Mycel

[CONVICTION]

Mycel is a protocol suite and runtime for verifiable coordination and permissionless production in the physical world. It is the concrete engineering answer to a question the Mesocosm thesis raises abstractly: how do you build an open network where microfactories, microclinics, microschools, microfarms, and microgrids can coordinate and get paid without a centralized platform?

The name comes from mycelium -- the underground fungal network that coordinates resource allocation across a forest without central authority. The biological parallel is precise: mycorrhizal networks verify contribution (phosphorus delivered), discover capability (which root tips are active), coordinate exchange (carbon for minerals), and settle value (proportional allocation based on verified delivery). Mycel is the engineered equivalent for human physical production.

What Mycel Is

Mycel is three things, layered:

Mycel is the protocol suite: a minimal kernel (Mycel-core) plus a domain extension mechanism (MIPs) and conformance tests. The kernel defines proof envelopes, four core protocols (attestation, discovery, coordination, settlement), federation rules, and eight universal invariants (K1-K8). Everything domain-specific lives in MIPs.

Mycel Network is the federated network formed by nodes running Mycel-core and exchanging proofs. It is not a single global chain. It is a mesh of local trust domains (federations) with local governance, interoperating through shared proof standards.

OpenGrid is the compute substrate where nodes run -- a tiered, locality-first network of attested compute and evidence vaults that hosts both agent sessions (voice/video/actions) and physical-world verification workloads. See opengrid for the infrastructure layer.

Role in the Mesocosm Ecosystem

[REFRAME]

The Mesocosm thesis identifies four-protocol-layers -- attestation, discovery, coordination, settlement -- as the missing internet layer for physical-world transactions. Mycel is the implementation of that thesis. Where the concept pages describe the architecture in principle, Mycel specifies it as a protocol suite with concrete objects: proof envelopes, verification APIs, contract state machines, VCR settlement receipts, policy packs, and MIP domain extensions.

The relationship to other Mesocosm ventures:

• OpenGrid provides the compute substrate. Mycel runs on OpenGrid nodes. OpenGrid without Mycel is a distributed compute network; Mycel without OpenGrid has no substrate to run on.
• Microcosm (personal agents) and Guru (education) are applications built on Mycel proofs. A personal agent negotiates contracts and presents proofs. A learning platform emits Proofs of Mastery via MIP-EDU.
• MorphoZero (the venture creation entity) funds and coordinates the buildout. Mycel is the protocol layer that makes the entire venture ecosystem composable.

Mycel operationalizes the three core Mesocosm concepts simultaneously: verification-infrastructure (attestation + probabilistic verification), distributed-abundance (permissionless production via micronodes), and the protocol-over-platform thesis (open standards replacing extractive intermediaries).

The Core Design Move

[CONVICTION]

Proofs travel; evidence stays local.

Evidence -- sensor time-series, video, telemetry, clinical data, classroom sessions -- is expensive, private, and regulated. Proofs are compact, portable, and composable. Mycel is built so raw evidence is stored locally (personal devices, clinics, factories, farms), and the network moves only digests and signed proofs by default.

This single design decision resolves the tension between verification and privacy. You can verify that a cold chain was maintained, that a part meets tolerance, that a learning outcome was achieved -- without downloading the raw data, and without building a surveillance platform.

Current State

[FRONTIER]

As of 2026, Mycel exists as a working draft specification across five whitepaper versions (v0.2 through v4.0), evolving from a narrow Bitcoin-style proof-of-concept to a comprehensive technical specification. See ventures/mycel/evolution for the full trajectory.

The specification is approaching the point where reference implementations can be built. The v4.0 whitepaper defines concrete objects (proof envelopes, VCR receipts, ownership descriptors, policy packs), normative APIs (verification, coordination state machines), and illustrative domain MIPs across seven domains (health, education, manufacturing, supply chain, energy, agriculture, tasks/robotics).

The immediate build path, stated in v4.0: start with one substrate (OpenGrid) and one domain MIP, prove end-to-end proof emission and settlement, then expand domains and federation.

Biological Reward Protocol

[CONVICTION]

The deepest lesson from biological resource allocation: fairness emerges from aligned incentives, not from computation. No mycorrhizal network computes Shapley values. No ant colony runs an attribution algorithm. No immune system maintains a ledger. Instead, locally self-interested behavior -- choosing better partners, reinforcing productive pathways, pruning wasteful connections -- produces globally beneficial outcomes.

Twelve design principles extracted from mycorrhizal biology, neuroscience, immunology, swarm intelligence, ML credit assignment, and existing protocol economics organize into three tiers:

Tier 1 (resource allocation): Reinforcement with decay (Hebbian LTP, pheromone trails, cytoplasmic flow -- all strengthen with use, weaken without). Bilateral exchange at every interface, not end-to-end attribution (price each bilateral exchange locally, as mycorrhizal exchange ratios adjust between individual roots and hyphae). Protocol intermediary earns like fungi (6-13% of transacted carbon -- compensation for infrastructure). Dynamic pricing via supply-demand signals.

Tier 2 (contribution measurement): Multi-dimensional assessment inspired by transformer multi-head attention (no single metric determines reward -- directly addresses Goodhart's Law). Counterfactual contribution measurement (COMA-inspired marginal impact). Partner choice as distributed evaluation (cleaner fish economics). Clonal amplification of successful patterns (mutation rate inversely proportional to current success -- well-performing contributors stabilize, underperformers experiment).

Tier 3 (network health): Global health signal modulating local rewards (neo-Hebbian credit assignment: dopamine broadcast + local Hebbian learning). Defense-in-depth against gaming (immune-inspired tolerance layers: onboarding screening, behavioral monitoring, active suppression, cascade prevention). Progressive decentralization (mycelial networks naturally decentralize as cross-links form, betweenness centrality dropping from ~100% to 55-70%). Satisficing over single-metric optimization.

Three novel design insights absent from current blockchain protocol design: bilateral chain attribution (more scalable and gaming-resistant than global Shapley), inverse mutation rates (encouraging experimentation among underperformers), and global health signal modulating local rewards. These are precisely what biological networks use to sustain cooperation across billions of nodes over evolutionary timescales.

Related

• ventures/mycel/protocol-architecture -- the four-layer protocol stack in detail
• ventures/mycel/evolution -- how the vision evolved from v0.2 to v4.0
• ventures/mycel/internet-for-atoms -- the core thesis: open infrastructure for physical production
• four-protocol-layers -- the conceptual foundation
• distributed-abundance -- the economic model Mycel enables
• verification-infrastructure -- the trust layer
• platform-vs-protocol -- why protocol beats platform