Small decisions. Interesting possibilities.Submit contentSubmit
OpenWork Jev Verification preview

OpenWork Jev Verification

OpenWork uses Jev as a typed verification judge for agent-produced work in its evaluation testkit.

Added to Jevfast

How it uses Jev

OpenWork’s testkit can use Jev as an evaluator when compiling a verification intent against a dictionary of checked-in checks. The evaluator passes the compiler’s state and questions to the configured gateway; the selected checks are then replayed through real deterministic steps without further model calls.

What Jev decides

Select a requested verification checkExample answers · not a recorded Jev response · Source ↗
Question 1 · check 0
YOUR APP
INSTRUCTION

Does the user’s intent require the composer check? Judge which checks were requested, not whether the app already passes them.

STATE

A draft-review user intent and a versioned dictionary of requested checks, including a check that the composer is editable and says “Review notes”. The app screen itself is not sent.

JEV · BOOLEAN
YesNo

App workflow

  1. Define a check dictionary

    Provide uniquely identified checks with author-written descriptions and executable observation bindings.

  2. Compile the requested intent

    Ask Jev whether the dictionary covers the intent and which checks are relevant or necessary.

  3. Review the compiled plan

    Reject incomplete or malformed coverage and inspect the selected fixed checks.

  4. Replay deterministically

    Execute selected checks in dictionary order against current application state without calling Jev again.

Why it is interesting

Jev selects from a closed set of author-defined checks and cannot create assertions or perform the verification itself; the resulting plan remains reviewable and replayable.