The Mozi Cooperative

Abstract

The dominant model of AI infrastructure development...venture-backed, corporate-owned, extraction-oriented...concentrates both economic returns and governance power in a vanishingly small number of entities. This paper argues that cooperative ownership of AI infrastructure is not merely ethically preferable but economically superior for the long-term health of the technology. Named for Mozi (c. 470-391 BCE), the Chinese philosopher who advocated jian ai (universal love) and pragmatic cooperation over hierarchical power, we propose a cooperative economic framework for decentralized AI networks. The framework addresses three failures of the venture model: misaligned incentives between capital providers and value creators; extraction of creative and data labor without proportional compensation; and concentration risk that makes the entire AI ecosystem fragile. We formalize cooperative mechanisms for model ownership, inference revenue sharing, and infrastructure governance, drawing on cooperative economics literature, platform cooperativism theory, and game-theoretic analysis. The framework is designed for implementation on the Citrate Network (Paper I), where LoRA adapter contributions, validator operations, and data provisioning are the measurable units of cooperative value creation.

Keywords: cooperative capitalism, AI infrastructure, venture capital critique, platform cooperativism, decentralized ownership, Mozi, jian ai, cooperative game theory

1. The Problem: Venture Capital and the Enclosure of AI

1.1 The Extraction Thesis

The venture capital model of AI development operates on a simple exchange: capital for equity. Investors provide compute budgets necessary to train large models, and in return receive ownership shares. The data...the raw material...comes from individuals who receive no ownership stake: internet users whose text and images constitute training corpora; content creators whose works are ingested without licensing; and data laborers who clean and curate datasets for wages bearing no relationship to the value they create [2, 3]. Zuboff [4] documented the surveillance capitalism model where user behavioral data is extracted and sold without meaningful consent. The AI era intensifies this: foundation model companies monetize the generative capacity derived from absorbed human output.

1.2 Concentration Risk

As of early 2025, three cloud providers control approximately 65% of global compute [6]. Fewer than ten organizations produce the most capable foundation models. GPU supply is constrained by a single manufacturer. This creates systemic fragility: a policy decision by one cloud provider or a supply disruption from one chip manufacturer cascades across the entire ecosystem. The Harvard Ash Center [8] identifies this concentration as a governance failure requiring alternative ownership structures.

1.3 The Open-Source Exception

Open-source AI...Llama, Mistral, Stable Diffusion...has been the only consistent exception to extraction. Contributors receive usable tools, transferable skills, and community reputation. But open-source faces a sustainability crisis: training frontier models costs $10-100+ million, and projects depend on corporate sponsors whose interests may diverge from the community’s. When Stability AI’s corporate struggles threatened Stable Diffusion, the community had no governance mechanism to ensure continuity [9]. The cooperative model addresses this gap.

2. The Mozi Framework: Universal Love as Economic Design

2.1 Philosophical Foundation

Mozi (c. 470-391 BCE) advocated jian ai (兼爱)...commonly translated as “universal love” but more precisely understood as “impartial care”: caring equally for all people rather than privileging one’s in-group at others’ expense [10]. Mozi was also a pragmatist and engineer. Applied to AI infrastructure: value created by the network should benefit all contributors proportionally, not privilege early investors disproportionately. And the cooperative must be economically viable, not merely ideologically appealing.

2.2 Core Principles

Contribution-proportional ownership. Ownership is earned through measurable contributions: running validator nodes, providing training data, developing LoRA adapters, building applications, provisioning compute. This inverts the VC model where capital...the easiest resource for the already-wealthy to deploy...determines ownership.

Revenue flows to contributors. Inference fees, bridge fees, and marketplace commissions are distributed algorithmically to the contributors who make services possible. Distribution is transparent and on-chain.

Governance follows contribution. Voting power is weighted by contribution history, not token holdings alone. A validator operating honestly for a year has more governance weight than a speculator who purchased tokens yesterday. This prevents plutocratic capture...the failure mode of most DAO governance systems [11].

