The MADE Rules
Six parts, in the open. Every rule change is dated in the changelog; every interpretation is published as a numbered ruling. Questions about a rule become public answers. That is itself a rule (§ 5.2).
Every AI detector can be beaten, and every semester it gets easier. Schools are left choosing between banning the most important tool of the decade and grading work no one can trust. These rules take the third option: stop trying to detect what a machine did, and start witnessing what a student can do. Nothing witnessed can be faked.
Purpose & Principles
The league exists to assess and credential authentic capability: what a student can build, verify, defend, and explain, with any tools, in front of witnesses. Its competitive format is MADE: Make, Audit, Defend, Explain.
The league does not attempt to detect AI-produced work, and no MADE rule may depend on detection. The league witnesses instead: no round is scored on unwitnessed work alone, and every score traces to something a judge, auditor, or opponent saw happen live.
Any tool, any model, any assistance is permitted in the Make round, and all of it must be disclosed in the build log (§ 2.2). Concealing assistance is the only tool-related offense, and it is scored as dishonesty under § 4.2.
Five things, and nothing else: the quality of the artifact; the accuracy of the team's audit of another team's work; performance under live questioning; the clarity of explanation; and honesty, scored under Part 4.
Defenses, red-team rounds, and awards happen live, before judges and opposing teams. Recordings exist for the record; they are never a substitute for the room.
The Season & the Rounds
Every event opens with a sealed challenge, published to all teams at the same moment and secret until the clock starts. Challenges specify a real deliverable, and where possible a real client or user whose problem it solves.
Teams build the deliverable inside the event window, using any tools (§ 1.3). Every team keeps a build log: what was made, what was generated, what was verified, and by whom. The build log is part of the submission and is what the auditing team receives.
Each team receives another team's submission and build log, and files an audit: claims checked, claims verified, claims broken. Audits are attributed: the auditing team signs its findings and answers for them at the defense (§ 4.4).
Each team defends its work live: a short presentation, then open questioning from judges and from the team that audited them. Any member may be asked about any part of the work. "I don't know. X built that with tool Y, and here is how we verified it" is a legitimate, scored answer; a confident wrong answer is not.
An opposing team attempts, live and within bounds published with the challenge, to break the work: hostile inputs, edge cases, contradictions, factual challenge. The defending team responds in real time. Attacks outside the published bounds forfeit the round for the attacker.
Every sanctioned event names a champion, and also awards the Verifier's Prize (best audit), Best Honest Failure (the disclosure that scored), and Client Impact. Open events (§ 3.4) publish standings and feed the individual record but award no titles.
A team that misses its own defense scores zero for the Defend and Explain rounds of that event. The audit it filed still counts, for it and for the team it audited. Withdrawing before the challenge drops carries no penalty and no record entry.
Chapters, Sanctioned Teams & Titles
A chapter is free, forever. It requires three to six students and the signature of one faculty or adult advisor. Chapters may be school-based or community-based: robotics teams, clubs, homeschool co-ops, and library groups all qualify.
A school may have any number of chapters. Chapters are crews, not varsity squads: no tryouts, no cuts, no bench. Nothing in these rules permits a school to limit its students to one chapter.
To enter a sanctioned event, a chapter registers a team, ratified by its school or sponsoring organization. The entry fee starts at $250 per event, set exactly in each event's program, and is the only money the league ever requires (§ 6.2). Need-based waivers are granted on application and are not published.
Open events, called Sprints, are open to any eligible student (§ 3.6). They require no chapter, no school, no advisor, and no fee. Results at open events count toward the individual defense record (§ 4.5) but award no titles.
Only sanctioned events award titles. A chapter is not a title; a chapter is permission to try to earn one. No rule, ruling, or officer may waive this clause.
The league is for high school students: grades 9 through 12, or the homeschool and gap-year equivalent, aged 13 or older. A competitor's eligibility ends with the last sanctioned event of their graduating spring. Anyone may watch; adults may judge and advise but may not compete. Where a competitor needs an accommodation to compete on equal footing, the event program provides one and says so.
Scoring & the Record
Every event publishes a box score: per-team, per-round results across Make, Audit, Defend, and Explain, plus honesty adjustments. Box scores are public and permanent. Ties break on the honesty score (§ 4.2), then verification accuracy (§ 4.4); a tie that survives both stands as a tie.
Claims made against claims sustained. A claim broken in audit or defense costs more than uncertainty admitted up front: "we did not get this working, and here is why" scores better than a confident claim that collapses under questioning.
Attacking teams are scored on signal, not noise: a confirmed break scores; an accusation refuted by the defending team costs. Volume of attack earns nothing by itself.
Audit findings are tested at the defense. Findings that hold adjust the auditing team's score upward; findings that collapse adjust it down. Signing an audit means answering for it.
Every student accrues an individual record across all events, open and sanctioned: defenses survived, career honesty score, red-team accuracy, verification accuracy. The record is portable across chapters, schools, and seasons, and it belongs to the student, not the team, not the school, not the league.
Conduct, Questions & Rulings
Adversarial rounds attack the work, never the person. A question about a claim is always in bounds; a remark about the claimant never is. Judges stop personal attack immediately and may score it as an out-of-bounds attack under § 2.5.
Any student, advisor, or judge may submit a rules question. Answers are published as numbered public rulings with the question quoted verbatim, and bind all subsequent events until superseded. There are no private interpretations of these rules.
A protest is filed within thirty minutes of the round it concerns. The head judge rules before awards, and the ruling is published with the event's record.
Rule changes are dated and public. A reversed rule stays in the changelog, struck through, with the reason it was reversed. The league does not quietly rewrite its own record.
A judge who shares a school, chapter, family, or coaching relationship with a team does not score that team. Recusals are noted in the event record. A judge who is unsure asks the head judge, in front of the room.
Governance & Independence
AI-tool vendors may not sponsor, name, or influence any event, round, or award. No challenge may require a specific commercial tool. This clause is chartered: amending it requires public notice one full season in advance.
The rules, the training materials, and the practice challenges are free. The league never requires a purchase to compete. The only fee that exists anywhere in these rules is sanctioned-event entry (§ 3.3), and it is waivable for need.
Sanctioned events run under the league's fiscal sponsor, with insurance and named adult officers of record. No sanctioned event may run without them. Open events publish their supervising adults in the event program.
Any commercial arrangement between the league and a company owned by a league officer, including the founder, must be disclosed on this site, priced at arm's length, and open to rival providers on equal terms. Rival training providers are welcome at every event, and always will be. No such arrangement exists as of this version.
Everything a competitor builds at a MADE event remains the competitor's property. The league keeps the record of the event: scores, logs, footage, and rulings. It never takes ownership of the work itself, and it may not sell or license a competitor's work to anyone.
RULES v0.2.1 · EFFECTIVE 2026-09-01 · CHANGES: CHANGELOG · INTERPRETATIONS: RULINGS