In 1937, Ronald Coase explained why firms exist: the transaction costs of coordinating activity through markets (finding trading partners, negotiating contracts, enforcing agreements) exceed the internal costs of doing it within a firm. Reduce transaction costs, and the optimal firm size shrinks. Reduce them to near zero, through smart contracts, AI agents, and programmable protocols, and the firm dissolves into a mesh: a continuously reconfiguring network of autonomous participants coordinating through protocols rather than management hierarchies.
The Coasean Logic
Coase's insight was counterintuitive: markets are efficient, but using markets has a cost. Finding a supplier, negotiating terms, writing a contract, monitoring performance, resolving disputes. Each step carries "transaction costs" that make it cheaper, beyond a certain threshold, to bring the activity inside a firm and manage it through employment relationships and hierarchical coordination.
The 1991 Nobel Prize was awarded to Coase for this observation, which had been hiding in plain sight for 54 years. His paper "The Nature of the Firm" is one of the most cited in economics because it answers the question every economist should have asked but didn't: if markets are so efficient, why do firms exist at all?
The answer: because markets are not free. They carry friction. And that friction determines the boundary between what happens inside organizations and what happens between them.
Every reduction in transaction costs shifts the boundary between firm and market. The internet reduced search costs (you can find a supplier in Shenzhen in seconds). Email reduced negotiation costs (a contract discussion that took weeks of postal exchange now takes hours). E-commerce reduced distribution costs (a single seller can reach global markets without physical retail). Each reduction made it viable to outsource activities that were previously internal, and the gig economy, contractor culture, and platform marketplaces followed.
Transaction Cost Collapse
How protocols eliminate each category of Coasean friction
Coase, R. (1937). "The Nature of the Firm." Economica. Protocol-era cost estimates based on DeFi operational analysis.
Smart contracts, AI agents, and cryptographic protocols reduce transaction costs by another order of magnitude. The key categories of cost and their protocol-era equivalents:
- Search costs: Zero. AI agents discover counterparties algorithmically across global liquidity pools.
- Negotiation costs: Zero. Terms are defined in protocol parameters; participants accept or reject.
- Contract writing costs: Near zero. Smart contracts are deployed once and reused millions of times.
- Monitoring costs: Zero. Execution is automatic and verifiable on-chain.
- Enforcement costs: Zero. Smart contracts are self-enforcing; mathematical guarantees replace legal ones.
- Dispute resolution costs: Near zero. Outcomes are deterministic by construction; where ambiguity exists, decentralized arbitration protocols like Kleros provide crowdsourced resolution.
At these costs, the firm as an organizing principle becomes optional for an increasing range of activities.
Decentralized Finance as Proof of Concept
DeFi is the first large-scale implementation of mesh economics, and the results are extraordinary.
Uniswap, a decentralized exchange protocol, surpassed $3 trillion in cumulative trading volume in May 2025, becoming the first DEX to reach this milestone. It executes approximately $2 billion+ in daily trading volume and commands roughly 36% of the total DEX market share, all with no employees, no central operator, and no physical office. The protocol is a set of smart contracts that automatically match buyers and sellers, determine prices through an algorithmic market maker (the constant product formula x*y=k), and distribute fees to liquidity providers. The entire value chain of a financial exchange, listing, matching, clearing, and settlement, is executed by code.
The mesh economy is what emerges when coordination costs approach zero. The firm, which Coase showed exists to minimize transaction costs, becomes unnecessary when those costs are eliminated by automation. What remains is a protocol: a set of rules that participants follow voluntarily because the rules are transparent, enforcement is automatic, and exit is free.
Aave, a decentralized lending protocol, manages billions in deposits and loans without a single loan officer, credit committee, or compliance department. Interest rates adjust automatically based on supply and demand for each asset. Collateral is managed by smart contracts that liquidate positions automatically when collateral ratios fall below threshold. The protocol has processed over $50 billion in cumulative lending volume since inception.
MakerDAO (now Sky) manages one of the largest stablecoin systems in crypto, with DAI maintaining its dollar peg through an automated system of collateralized debt positions, liquidation bots, and governance votes. The system has operated continuously since 2019, through multiple market crashes, without the intervention of a single human risk manager.
These protocols demonstrate that complex economic coordination, the kind that historically required firms with hundreds of employees, regulatory licenses, and physical infrastructure, can be performed by software executing on a public blockchain.
The DAO Treasury Experiment
Decentralized Autonomous Organizations (DAOs) extend the mesh model from execution to governance.
