effect, and to recording the result here. (Education and learning-by-doing are named principles of the cooperative; "always have an active experiment or challenge" is a proposed cultural norm.)
Membership
Decision-making: consent
Working groups (do-ocracy, horizontal)
Coordination: delegates and a spokescouncil
Conflict
Labor & compensation
open-questions.md, not here.The "super-governor" risk
The co-op must not be able to move only as fast as its technical team, nor let skilled maintainers accrue unaccountable power. The answer is publication, not hierarchy: everything a group does is visible, and any member can trigger a review. A maintainer accruing power is fine; a maintainer doing so invisibly is the failure this charter guards against.
Transition at 50 members
How this passes
This takes effect if nobody objects within 14 days. The full draft and the list of open questions live in the docs repository.
A couple thoughts:
To be clear, this has not been formally proposed. That would happen in the forum.
More:
Working groups should be able to make operational decisions within their purview. For instance, the infra group should be able to swap models or make platform changes as necessary. What needs to go through the decision process are policies, which impose constraints or requirements on members' freedom to contribute.
An important value that came up in the first meeting is experimentation. We want to make sure to encourage that as a core value, while also ensuring reliable infrastructure.
A further thought: The document needs to acknowledge that our governance is subject to the Metagov fiscal sponsorship policy and our agreement with them as the fiscal host: https://metagov.pubpub.org/pub/7bj5ia5g/release/1 This should involve a commitment to honor that agreement and to operate in a spirit that seeks to avoid causing harm to our fiscal host. If that commitment is at odds with the co-op's evolution, it should move to a different fiscal host.
This might also involve recognizing that Metagov can veto a decision if it is deemed at odds with the organization's policies.
I much prefer that a proposal requires a 2nd. >= 2 people should have some idea what's going on for each change.
OTOH, that would perhaps bring progress to a screeching halt. I expect nobody has been around to work with the general manager on most things.
Another possibility: notice-and-appeal. GM (or other proposer?) posts substantive changes to a well-known place. Normal things would stay queued for 2 to 3 days, but out of expedience, the proposer can go ahead at their discretion. If a member appeals within a week or two, it rolls back until the next meeting.
I guess we want to get the GM out of the center...
Is rollback possible in general? What if the change involved spending money?
I dunno... I guess I'm not coming to a clear conclusion. Just thinking out loud.
@dckc Thanks for this. I suppose a big question is whether we should retain the GM role or not. But it might be a good idea to retain it a while longer, but to take some more serious steps toward member participation and power.
I think the most (small-c) conservative approach, on the theory of reflecting actual practice in written governance design, would be to retain the GM role but have some increased member checks and options. For instance, a member proposal process could become binding on the GM. It could follow the consent model proposed above, plus the seconding approach that dckc suggests. We should define precisely how it should work in Loomio.
👍
Charter (revised draft, for discussion)
A proposal to react to, not adopted language.
Preamble
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter sets out how we make decisions together. It is not a legal document; our bylaws are separate. Where this charter is at odds with our agreement with our fiscal host, Metagov, that agreement prevails, and we act in a spirit that avoids harm to Metagov. If this commitment conflicts with the cooperative's evolution, we seek a different fiscal host.
Values
We govern by these commitments:
Roles
Decisions
Policy. Any member may propose a policy. A proposal requires a second from another member, then stays open in Loomio for 14 days. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote of three-quarters of all members.
Charter. Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a formal proposal, a longer comment period, and a three-quarters supermajority vote of all members. The constitutional structure changes more deliberately than ordinary policy.
Conflict and conduct
Disagreements are managed, not resolved. Members try to work things out directly first; a rotating conflict group of facilitators steps in if needed. Violations of our Code of Conduct are handled under this process.
Institutional members
Institutional members may hold a vote and a voice, though never an outsized one, as policy determines.
Some modifications:
New draft incorporating your Oct 10 modifications. Summary of what changed from the previous version:
Open question I'd flag: the charter-amendment threshold for leaving Metagov — I set it at the charter-amendment bar (14 days + 3/4 of voting members). Comfortable, or should that be a different threshold?
Draft charter (v3, for discussion)
A proposal to react to, not adopted language. This revision applies Nathan's Oct 10 modifications: Disputes working group, no separate bylaws, "Legal entity" section at the end, working groups left undefined in the charter, 7-day policy window (14 for charter changes), supermajority of voting members only, monthly spokescouncil, and references to the Code of Conduct / ToS / privacy policy.
