Topological Mass Consensus

Consensus from topology, not economics

A post-quantum protocol that detects Sybil attacks through discrete curvature on transaction graphs. No token. No mining. No staking. Security from geometry.

9
Rust Crates
349
Tests Passing
~460K
Lines of Code
NIST L5
Post-Quantum
Consensus has two options. Both are broken.
⛏️

Proof of Work

Security from thermodynamics. Requires burning energy to prove computational effort. Miners concentrate into pools. Hardware becomes the moat.

⚠ Vulnerable to Shor's algorithm. 51% attacks require only capital.
🏦

Proof of Stake

Security from economics. Requires locking capital to participate. Wealth concentrates over time. The rich get richer. Plutocracy by design.

⚠ Nothing-at-stake. Long-range attacks. Quantum-vulnerable signatures.
🤖

Agent Economy

Autonomous AI agents need identity and trust infrastructure. No existing protocol treats non-human participants as first-class citizens.

⚠ No substrate-independent identity. Tokens enable agent capture.
Security from geometry
κ = +0.33 dense · coherent κ = −0.33 sparse · incoherent bridge w = ε = 0.01 Honest Network Full weight · High mass Sybil Cluster Throttled · 99% reduced

Sybil attacks create geometric defects. Fake identities must connect to the honest network through bottleneck edges. These edges have low neighborhood overlap — detectable as negative discrete curvature.

Curvature throttling reduces influence by ~99%. Edges connecting Sybil clusters are automatically weighted down. Even a 5:1 attacker ratio yields only ~7% of honest mass.

κJ(u,v) = 2 · J(N(u), N(v)) − 1

edge_weight = clamp(1 + α·κ, ε, 1.0)
where α=3, ε=0.01

Curvature is immutable by construction. Computed from ancestor sets (parents + grandparents) — committed transaction data that never changes. All honest nodes agree on curvature for every edge, regardless of their view of the DAG's tips.

One geometric primitive unifies: Sybil detection, trust chain analysis, capability misuse detection, and governance capture resistance.

What makes this different
Value

No Token

The system's native value measure is topological mass — a non-transferable, non-purchasable, non-extractable structural property of coherent participation. Parasitic extraction is structurally impossible, not merely disincentivized.

Identity

AI-Native Agent Identity

Human and non-human agents (did:agi:*) participate under identical rules. Coherence detection applies uniformly regardless of substrate. Object capabilities with geometric misuse detection. Introduction chains are DAG transactions.

Security

Post-Quantum from Genesis

NIST Level 5: ML-DSA-87 signatures, ML-KEM-1024 key encapsulation, SHA3-256 hashing, Plonky3 STARKs for zero-knowledge proofs. No elliptic curves in the consensus-critical path. No migration debt.

Governance

Coherence-Weighted Governance

Vote weight equals topological mass. Manufacturing governance weight requires genuine structural integration with the majority of the network. Plutocratic capture is geometrically impossible.

Production Rust. 349 tests. Zero failures.
crypto
ML-DSA, ML-KEM, SHA3
17 tests ✓
dag
Transaction graph + curvature
43 tests ✓
consensus
Mass computation
24 tests ✓
zkp
Plonky3 STARK circuits
47 tests ✓
identity
DIDs, petnames, capabilities
79 tests ✓
p2p
libp2p + PQ transport
25 tests ✓
simhash
Structural LSH
9 tests ✓
node
PoW anti-spam, mempool, RPC
55 tests ✓
cli
Command-line interface
50 tests ✓
~19,000 lines Rust core · ~460,000 lines total · 349 tests passing · 0 failures
A third approach to consensus
DimensionProof of WorkProof of StakeTMC (Disentangle)
Sybil resistanceEnergy costEconomic stakeTopological structure
Quantum securityVulnerableVulnerableNIST PQ from genesis
Value modelTransferable tokenTransferable tokenNon-transferable mass
IdentityPseudonymous keysPseudonymous keysDID + capabilities + petnames
Agent supportNoneNonedid:agi:* first-class
GovernanceNone on-chainToken-weightedCoherence-weighted
Entry barrierMining hardwareCapital stakeParticipation (free)
EnergyEnormousModerateNegligible
Read the paper. Run the code.

Disentangle is open research from QuantumEdge LLC. The paper, implementation, and protocol specification are available now.