Did the machine perform verified work?
Another provider can use an admitted meter, instrument, telemetry source, or independent verifier. This establishes resource value and its physical ceiling.
A customer defines the outcome and comparison before execution. An admitted provider performs the intervention. Independent evidence determines what actually changed. Only the attributable, policy-admissible result becomes eligible for settlement.
Another provider can use an admitted meter, instrument, telemetry source, or independent verifier. This establishes resource value and its physical ceiling.
The preregistered comparator and independent observation determine the attributable effect. Verified activity can exist without becoming causally settleable customer value.
Most AI learns patterns from the past and predicts what is likely next. A pattern can show that two things happened together without proving that one caused the other. Causal AI declares the action and fair comparison first, measures both results, shows its uncertainty, and refuses to claim causation when the evidence is not strong enough. That is why it can support outcome-based payment.
The customer fixes the treatment, comparator, desired outcome, time horizon, risk ceiling, proof rules, and maximum settlement before results are known.
The selected executor commits its action, environment, control assignment, software, hardware, and execution window before observing the outcome.
An admitted measurement provider reports treatment and control outcomes, uncertainty, calibration, missing data, and regime changes.
crwdUnit estimates the causal effect, grades identifiability, applies evidence and delivery quality, and refuses settlement when the effect cannot be attributed.
QCU supplies an integrated path in which its Causal technology executes the admitted intervention, it independently measures physical delivery, Causal AI moves work through the declared topology, and anchors proof, and SEA reports reserve coverage.
A provider does not need anything but Prometheus to participate. It can use its existing executor, telemetry, verifier, witness, and settlement rail as long as each role is declared, independently identifiable, and capable of producing the canonical evidence contract.
Defines the subject, treatment, comparator, outcome, population, horizon, confounders, risk limit, and settlement rule.
Commits the executor, action, control assignment, execution window, and hardware/software identity before the outcome is known.
Proves what ran, when it ran, its environment, actual inputs and actions, and the hashes of the raw evidence.
Reports treatment and control results with units, method, calibration, uncertainty, missing data, and verifier identity.
Reports estimated effect, confidence interval, assumptions, sensitivity, identifiability, settleable effect, and settlement eligibility.
Missing, invalid, late, or policy-inadmissible proof makes the result ineligible for settlement. Stronger proof cannot increase the amount charged.
QCU value reflects verified physical work, useful computation, delivery performance, and independently supported outcomes.
Only the conservative, evidence-supported portion of an attributable outcome may become eligible for settlement.
Without admissible proof, no value is settleable.
Did a GPU/CPU cause a measurable improvement over the best declared classical alternative?
Did a scheduling intervention reduce cost without falling below the required reliability floor?
Did a blockchain improve finality, recovery, integrity, or cost over the admitted alternative?
Did a control policy reduce joules, cooling cost, or carbon intensity per completed workload?
Did an intervention reduce incident frequency, loss severity, downtime, or recovery time?
QCU is the premium assured customer and implementation. Every other provider can join without surrendering technological sovereignty. crwdUnit standardizes attribution, not the provider’s hardware.
Monitoring tools integrate power, stress, hazard, and reliability so QCU begins with observed infrastructure behavior rather than nameplate capacity.
Five enforced gates can reduce QCU issuance to zero. Measurement precedes the financial claim.
SEA settlement cannot exceed usable capacity. An empty reserve produces zero payout by construction.
More independent measurement nodes add verified output and reduce aggregate measurement variance when the reserve wire is complete.
CPU hours, GPU hours, quantum shots, storage, network activity, and verification are different machine measurements. QCU gives them one resource-value language and allows the customer to authorize one bounded budget for the outcome instead of deciphering every provider’s machine bill.
The customer defines the outcome, maximum budget, deadline, jurisdiction, reliability floor, and proof requirements. Quantum Core can optimize the provider mix inside those boundaries without changing the customer’s authority.
The maximum authorized price for the defined outcome. A quote is a ceiling, not permission to consume the entire amount.
The portion temporarily committed while compliant capacity is secured and the intervention is executed.
Only the portion supported by admissible evidence and independently attributable delivery. Unused or ineligible QCU is released.
The customer authorizes a maximum QCU budget. Quantum Core may remove execution constraints inside that mandate, but only independently attributable results become final QCU consumption. No admissible proof means no settleable causal value.
crwdunit, Inc. filed a Form C offering statement on September 27, 2022 and a Form C/A amendment on January 5, 2023 under SEC file number 020-30999.
QCU’s gated issuance chain is implemented. SEA’s coverage, guarantee-pool, settlement-capacity, and payout caps are implemented, but Ereserve is still declared rather than derived from admitted monitoring telemetry inside the audited measurement chain.
Ereserve(T) = ∫ P(t) · m(t) · η(t) dt