GenLayer studionet · intelligent contract · live snapshot

Work goes in.
A number comes out that anyone can check.

Every score below was settled by a validator set reading the work under consensus. The part a model is good at — whether a piece reads well — is the only part it was asked for, and it answers with one rung of a published ladder that the validators must agree on exactly. Every deduction, every disqualification, the ranking and the payout were computed by code, inside the contract, from that rung. Two readers who see the same work the same way cannot write two different outcomes, however differently they word it.

How a score is settled

A model compiles the sponsor's rubric into Python expressions. They run in a sandbox against the submitted work. Only then is the work read, with the sandbox result handed over as ground truth the reader is told it may not contradict.

Sandbox
A mandatory criterion is unsatisfied
score 0 · disqualified
Sandbox
A preferred criterion is unsatisfied
-15 each
Sandbox
Token overlap with an earlier entry reaches 80%
score 0 · disqualified
Reader model
Which rung of the ladder below the work sits on, and why
one rung, agreed exactly

The reader is asked for a rung, not a number, and a grade is only written when the validator set agrees on that rung exactly — along with the gate, the disqualification and the count of missed criteria. The points, the ranking, the winner and the payout are arithmetic over those, so there is no reading the arena accepts that could hand the prize to a different entry.

The board

The record

Written by consensus, readable synchronously by any other GenLayer contract, and impossible to write by hand: a higher-stake arena or a hiring contract can gate on it.

SolverEntriesGradedWins DisqualifiedBestAverageGEN earned