Preamble
The Inference Cooperative exists so that the people who use its AI infrastructure also govern it. We came together to build and run an alternative to extractive, corporate AI — an inference service owned and directed by its members, hosted on renewable energy, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
Values
We govern by these commitments:
Roles
Decisions
Policy. Any member may propose a policy in Loomio. A proposal requires a second from another member, then stays open for 7 days. It passes by consent: if no member raises a block — a reasoned objection that the proposal would harm the cooperative — it becomes policy and is binding on the GM. An objection that is not a block is an invitation to improve the proposal, not a veto. A proposal that draws repeated blocks may be revised and re-proposed; after three failed attempts it may go to a vote.
Voting. A vote that goes to the members passes by a three-quarters supermajority of voting members — those who cast a ballot. We do not hold ourselves hostage to members who do not participate.
Charter. Changes to this charter — its roles, its decision rules, the GM role itself, and its values — require a proposal open for 14 days, followed by a three-quarters supermajority vote of voting members. The core governing document changes more deliberately than ordinary policy.
Disputes
Disagreements are managed, not resolved. Members try to work things out directly first. The Disputes working group steps in when they cannot — for conflict between members and for disagreements with co-op management alike, including disputes over whether the GM has implemented a member decision. The Disputes group's process is defined by policy. Violations of the Code of Conduct are handled under this process.
Institutional members
Institutional members may hold a vote and a voice, though never an outsized one; their specific standing is set by policy.
Related documents
This charter is the core governing document. Members are also bound by the Terms of Service and Privacy Policy, and the community's conduct standards live in the Code of Conduct. There are no separate bylaws: as a fiscally sponsored project, the charter is our founding and governing text.
Legal entity
The Inference Cooperative operates under the fiscal sponsorship of Metagov, and our governance is subject to Metagov's fiscal sponsorship policy and our agreement with them. We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may veto a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution — if the co-op outgrows its host or the relationship stops serving members — the cooperative may move to a different fiscal host. Leaving Metagov is a decision of the same weight as amending this charter, taken through the charter-amendment process above.
Draft charter [2026.10.13-ntnsndr]
Preamble
The Inference Cooperative exists to enable the people who use its AI infrastructure to co-govern it. We came together to build and run an alternative to extractive, profit-driven AI—a service owned and directed by its members, committed to pursuing ethical practices, and governed cooperatively. This charter is our core governing document. It sets out how we make decisions together.
Values
In addition to the international cooperative principles, we share these basic commitments:
Roles
Decisions
The co-op operates through the practice of modified policy governance. Member decisions should seek to instill responsibility and trust in role-holders, while guiding them only through goal-setting and necessary constraints. Role-holders should act on the basis of their expertise and the co-op's values, as well as through consultation with members.
Policy. Any member may propose a policy in the co-op's forum. A proposal requires at least one vote from another member and stays open for 7 days. It passes by consent as long as it receives support from at least 60% of voting members. If a member raises a block—a reasoned objection that the proposal would seriously harm the cooperative—it requires 75% support. The proposal's author is always encouraged to revise the proposal based on feedback from other members and restart the process.
Charter. Changes to this charter require a proposal open for 14 days, along with approval by 75% of voting members, regardless of blocks. Policy decisions must not contradict the charter.
Disputes
Disputes are managed, not resolved. Members try to work things out directly with each other first. When necessary, disputes either among members or between a member and the co-op may be brought to the disputes working group—or, if one does not exist, the GM. Disputes that cannot be resolved within the co-op will be escalated to the parent legal entity, which may forward them to the relevant legal jurisdiction.
Agreements
This charter is the core governing document. Members are also bound by the following agrements:
Legal entity
The Inference Cooperative operates under the fiscal sponsorship of Metagov, and our governance is subject to Metagov's fiscal sponsorship policy. We commit to honor that agreement and to operate in a spirit that avoids causing harm to our fiscal host. Metagov may rescind a decision it deems at odds with its policies.
If that commitment is ever at odds with the cooperative's evolution, the cooperative may move to a different fiscal host. Changing fiscal hosts should occur through a revision of this Charter.
What does it mean to manage but not resolve disputes? The later phrase "Disputes that cannot be resolved within the co-op" seems to contradict it.
@lucio great point! Let's massage that language to avoid the contradiction. Does this help?
Yeah, that makes more sense to me
This all makes sense to me! @lucio My thought with that initial “manage” language came from the finding that past a certain point, high temperature conflicts cannot ever be truly resolved, but people still need to be able to work together. So at that point the goal is not to resolve the conflict because the cat is out of the bag, so to speak. The goal is to create a plan to manage the conflict and prevent it from boiling over.
I added some language on policy governance to stress that the purpose of member decisions is not to micromanage operations but to establish policy guidelines for role-holders.
Approved! https://forum.inference.coop/d/UZwFCgsi/first-governance-transition/7