No single global race
Tokyo ground motion, an Osaka building and a regional propagation layer are different useful tasks. Nodes can solve them in parallel rather than destroying one another’s work in a single winner-takes-all race.
NAMAZU CHAIN / PoTGW
Namazu Chain is a research architecture in which nodes earn trust and consensus weight by producing verifiable improvements to earthquake, regional-propagation and building-response tensors—not by winning a wasteful global hash race.
The concept separates immediate earthquake communication from later scientific verification and permanent ledger finalization. It is not presented as a deployed public warning system.
A node should gain eligibility because it explains committed physical data more accurately and reproducibly—not because it searched more random nonces than everyone else.
Tokyo ground motion, an Osaka building and a regional propagation layer are different useful tasks. Nodes can solve them in parallel rather than destroying one another’s work in a single winner-takes-all race.
A validated improvement can unlock a wider-area tensor, a higher layer, a building-state update or a floor-level simulation. Progress expands a graph of useful work.
Every node carries a signed record of correct claims, challenges, rejected calculations and later physical-state changes. Trust becomes specific, inspectable and domain-aware.
The ledger does not accept a claimed accuracy number on faith. Other nodes reproduce the measurement against the same committed inputs and defined calculation rules.
Consensus credit comes from validated improvements: observed fit, held-out prediction, physical plausibility, stability and appropriate model complexity.
Maintenance tasks and low-cost validator heartbeat checkpoints keep the ledger alive. Minimal hash work, if retained at all, is only a clearly marked emergency fallback—not the economic center of the system.
Accepted tensors become inputs for later tasks. The network can progressively connect raw measurements, regions, structures and resident-level estimates.
Signed 3-axis acceleration streams from IoT devices, gateways and participating smartphones.
Canonical coefficient models for selected groups such as 8 or 80 geographically related points.
Validated local tensors become virtual observation points for wider propagation calculations.
Ground motion is combined with a versioned structural tensor to estimate motion by building and floor.
A building tensor can be correct for one physical state and later stop matching after damage, renovation, sensor failure or other structural change. Namazu preserves both the old truth and the transition.
Verified, challenged, invalid calculation, false input, superseded model, physical-state change, sensor failure suspected, or insufficient data.
Reliability is tracked by region, sensor type, tensor class, algorithm family and building class—not reduced to one simplistic global score.
Past coefficients, inputs, challenges and transition evidence remain available for later review and disaster analysis.
Public safety cannot wait for block confirmation. Namazu therefore separates a fast event path from the slower process of independent verification and permanent finalization.
Signed local detections are broadcast immediately. Nearby devices and gateways combine observations, form preliminary regional estimates and deliver progressive-confidence information to participating apps.
Nodes improve and challenge the tensors, issue work certificates, record building-state transitions and finalize a post-quantum audit trail after the immediate event.
The PoTGW-first four-node local devnet has now been implemented and tested with deterministic synthetic tasks. The next scientific step is replay against recorded accelerometer datasets and independent review.
A Java research prototype generated synthetic 3D acceleration records, fit them into coefficients, checked error, committed result roots and constructed a locally verified proof block.
The project reports 30 finalized checkpoints, 20 valid work certificates across 11 task domains, and 24 passing tests with primary hash mining disabled. All four node processes reached the same finalized tip; false and copied claims were rejected; a simulated building-state transition was preserved.
Recorded accelerometer datasets, calibrated sensors, controlled building experiments, latency studies, external scientific review and public-safety governance are still required.