Interactive · Play with it live
The Jev sandbox
Feed in a state, ask Noul / Choice / Score questions, and watch typed decisions come back with calibrated confidence. Edit the text and re-run to build intuition for how a System One model behaves.
Note: this is an offline teaching simulation using simple keyword heuristics — it mimics the shape of a Jev call, not the real model’s intelligence.
Pick a scenario
Classic Jev job: urgency, routing and frustration in one call.
Try editing the text — add words like “urgent”, “refund”, or “rm -rf” and watch the probabilities move.
noulDoes this need immediate escalation?
choiceWhich department should handle this?
scoreCustomer frustration (0–2)?
Auto-act threshold0.70
Answers at or above this confidence are auto-actioned; below it, they escalate to a human.
Typed decisions
Hit Run Jev to see typed, calibrated decisions come back — all evaluated in parallel.
See the API request & response
Request
{
"model": "jev-latest",
"state": "Hi, my Stripe payouts have been failing for 3 days and I'm losing sales. This is really urgent, please help ASAP!",
"questions": {
"is_urgent": {
"type": "noul",
"question": "Does this need immediate escalation?"
},
"department": {
"type": "choice",
"question": "Which department should handle this?",
"choices": [
"billing",
"technical",
"account"
]
},
"frustration": {
"type": "score",
"question": "Customer frustration (0–2)?",
"scale": "0-2"
}
}
}