Ask several questions of the same evidence

Send a batch of support reports once. A Choice, a yes/no question and a Score each read the same records.

  • General decision tool with code counts
  • Tool sysone_decide
  • Recorded examples here. Live runs through your account or engine.

Recorded example: Upload problems. The interactive version loads with JavaScript.

Evidence and options

  • R-1041 iOS, 4.2.0: Photo upload spins forever on hotel Wi-Fi. It works on mobile data.
  • R-1042 Android, 4.2.0: The app closes when I tap Upload on a photo larger than 12 MB. Every time: open an album, pick a large photo, tap Upload.
  • R-1043 Android, 4.2.0: Upload crashed the app twice today.
  • R-1044 Web, 4.2.0: Love the new album layout!
  • R-1045 Web, 4.2.0: How do I change the email address on my invoices?
  • R-1046 Android, 4.2.0: After updating, big video uploads fail with 'file too large'.

Recorded Jev answer, September 24, 2026

  • Choice area: uploads, >99%
  • Yes/no steps: P(yes) 98%
  • Score impact: 2.00 of 3

The reports mainly concern uploads; impact is 2.0 of 3.

Page the on-call engineer

Impact meets your 2.0 rule. Link the report IDs behind the finding and include the reproduction steps.

Measured when recorded: 208 ms in the engine, 1 model call, 835 input tokens. A recorded answer, not a live run.

Exact typed request for this example
{
  "state": {
    "source": "Support reports excerpt: 6 of the reports received this week",
    "reports": [
      {
        "id": "R-1041",
        "platform": "iOS",
        "version": "4.2.0",
        "text": "Photo upload spins forever on hotel Wi-Fi. It works on mobile data."
      },
      {
        "id": "R-1042",
        "platform": "Android",
        "version": "4.2.0",
        "text": "The app closes when I tap Upload on a photo larger than 12 MB. Every time: open an album, pick a large photo, tap Upload."
      },
      {
        "id": "R-1043",
        "platform": "Android",
        "version": "4.2.0",
        "text": "Upload crashed the app twice today."
      },
      {
        "id": "R-1044",
        "platform": "Web",
        "version": "4.2.0",
        "text": "Love the new album layout!"
      },
      {
        "id": "R-1045",
        "platform": "Web",
        "version": "4.2.0",
        "text": "How do I change the email address on my invoices?"
      },
      {
        "id": "R-1046",
        "platform": "Android",
        "version": "4.2.0",
        "text": "After updating, big video uploads fail with 'file too large'."
      }
    ]
  },
  "questions": {
    "area": {
      "type": "choice",
      "instructions": "Which product area do the problems in these support reports mainly concern? Use only the report text. Choose none when the reports describe no problem.",
      "criteria": {
        "uploads": "Uploading photos or videos",
        "billing": "Payments, plans, invoices or charges",
        "performance": "Slowness or battery use without a failure",
        "none": "The reports describe no problem"
      }
    },
    "steps": {
      "type": "boolean",
      "instructions": "Does at least one report give steps that reproduce a problem? Steps name the actions taken and the failure they cause.",
      "criteria": {
        "true": "A report lists the actions taken and the failure they cause",
        "false": "No report lists actions that reproduce a failure"
      }
    },
    "impact": {
      "type": "score",
      "instructions": "How badly do the problems in these reports stop customers from using the app? Judge only what the reports say.",
      "criteria": [
        "No report describes a problem",
        "Customers are slowed down but can finish what they started",
        "Some customers cannot finish a core task, such as an upload",
        "Customers lose money or data"
      ]
    }
  }
}

Three independent questions over one shared state: a Choice for the product area, a yes/no for reproduction steps and a four-level Score for impact.

How this experiment works

  1. Code Keep record IDs stable so every answer can be traced to its evidence.
  2. Code Count reports by platform and version with ordinary code.
  3. Code Mark earlier answers stale when the records or a question change.
  4. Jev Three independent questions over one shared state: a Choice for the product area, a yes/no for reproduction steps and a four-level Score for impact.
  5. You Choose the records to send and say they are an excerpt.
  6. You Read the answers together; no question can see another answer.
  7. You Open the records behind a finding before reporting it.

Limits

  • Each question reads the records, not the other answers.
  • Counts and totals come from code. Jev is not asked to count.
  • Six reports are an excerpt. The answers do not describe every report received.
  • Recorded examples are real Jev answers replayed only for an identical request. An edit that changes the request needs a live run.