A Global AI Competition Closes in Four Days. Here Is What a Caribbean Team Actually Needs to Enter.

A global AI competition for young people closes its application window on September 15, 2026. Run by the UNU Global AI Network, an arm of the United Nations University, the AI for SDGs Global Youth AI Future Innovation Competition asks entrants for one thing most youth competitions do not: a working prototype, tested somewhere close to real conditions, not a concept. Applications are free, the theme this year is AI and education, and the global final happens in Macau SAR, China, on November 26.

Most Caribbean students will never hear about this specific deadline in time to enter it. That is fine. What is worth four minutes of reading, whether or not a single entry gets submitted from Kingston, Bridgetown, or Port of Spain this week, is what the competition's actual rules reveal about the gap between a school project and something a global judging panel treats as real.

Four Days, a Free Entry, and a Trip to Macau

The timeline is public and specific. Applications close September 15. Judges run an initial round from September 16 to 30. Semi-finals run October 1 to 15. The global final and award ceremony land in Macau on November 26, where finalists receive complimentary hotel accommodation and a travel allowance of up to $1,000 USD per team before tax. There is no registration fee, no training fee, and no participation fee at any stage.

Three tracks run in parallel: AI for Education, AI for Social Innovation, and AI for Less Developed Countries. Eligibility is genuinely global, open to individuals, students, youth groups, research teams, startups, and companies, with submissions accepted in English only. One filter applies specifically to registered businesses: a company entering must have been established for less than ten years. That clause does not touch a student or an informal team entering as individuals, which is how most Caribbean youth would realistically apply.

The Bar Nobody Advertises: TRL6

Buried in the eligibility rules is the detail that matters most: every project must sit at Technology Readiness Level 6 or above. TRL is an engineering scale, originally built by NASA and now standard across research funding worldwide, that runs from TRL1, a raw scientific idea, up to TRL9, a fully proven system already in routine use. TRL6 sits past the midpoint. It means a prototype has been built and demonstrated working in conditions close to how it would actually be used, with real inputs and a real user, not just a clean run on a developer's own laptop.

That single requirement disqualifies most of what young people are usually praised for at school AI fairs: a chatbot that only ever answered the questions its builder typed, a model trained and shown once, a slide deck describing what the app would eventually do. None of that clears TRL6. What clears it is a tool that has been handed to someone outside the build team, used for its actual purpose, and shown to work, warts and all.

What Caribbean Teams Already Have Sitting in a Folder

This is the part worth sitting with. The Genius Project's own AI Solution Challenge already asks teams of Caribbean teenagers to do something close to a TRL6 test: build a working AI tool for a real problem from their own community, try it on an actual person outside the team, then demo it live for three minutes. That is not the same rigour a Macau judging panel applies to a funded startup, and nobody should pretend it is. But it is the same category of evidence: a tested prototype, not a described one.

The same is true of the teams that built disaster-relief prototypes during the region's Hurricane Melissa challenge, and of the 40 teams from 32 countries who spent recent weeks building agentic AI for Future Caribbean's buildathon. None of those projects were built with this specific Macau deadline in mind. All of them cleared a bar most school AI projects never reach: someone outside the build team actually used the thing and it did what it was supposed to do.

That is worth saying plainly. A Caribbean teenager who finished a build-based programme this year is, in a narrow but real sense, closer to a global competition's actual entry bar than a classmate who took an AI course that never left the slide deck stage. The gap most Caribbean students face is not talent. It is exposure to the fact that this kind of bar exists at all before the week a deadline lands.

Four Days Is Tight. Here Is What a Real Attempt Looks Like

For a Caribbean team that already has a tested project and wants to try anyway, the honest version of a four-day plan looks like this. First, pick the project that has already been used by someone outside the team, not the most technically impressive one sitting unfinished. Second, write the submission in plain English around the AI for Education or AI for Social Innovation track, since those two fit most youth-built projects far better than the Less Developed Countries track, which is scoped to a specific UN list of countries. Third, document the actual test: who used it, what happened, what broke, and what got fixed. A judge scoring for TRL6 is looking for that evidence far more than for polish.

For everyone else, the more useful four days is spent doing something the competition itself cannot teach: starting the next tested build now, months before any deadline, so the next global opportunity that lands with four days' notice is not a scramble. The Genius Project's Solution Challenge, and the wider network of Caribbean AI programmes around it, exist specifically to close that gap before the next deadline shows up uninvited.

Back to School Means Build Season, Not Just Study Season

This deadline lands in the same week Caribbean schools are settling into the new academic year, with AI already folded into how students study, submit coursework, and prepare for CXC exams. A global competition asking for tested prototypes rather than essays is a useful reminder that the same AI skills schools are now teaching for study support translate directly into something a student can point at a real problem and build, not just something that helps with homework.

That distinction, between using AI to study faster and using AI to build something someone else can use, is the one Adrian Dunkley has built The Genius Project around. Dunkley founded StarApple AI, the first AI company established in the Caribbean, and is widely regarded across the region as its leading AI voice. The wider network built around that work includes the Caribbean AI Association, which tracks opportunities like this one and pushes for Caribbean teams to actually hear about them in time, and Jamaica AI, which covers how global AI competitions and policy shifts land specifically for Jamaican students and schools.

Whether or not a Caribbean team makes the September 15 deadline is a smaller question than it feels like this week. The bigger one is whether the region's young builders keep hearing about global AI opportunities four days before they close, or start hearing about them four months before, with a tested project already sitting ready to submit. Nobody has fully solved that yet, including this competition's own outreach into the region. That gap is the one worth building toward next.