Run Receipts live on a preset task

Nemotron 3 Super writes a fix on Nebius Token Factory; a Token Factory Sandbox forked from before the agent started runs the tests again; the gate, plain Python, decides. Each run takes one to five minutes. Live runs can come out differently from the recording. Keep this page open until the run finishes: closing it stops the run.

Judge code

Task

The code is in the testing notes attached to the Devpost submission. Without a code, the recorded runs are at zeyusuntt.github.io/receipts.

Live runs are open.