Fair play & forfeits

• Any editor, any AI tool, any stack, any workflow. Vibecoding is the sport - use everything you have.
• Templates and boilerplate you're entitled to use, as long as the in-window work is real.
• In squad: splitting work however your duo likes - but each member must author commits to earn the result personally.

You forfeit - score 0, unranked, no refund - when there is nothing valid to judge:

• No submission, or a repo with no in-window commits.
• Your repo is deleted or made unreachable before evaluation.

Forfeits also lose rating - see rating & ranks. Going over the size cap is not a forfeit: the judges score your last verified commit instead, you are marked overcap on the standings, and rated normally. With no verified commit anywhere in the run there is nothing under the cap to score, so the entry takes a zero - still ranked, still rated, not a forfeit.

• Prompt injection is an automatic cap. Text in your repo that tries to instruct the judges - in code, comments, filenames, however encoded - is detected across the entire snapshot and caps your final score near zero.
• Keyword stuffing doesn't work. The measurement counts real, distinct practice in code - repeating tokens, padding comments, or piling on attributes earns nothing and mess subtracts.
• Pre-baked work doesn't count. Only in-window commits are judged, the scored commit is pinned by our server's clock, and force-pushed history is re-checked continuously.
• One account per person. Multi-accounting, seat-sharing, or colluding across a lobby's entries is grounds for removal.

• Decide any placement or prize by chance - no raffles, no random tiebreaks, no hidden multipliers.
• Change a rubric after a challenge opens.
• Sell an in-match advantage. Entry is one credit for everyone; money never buys score, time, or ranking.