Knowledge sharing is value creation. The Mentorship Protocol (Paper III) formalizes knowledge transfer as a first-class economic activity. When expertise is encoded as a LoRA adapter that improves a weaker node, the adapter creator receives attribution and revenue share. This creates a positive-sum game: sharing expertise increases total network value, and the sharer is compensated [12].

3. Economic Mechanisms

3.1 Contribution Taxonomy

Table 1. Contribution Types and Value Metrics

Contribution Type

Measurable Metric

Reward Mechanism

Citrate Implementation

Validation

Blocks produced, uptime, blue score

Block rewards (SALT)

GhostDAG mining rewards

Model Hosting

Inference served, accuracy

Inference fee share

MCP marketplace revenue

Adapter Creation

LoRA adapters, adoption rate

Royalty on adopters’ earnings

LoRAFactory attribution

Data Provision

Quality score, volume, uniqueness

Data licensing revenue

On-chain data registry

App Development

dApp usage, transactions

Developer fee share

LVM contract metrics

Bridge Infra

$SNAP NFT ownership, operation

Bridge fee share

FeeDistributor pro-rata

Governance

Votes cast, proposals authored

Incentive bonus

GovernorVault tracking

3.2 Revenue Distribution Model

Base layer: SALT block rewards (50% of 1B total supply over the mining schedule), distributed to validators proportional to blocks produced.

Inference layer: 70% of inference fees to model hosts, 15% to adapter creators whose LoRAs improved the serving model, 10% to the DAO treasury, 5% to data providers.

Bridge layer: 70% of bridge fees to $SNAP NFT holders, 20% to treasury, 10% to Chainlink node operators [13].

Every participant in the value chain receives compensation proportional to contribution. No participant is purely extractive.

3.3 Game-Theoretic Properties

In a cooperative game theory formulation [14], each contributor’s Shapley value increases as the network grows, because each contribution’s marginal value depends on the contributions it interacts with. A model host’s value increases with better adapters. An adapter creator’s value increases with more hosts. A validator’s value increases with more application demand. The cooperative converts zero-sum extraction into positive-sum cooperation...Mozi’s jian ai expressed as incentive alignment.

4. Comparison to Existing Models

Table 2. Economic Model Comparison

Dimension

VC-Backed AI

Open-Source AI

Bittensor ($TAO)

Mozi Cooperative

Ownership

Investors + founders

None (public domain)

Token holders

Contributors (weighted)

Revenue

Shareholders

None

Miners via subnets

All contributor types

Data compensation

None

None

Indirect

Direct via licensing

Governance

Board of directors

Maintainer consensus

Senate + token vote

Contribution-weighted DAO

Sustainability

VC rounds

Corporate sponsors

Token emissions

Usage fees

Concentration risk

Very high

Medium

Medium

Low (by design)

Bittensor [15] is the closest existing implementation, with subnet-based competition and token-incentivized contribution. However, Bittensor concentrates rewards in mining and validation without addressing the broader contributor taxonomy...data providers, adapter creators, application developers, and bridge operators receive no direct compensation. The Mozi Cooperative extends this model by recognizing value creation across the entire knowledge lifecycle.

5. Implementation on the Citrate Network

On-chain contribution tracking. Every block produced, inference served, adapter created, and transaction processed is recorded in the BlockDAG. Contribution metrics are deterministically computable from chain history, requiring no trusted third party.

Algorithmic revenue distribution. Smart contracts on the Lattice Virtual Machine implement the multi-layer revenue split. RewardDistributor, LoRAFactory, and MCP marketplace contracts enforce cooperative economics without human intermediation.

Contribution-weighted governance. GovernorVault weights voting power by a composite of blue score (validation history), adapter adoption (knowledge contribution), and participation streak (governance engagement).

Immutable cooperative history. The checkpoint chain provides an irreversible record of every contribution, distribution, and governance decision. The cooperative’s history is auditable by anyone, at any time.

6. Honest Limitations

The cooperative model faces real challenges that this paper does not solve. Free-rider problems: contributors who do minimal work but accumulate governance weight through longevity. Sybil attacks: a single entity creating multiple identities to claim disproportionate contribution credit. Governance paralysis: contribution-weighted voting may be slower than executive decision-making, creating competitive disadvantage. Cold start: the cooperative needs sufficient initial participation to generate meaningful revenue, creating a bootstrapping problem identical to any marketplace. These are not unique to cooperative AI...they are the known challenges of cooperative economics [16, 19]. We propose that on-chain transparency and algorithmic enforcement mitigate some of these challenges, but we do not claim to have solved them.

