A document arrives.
Which review next?

Discovery means finding the material that matters in a collection of documents.

Here, a review workflow asks Braess to choose where each task goes. Follow the decision, the checks and the recorded result.

Loading the recorded run…

Four tasks. One recorded experiment. Scripted Jev decisions exercise standard review, deeper review and local fallback. A fourth task stops at its budget limit. The router and adapter ran locally; providers and documents are synthetic.

Read the branches. Dashed paths show candidate routes; the solid path shows the recorded outcome. Choose a task, then use Start and Play replay to follow its events. Each task takes one outcome path.

Inspect the result. One review passed source-span validation, one remained uncertain, one returned locally and one was deferred. Validation checks evidence structure, not legal accuracy. This page makes no model calls and contains no private documents.

Loading evidenceInspect —

Recorded tasks—

BRAESSROUTING GATE

Candidate routeRecorded outcomeCrossed endpoint: preference held by gate

Jev preferenceNot observed yet

Policy gateNot observed yet

Recorded outcomeNot observed yet

Paths illustrate decisions, not extra dispatches or fan-out. Scores and gate results appear with the recorded response; gateway timings are shown separately below.

Recorded time—Speed

— completed

— uncertain

— deferred

— pending

Total cost Not reported

Loading replay data. No provider calls are made by this page.

Across the routes

Waiting for observations

Compare the work visible at this point in the recording. Timing medians describe this sample; they do not establish a speedup.

Jev and handler intervals include transport and validation. Missing intervals remain unknown. Inspect a task below for cost receipts; total cost and savings are not established.

The record

Choose a task to inspect

A handler completion confirms a returned result. It does not establish review accuracy.

Why this route

Select a recorded task to inspect its decision.

Inside the gateway

No timing evidence selected.

Offsets use the gateway’s own clock. Send-to-validation intervals include local overhead. Missing endpoints remain unknown.