Every hackathon has a rumour. Who can answer it?
Somebody always suspects the result was decided before the pitches were heard. Usually there is no way to answer them — the organisers can only ask to be trusted. And when software helps read the entries, a second question appears: who answers for what it did?
"Trust us" is the only answer
When a losing team asks whether the goalposts moved, the organiser has nothing to show but its word. There is no record anyone can re-run.
Nobody reads the two-hundredth pitch the same way
A human judge cannot honestly promise equal attention to every entry. Software can read them all the same way — but software cannot be asked to answer for a decision.
Panels and scores can change quietly
Who picked the judges, whether a score was altered afterwards, whether an entry was polished after the deadline — today none of this can be checked.
Three rounds. The AI narrows; people decide.
The hackathon runs on the NEO stack's Chamber — the same three-tier escalation the stack uses to settle disputes, run on submissions instead. The AI only ever advises. Every decision is a signed record that anyone can verify on their own.
A record of who judged what, by which criteria, and when — that the organiser cannot rewrite and that any team can check.
Fair is not a promise. It is a procedure anyone can re-run.
For participantsLose, and still know it was fair
The rubric existed before you entered. Your submission is what you submitted. You were on time, provably. Nobody picked your panel. Nobody rescored you afterwards. And you can read why. None of it requires trusting the judges — it is arithmetic you re-run yourself.
For the organiserA record instead of a reputation
When the result is questioned, the organiser shows the record rather than asking for trust. The AI carries the reading load; the panel carries the signature; every cut can be defended with its own receipt.
For the regulatorAI that advises, people who answer
Every AI output is labelled and fingerprinted; every human decision references it; a tampered advisory is rejected. This is the AI Act Article 22 path, built into the Chamber rather than layered on for the event. AI-triaged is not finally adjudicated; Verified is not the same as good.
For the NEO stackThe dispute machinery, exercised in the open
A hackathon is not a bespoke process invented for the occasion. It exercises the exact three-tier escalation the Chamber uses when real money is in dispute — the same machinery, run on submissions instead of disputes.
Enter. Be read. Be judged by a panel nobody picked. Then check it all.
The criteria are committed before entries open. An AI reads every entry the same way and writes its reasons down. Mentors and judges work with the teams that get through. A drawn panel picks first, second and third. Afterwards, every team — winners and losers — can re-run the arithmetic.
Round one — AI triage
Your entry is sealed at the deadline and read by the AI alongside all the others, with reasons recorded. The shortlist is a recommendation; people sign it before it counts. An entry that trips an integrity flag goes straight to the panel.
Round two — mentors and judges
Assigned mentors and judges work with the teams that got through, alongside the AI, not replaced by it. Who advised whom, and when, is recorded rather than remembered.
Round three — the panel decides, and you verify
A drawn panel picks first, second and third. The AI advises; the signature is human. Then you take the published inputs, re-run the draw, and read the reasoning bound to every decision about you.
The three rounds, the AI binding, the drawn panel and the signed receipts run end to end as one command against the live Chamber tier engine; that run used synthetic entries — no real venture is shown — and the next edition's dates, team size, criteria and prizes are still being set.
AI Hackathon — COTRUGLI Tech, part of COTRUGLI Business School, the leading business school of Southeast Europe. Built on the NEO stack's Chamber. Lose, and still know it was fair.