Playground/SMT solver
API ACTIVITY

Usage & budget

Your TypeSafe API key

Import your key to override the shared server default.

Saved on this browser until removed or site data is cleared. Masked on screen; sent only with Jev requests through this app. The saved value is never displayed.

Connecting
EXACT VERIFICATION + PROBABILISTIC TRIAGE

SMT solver lab

Can every constraint be true at once? Compare Jev’s prediction with Z3’s exact check.

WORKSPACE GUIDE

SMT solver

Compare Jev’s prediction with an exact solver on structured constraints.

ConstraintsJev + Z3Verified result

What you provide

A preset puzzle or constraints written in the supported syntax. Variable types and assumptions define the exact problem.

What happens

Jev predicts an outcome while Z3 checks the constraints. The comparison keeps probabilistic classification separate from solver evidence.

What you get

A satisfying assignment, an unsatisfiable result, or an unresolved check. A definitive Z3 result takes precedence when the model disagrees.

Walk through a run

  1. 01

    Pick a puzzle

    Choose a scenario or write supported constraints.

  2. 02

    Compare the approaches

    Ask Jev for a prediction and let Z3 check the constraints.

  3. 03

    Trust the exact check

    Inspect a satisfying assignment or an unsatisfiable result. Missing or failed checks remain unresolved.

Try this

Choose Impossible bounds, inspect the contradictory constraints, and compare the model prediction with Z3's exact check.

Execution limits

Z3 performs a real solver check. A model prediction cannot override a definitive solver result.

Explore the example. Check the evidence.

Build your logic puzzle

Z3 + Jev
THE QUESTION

Can one number fit two impossible bounds?

A constraint is a rule that must hold. Add one per line, then compare a fast prediction with an exact check.

Use true / false or whole numbers. Up to 60 rules and 40 variables. ⌘ / Ctrl + Enter runs the check.

Check options Exact verification always on
Need help writing a rule?

Use =, ==, !=, <, <=, >, >=, +, -, multiplication by an integer, parentheses, !, &&, || and =>.

(a_end <= b_start) || (b_end <= a_start)

This explicitly requires two meetings not to overlap. Availability flags alone do not create scheduling rules. Unsupported syntax is rejected; constraints are never executed as code.

Jev predicts; Z3 verifies every rule. Uncertain predictions stay uncertain until the exact check completes.

Comparison

Your rules→Jev predicts→Z3 verifies

One puzzle. Two approaches.

Choose a scenario or write your own rules. You’ll see whether they can coexist, how confident Jev was, and which method took longer.

Satisfiable: a solution exists Unsatisfiable: the rules conflict Unknown: more checking needed

Seeded benchmark

Five measured cases · API calls use your configured key

No benchmark measurements yet. Latency includes network and initialization; these examples are not a general accuracy claim.

ExampleZ3JevConfidenceZ3 serverJev round tripAgree
Impossible bounds——————
Team availability——————
Equality contradiction——————
Ordered tasks——————
Overlapping meetings——————