7. Relationship to the Gradient Papers Series

Paper I (Citrate Technical Paper) provides the technical primitives: on-chain contribution tracking, smart contract revenue distribution, and contribution-weighted governance are all implementable within the Citrate architecture.

Paper III (Mentorship Protocol) provides the organizational learning theory. Knowledge sharing as value creation (Principle 4) is the Mentorship Protocol’s core thesis operationalized as economics.

Paper VI (Memetic Money Portal) implements cooperative bridge ownership. $SNAP NFT holders are not passive investors but active infrastructure operators...the Mozi framework applied to cross-chain infrastructure.

Paper VIII (BR1J Constitution) implements cooperative governance. The DAO governance framework is the legal and procedural realization of the Mozi Cooperative’s economic philosophy.

8. Conclusion

The venture capital model has produced remarkable AI capability at the cost of extreme concentration, systematic extraction, and governance failure. The Mozi Cooperative proposes an alternative: infrastructure owned by contributors, revenue flowing to value creators, governance weighted by engagement, and knowledge sharing as the primary mechanism of value creation. The cooperative mechanisms described here are implementable on existing blockchain infrastructure, enforceable through smart contracts, and auditable through on-chain transparency. What remains is the conviction to build infrastructure that serves the many rather than extracting from them.

References

[1] Klosowski, L. (2026). Citrate: Protocol Specification. Gradient Papers No. I.

[2] Perrault, R., et al. (2024). The AI Index 2024 Annual Report. Stanford HAI.

[3] Zuboff, S. (2019). The Age of Surveillance Capitalism. Profile Books.

[4] Zuboff, S. (2019). Surveillance capitalism and collective action. New Labor Forum, 28(1).

[5] Klosowski, L. (2024). Venture and AI: to Whom? Cnidarian Foundation Essay.

[6] Synergy Research Group. (2025). Cloud Infrastructure Service Spending.

[7] OpenAI. (2024). Infrastructure is Destiny. Policy Paper.

[8] Hubbard, S. (2025). Cooperative Paradigms for AI. Harvard Ash Center.

[9] Stability AI. (2024). Corporate restructuring. Industry analysis.

[10] Fraser, C., & Campagna, D. (2003). Mozi: Basic Writings. Columbia University Press.

[11] Simoncic, K., & Jerele, T. (2023). Democratizing AI Governance. Palgrave Macmillan.

[12] Klosowski, L. (2026). The Mentorship Protocol. Gradient Papers No. III.

[13] Klosowski, L. (2026). The Memetic Money Portal. Gradient Papers No. VI.

[14] Shapley, L. S. (1953). A value for n-person games. Contributions to the Theory of Games II.

[15] Rao, J. (2021). Bittensor: A peer-to-peer intelligence market.

[16] Scholz, T. (2016). Platform Cooperativism. Rosa Luxemburg Stiftung.

[17] Scholz, T., & Tortorici, S. (2025). 5 Ways Cooperatives Can Shape AI. HBR.

[18] Piketty, T. (2014). Capital in the Twenty-First Century. Harvard University Press.

[19] Ostrom, E. (1990). Governing the Commons. Cambridge University Press.

Appendix A: Cross-Paper Parameter Consistency

Parameter

Value

Source

SALT total supply

1,000,000,000

Paper I, Section 5

Mining rewards

50% of supply over 10 years

Paper I, Section 5

Inference fee split

70% host / 15% adapter / 10% treasury / 5% data

This paper, Section 3.2

Bridge fee split

70% SNAP holders / 20% treasury / 10% oracle ops

This paper, Section 3.2

Contribution-weighted governance

Blue score + adapter adoption + participation

This paper, Section 5

BFT committee

100 validators, 67 signatures

Paper I, Section 2.3

Cnidarian Foundation • larry@cnidarianfoundation.org