Track assets, workloads, interventions, reserve capacity, causal outcomes and QCU status across regions and providers.
QCUCAUSAL VALUE NETWORKTHE CAUSAL TECHNOLOGY LAYERINTERACTIVE DEMO
Compute proved
the work ran.QCU proves what it changed.
A provider-neutral causal execution and verified-outcome settlement layer. Customers pay for conservative, independently attributable improvement, not activity, prediction or self-reported success.
SIMULATED EVIDENCEFuture-state contracts and interactions. No production execution or settlement authority is represented.
VERIFIED CAUSAL OUTCOME13.3%CONSERVATIVE ATTRIBUTABLE IMPROVEMENT
SETTLEMENT ELIGIBLE
τ̂ − zα στ
TREATMENT
COMPARATOR
CONSTITUTIONAL BOUNDARYQCU measures and governs value. It does not infer causality, certify its own actions or authorize its own settlement.
01 / CAUSAL VALUE OBSERVATORY
See the system. See what changed. Know what can be trusted.
DEMO ENGINE ONLINE
Toggle between complete telemetry and the smaller relationship set supported by identifiable mechanisms, controls and independent evidence.
Missing topology, weak excitation, conditioning, clock uncertainty and insufficient evidence remain visible limitations, not hidden assumptions.
Separate what was observed, estimated, predicted, intervened upon, independently verified and admitted for settlement.
PRIMARY STATUS LANGUAGEOBSERVEDESTIMATEDPREDICTEDINTERVENEDINDEPENDENTLY VERIFIEDSETTLEABLEINDETERMINATEREFUSED
CONNECTING ROUTING FEEDS…
GLOBAL OPERATING PICTUREINTERACTIVE 3D · SOURCE-DECLARED LAYERSREAD-ONLY QCU CONSUMER · QUANTUM ROUTING GLOBE + FEED ADAPTERS
SOVEREIGN FEED FABRICSource-bound observations
FUTURE-STATE STREAMQUANTUM VMDECENTRALIZED AUTHORITYREQUIRED
QSAINDEPENDENT WITNESSREQUIRED
QRTROUTING FEDERATIONREQUIRED
CMASWITNESS FEEDOPTIONAL
DATA SEAINGEST FEEDOPTIONAL
QHIIDENTITY AUTHORITYOPTIONAL
QORUMFINALITY STOREREQUIRED
Q-TOPOLOGYTOPOLOGY FEEDREQUIRED
QRMRESOURCE ADDRESSESREQUIRED
SOVEREIGN COMPONENTSCOMPONENT REGISTRYREQUIRED
GRAPH DBGEOMETRY INDEXREQUIRED
HEADSCALEMESH OBSERVATIONOPTIONAL
TELEMETRYRUNTIME OBSERVATIONREQUIRED
POLICYPOLICY OBSERVATIONREQUIRED
Feeds remain separate authorities. Their presence on this globe does not make them physically commensurate or grant QCU the right to infer causal truth.
ACTIVE CAUSAL CONTRACT
EcoSynQ Energy Optimization
PREREGISTERED ESTIMAND
- TREATMENT
- Causal cooling + workload intervention
- COMPARATOR
- Preregistered incumbent control policy
- OUTCOME
- Reliability-adjusted energy per useful compute
- MEASUREMENT
- Independent calibrated telemetry
ASSURANCE POSTURETIER A · INDEPENDENTLY CAUSAL
COUNTERFACTUAL DIFFERENCEkWh / workload
COMPARATOR · 100.0
TREATMENT · 81.6
ATTRIBUTABLE Δ
RAW EFFECT18.4%
COMBINED σ±3.1%
CONSERVATIVE EFFECT13.3%
SETTLEABLE QCU18,420
FIELD NOTES · THE CRWDUNIT STORY
Why this layer exists in the real world.
Short operating stories connect the doctrine to a decision a customer can recognize.
Two lines moved together. That was the beginning of the question, not the answer.
ObservationGraphV1 records interactions actually seen. CausalGraphCandidateV1 records a hypothesis that must survive topology, temporal order, controls, intervention evidence and uncertainty. CrwdUnit never silently promotes one into the other.
FROM THE FIELDAt 2:13 a.m., GPU temperature and cooling-pump power rose together. The dashboard blamed the pump. The topology showed both were responding to a new workload batch. The alarm was real; the first causal story was wrong.
Sometimes the honest answer is: this instrument cannot see the direction that matters.
Covariate Fisher information and dark-axis analysis test whether the required causal direction is observable and identifiable before an effect is estimated.
DARK_AXIS_DETECTED / EFFECT_NOT_IDENTIFIABLEFROM THE FIELDA facility had exquisite server telemetry but no calibrated branch-circuit measurement. CrwdUnit could report the workload change, but it refused to attribute the building-level energy reduction.
The mathematical optimum is not automatically an admissible action.
Sparse identification and symbolic regression can propose interpretable laws. Affine control, Hamiltonian structure and symplectic invariants bound the intervention so optimization cannot violate physical, reliability or policy constraints.
FROM THE FIELDThe cheapest recommendation was to consolidate another 12% of GPU load. Reserve coverage would have fallen below the customer’s failover mandate, so the safe envelope rejected it.
A safe action is still not a successful action.
InterventionCommitmentV1 fixes the action, comparator, outcome window and authority before results are visible. OutcomeObservationV1 and CausalEffectReceiptV1 then separate attributable improvement from concurrent workload, weather and operating changes. Independent evidence must support the result before final QCU consumption.
NO VERIFIED CAUSAL OUTCOME / NO FINAL QCU CONSUMPTIONFROM THE FIELDA cooling setpoint change coincided with lower chiller power, but outdoor temperature also fell during the outcome window. The independent witness adjusted for weather and workload, admitted only the supported portion of the saving and left the remainder visible but non-payable.
THE NON-NEGOTIABLE PROMISE
No verified causal outcome, no final QCU consumption.
QCU can settle in QCU, fiat, stablecoin, credit, invoice or shared savings. The evidence, not the rail, earns the value.