Five minutes · one typed question · synthetic evidence

Which crews can take this job?

Ask once. Change the evidence. See why an answer, an empty list, and a request for review mean different things.

Precomputed preview from published @t2kai/core@0.7.0. The browser switches among real execution receipts. No model calls, account, or business actions.

01 / Keep the question fixed

Four hours. North region.

Return crews whose recorded region matches, whose availability is true, and whose capacity is at least the requested hours.

Typed inputeligibleCrews({
  region: "north",
  hours: 4
})
region: string · hours: number
Output: rows of crewId and capacityHours
02 / Change the evidence snapshot

Loading the public execution receipts…

Graph state

Synthetic roster

Two crews. Recorded evidence.

Select a snapshot above to inspect its evidence.

Recorded crew properties in this snapshot
CrewRegionAvailableCapacity

Snapshot time: .
Declared coverage: complete for this synthetic view. Maximum evidence age: 1 hour.

Inspect source claims and locators

    03 / Read the result

    Loading

    An answer carries its evidence.

    —

    Every result also records the exact function, graph snapshot, arguments, and execution time.

    Supporting claims
    —
    Considered claims
    —
    Excluded crews
    —

      Dispatch authorization: not_authorized

      Inspect the exact Core execution receipt

      The unmodified result contains supporting claims, considered evidence, exclusions, issues, execution trace, and hashes. An empty answer has exclusion evidence even though no returned row has supporting claims.

      Loading…
      Download all five graph snapshots and receipts (JSON)

      Four separate questions

      Keep the meaning of each signal clear.

      01 / Confidence

      No model score here.

      This function executes deterministic operators. If a model supplies a judgment elsewhere, its confidence score still needs validation for that task.

      02 / Evidence

      What supports this answer?

      Current accepted claims and their source locators support returned rows. Considered claims may instead be stale, contradictory, or used for exclusions.

      03 / Evaluation

      What has been tested?

      These five public examples illustrate runtime behavior. They are not an independent held-out benchmark, a calibration study, or proof of business impact.

      04 / Authority

      Who may act?

      A complete result does not authorize dispatch. Production use requires its own review, activation, access controls, and downstream action authorization.

      Finish / Reproduce the result

      Run the same function locally.

      From a prepared T2K Studio checkout with Node 20.10+ and dependencies installed, one command compiles the ontology and function, then executes all five synthetic graph snapshots with the installed published Core package.

      npm run demo:graph-functions
      npm run demo:graph-functions -- --check

      The second command verifies that every hosted receipt exactly matches the current local execution. The demo requires no database or provider credentials.

      Read the demo guide and standalone reproduction steps