From useful service
to verifiable contribution.

Begin with a precisely stated service.

A requester identifies the data and context it expects. A verifier checks a response against independently supplied expectations. Merely receiving a receipt with self-selected parameters is not independent authorization.

A cryptographic verification result concerns its exact statement and trust assumptions. It does not by itself establish physical storage duration, actual bandwidth, useful demand or unique economic contribution.

Public source: verification boundary

Different schedules.
The same accepted result.

For an already ordered candidate B = (o₀, …, oₙ₋₁), derive the read set Rᵢ and write set Wᵢ of every typed operation. An edge preserves order wherever a write could affect another operation.

This captures read-after-write, write-after-write and write-after-read conflicts. Operations at the same dependency depth form a wave. They read an immutable snapshot, produce isolated effects, and merge in canonical operation order.

let reference = execute_serial(&state, &candidate);
let parallel = execute_parallel(
    &state, &candidate, 4, &cancel
);

// Compare completed outcomes in controlled tests.
// A local abort is not candidate invalidity.
assert_eq!(reference, parallel);

The public finite model supports up to 1,024 existing keys, 256 typed operations and 1, 2 or 4 configured workers. State values are synthetic unsigned counters. No measured speedup or production throughput follows from these bounds.

Dependency analysis and worker coordination have costs. In a hot-key workload, parallelism may be limited or absent. CWC is an execution experiment, not a block-production or finality protocol.

Public sources: CWC model · CWC research paper

Count knowledge.
Not just identities.

An attacker can coordinate several identities, share backing data, retain earlier openings and reuse computation. A correct resource model has to account for the whole coalition and its accumulated information.

Here V is the complete allowed public view; Z has at most 2ᵐ possible values; and Γ contains at most Q replies, each from an alphabet of size at most 2ᵇ. This is a description-counting implication under bounded-channel assumptions—not a proof that a real participant has retained a given amount of memory.

When public history already reveals a target, its conditional guessing probability is one. A new nonce cannot erase that knowledge. The required link between this model and the actual service, timing and adversarial capabilities must be justified separately.

Public sources: finite information models · security scope

Keep the missing links visible.

A proposed service predicate, contribution identifier, eligibility rule, DAG order and execution engine answer different questions. They cannot be substituted for one another.

CWC explores the last question within a bounded model. The public verifier explores parts of the second. Their existence does not establish the remaining protocol.

Each stage earns the next.

1. Specify

   Write the service, adversary, public view and authority assumptions.

2. Implement

   Build bounded reference paths with unambiguous encodings and outcomes.

3. Attack

   Test replay, shared backing, omitted evidence, feedback and failure paths.

4. Reproduce

   Bind results to their source and check them with separate references.

5. Compose

   Investigate admission, ordering and execution without assuming the security being sought.

This is a research agenda, not a timed release roadmap. A future network depends on both the mathematical obligations and operational engineering being met.