How Judge Jev helps you finish on time
Jev reads every entry. Judges decide.
Judging runs late because judges start cold, read in arbitrary order and score against criteria nobody wrote down. Judge Jev does the preparation and leaves the judgment with people.
- 1Rubric
- PEOPLE
- Organizer writes or approves the questions.
- Drafting modelProposes questions for a criterion.
- JEV
- Checks each question: one judgment? Answerable from a write-up?
- 2Submission
- PEOPLE
- Team submits a write-up and a link.
- JEV
- Answers every rubric question for the entry in one call, with a probability on each.
- Checks whether the text tries to talk to the judges.
- CODE
- Tells the team what is missing.
- 3First pass
- JEV
- Answers sit beside each criterion.
- Low-confidence answers are marked UNSURE.
- 4Judging
- PEOPLE
- Each judge scores 1 to 5 per criterion and writes feedback, in a queue order unique to them.
- CODE
- Flags entries where judges differ by 2 or more.
- 5Results
- PEOPLE
- Organizer publishes. Each team gets a feedback card with reasons.
- CODE
- Weighted totals from the judges' scores. Jev's ranking shown beside, never instead.
- Reweighting needs no new Jev calls.
What Jev does
Answers narrow factual questions with a calibrated probability, checks the rubric's questions, and flags entries that try to talk to the judges.
What code does
Handles everything arithmetic: weights, totals, ranking, disagreement and each judge's queue order.
What people do
Write the rubric, judge each entry, decide the result and publish it.
Jev cannot generate text, count, or do date math, so it never writes feedback and never ranks entries on its own.
Open the demo event