DeepDAO data shows that DAO treasuries reached approximately $35 billion in assets under management at their peak in late 2024. By late 2025, market conditions had reduced this to approximately $13.6 billion, a figure that itself reveals an important structural insight: much of the "treasury" value was held in native governance tokens, making it circular (the DAO is "rich" in its own token, which derives value from the DAO's perceived value). Liquid, diversified treasury assets were closer to $12.3 billion.
DAO Treasury Concentration
~$13.6B in total DAO treasuries, heavily concentrated (late 2025)
Source: DeepDAO (late 2025 snapshot). Treasury values include liquid assets; native token holdings are market-dependent. Top 5 DAOs hold ~60% of total.
The concentration is extreme: the top 5 DAOs (Uniswap, Optimism, Arbitrum, Mantle, ENS) hold over 60% of all collective DAO treasury assets. This concentration mirrors traditional finance, where a small number of institutions manage the majority of capital. The mesh may distribute execution, but capital still concentrates.
What DAOs have successfully demonstrated is a new governance primitive: token-weighted voting on protocol parameters, treasury allocations, and strategic decisions. Uniswap governance has processed hundreds of proposals affecting billions in protocol activity. Optimism's Collective has distributed over $200 million in retroactive public goods funding through a novel bicameral governance structure.
The limitations are equally instructive. Voter participation in most DAOs hovers between 3-10% of token holders. "Whale" token holders (large institutional or early investors) can dominate outcomes. And governance through voting produces slow, consensus-driven decision-making that struggles with operational urgency. The mesh governs well at the strategic level (setting parameters, allocating capital) and poorly at the operational level (responding to crises, executing complex initiatives).
The Energy Mesh
The mesh model extends beyond finance into physical infrastructure.
Rooftop solar panels generate electricity at the point of consumption. When generation exceeds consumption, the surplus can be sold to neighbors: peer-to-peer energy trading. Power Ledger built blockchain-based platforms for local energy markets, where prosumers (producer-consumers) trade electricity directly without a utility intermediary.
The economics are straightforward: a utility charges consumers a retail rate and pays producers a wholesale rate. The spread is the utility's margin. A peer-to-peer energy mesh eliminates the spread by connecting producers to consumers directly, at a clearing price determined by local supply and demand. The utility becomes infrastructure (maintaining the physical grid) rather than intermediary (buying and reselling energy).
Germany's Energiewende has produced a grid where approximately 52% of electricity was generated from renewables in 2023 (Fraunhofer ISE), much of it distributed. The coordination challenge, matching intermittent generation (solar, wind) with variable demand, is increasingly managed by software rather than centralized dispatch.
Australia's Power Ledger has deployed peer-to-peer energy trading in multiple markets across Australia, Thailand, Japan, and Europe. The platform allows participants to set their own prices and trade surplus energy directly, reducing costs for buyers and increasing revenue for sellers compared to feed-in tariffs.
Energy Mesh Economics
Traditional utility vs. peer-to-peer energy mesh
Pricing data from Power Ledger deployments and European feed-in tariff schedules. Mesh protocol fees based on blockchain transaction costs.
The Labor Mesh
The gig economy is an early, imperfect version of the labor mesh.
Uber, DoorDash, and Upwork reduce the transaction costs of matching labor supply with demand. But they introduce a new intermediary: the platform itself, which extracts 20-30% of transaction value. This is not a mesh. It is a centralized marketplace with mesh aesthetics.
The distinction matters. A true mesh eliminates the intermediary. A platform marketplace replaces one intermediary (the traditional employer) with another (the platform operator). The worker trades an employment relationship for a platform dependency, often with less stability, fewer benefits, and no equity in the platform they help build.
A true labor mesh would coordinate labor through protocols rather than platforms: open-source matching algorithms, reputation systems that are portable across platforms (not locked to a single marketplace), and smart contract payment escrow that releases funds automatically upon verified completion.
Braintrust, a decentralized talent network, attempts this model: talent sets rates, clients post jobs, matching is algorithmic, and the network is governed by token-holding participants rather than a corporate employer. The platform take rate is near zero, compared to Upwork's 10-20% service fee.
Three Models of Economic Organization
Firm → Platform → Mesh: the Coasean progression
Framework adapted from Coase (1937) and Williamson (1985). Protocol-era data from DeFi protocol documentation and on-chain analysis.
The structural difference is governance: in a platform marketplace, the platform unilaterally sets take rates, resolves disputes, and determines algorithm parameters. In a mesh protocol, these decisions are made through governance mechanisms that include the participants themselves. The platform cannot unilaterally raise its take rate because the take rate is a protocol parameter that requires governance approval to change.
AI Agents as Mesh Participants
The convergence of AI agents with mesh protocols creates a new category of economic actor: autonomous software that participates in mesh coordination without human intervention.
An AI agent can discover a counterparty in a DeFi liquidity pool, negotiate terms algorithmically, execute a smart contract, and settle the transaction, all in seconds, all without human approval. The agent is a mesh participant that reduces the last remaining transaction cost: the human cognitive cost of evaluating options and making decisions.
Protocols like Anthropic's MCP and Google's A2A are building the communication layer for agent-to-agent coordination. Combined with mesh protocols for execution and settlement, this creates the possibility of fully autonomous economic coordination: agents discovering opportunities, negotiating terms, executing transactions, and allocating capital, all within protocol-defined boundaries, all without human intervention.
The mesh economy, originally conceived as a network of human participants coordinating through protocols, is evolving into a network of agents coordinating through protocols, with humans setting the parameters and monitoring outcomes.
The Limits of the Mesh
The mesh model has structural limitations that determine its domain of application.
Coordination that requires judgment. Activities that require contextual judgment, creative synthesis, emotional intelligence, or ethical reasoning cannot (yet) be reduced to protocol execution. Strategic planning, crisis management, mentorship, and conflict resolution require the kind of human processing that smart contracts cannot perform. AI agents are narrowing this gap, but the gap remains significant for high-stakes, ambiguous situations.
Activities that require trust beyond verification. A smart contract can verify that a payment was made. It cannot verify that a service was performed with care, that a relationship was maintained with integrity, or that a team member acted in good faith under ambiguous circumstances. The mesh is effective for verifiable transactions and less effective for trust-dependent relationships.
Regulatory requirements. Many economic activities require licensed intermediaries by law: banking, insurance, securities trading, healthcare. DeFi is increasingly integrating compliance (KYC/AML requirements, regulatory reporting), but the regulatory frameworks were designed for intermediated markets, not for protocols. The adaptation is ongoing and uneven.
Capital concentration. As DAO treasury data shows, even in mesh structures, capital concentrates. The top 5 DAOs hold over 60% of all treasury assets. Governance power follows capital: larger token holders have more votes. The mesh distributes execution but does not automatically distribute wealth or power.
The mesh economy promises to eliminate intermediaries. But the protocols that enable the mesh are themselves intermediaries, albeit transparent, automated ones. Uniswap charges a 0.3% fee on every trade. Aave takes a spread on every loan. The fee is lower than a traditional intermediary's, but it is not zero. And the governance of these protocols is concentrated among early investors and large token holders, replicating the power structures the mesh was designed to replace. The mesh does not eliminate hierarchy. It compresses hierarchy into protocol parameters and governance votes.
The Trajectory
The likely trajectory is hybrid: mesh coordination for verifiable, high-frequency, protocol-compatible transactions, and traditional organizational structures for judgment-intensive, relationship-dependent, and highly regulated activities.
The mesh does not replace the firm. It shrinks the firm to its irreducible core: the activities that genuinely require human judgment. Everything else, every verifiable transaction, every rule-based coordination, every parameter-driven allocation, migrates to protocols.
Coase predicted this implicitly. If firms exist because of transaction costs, and transaction costs approach zero, then firms shrink. The mesh economy is the Coasean endpoint: coordination without firms, value exchange without intermediaries, governance without management.
What Coase did not predict was that the agents doing the coordinating would themselves be software. The mesh is not just a network of humans trading through protocols. It is increasingly a network of AI agents operating within protocol-defined boundaries, executing transactions at speeds and volumes no human organization could match.
The mesh economy extends Coase's theory to its logical conclusion: as transaction costs approach zero through smart contracts and AI agents, the optimal firm size shrinks. DeFi demonstrates this at scale: Uniswap executes $2B+ daily volume with zero employees, surpassing $3 trillion cumulative. Aave manages billions in lending with no loan officers. DAO treasuries peaked at $35 billion (DeepDAO, 2024), though concentration remains extreme, with the top 5 DAOs holding 60%+ of assets. The mesh extends to energy (Germany's 52% renewable grid, Power Ledger's peer-to-peer trading) and labor (Braintrust's near-zero take rate vs. platforms' 20-30%). AI agents are becoming mesh participants, coordinating through MCP and A2A protocols. Structural limits apply: judgment-intensive activities, trust beyond verification, and regulatory requirements still require organizational structures. The mesh does not eliminate hierarchy. It compresses hierarchy into protocol parameters and governance votes.