Caribbean Teens Get Two Weeks to Build Real AI:
Inside The Genius Project's 2026 Solution Challenge

A course that ends in a quiz is a course. A course that ends with two weeks, a real problem, and a live demo is something else. That is the gap The Genius Project's 2026 cohort just walked into. The free online curriculum, six weeks for teens and five for preteens, closed out with a week on generative AI, transformers, and prompt engineering. What comes next is the AI Solution Challenge: pick something that actually annoys or hurts people you know, build an AI tool for it using whatever the course taught, test it on a real person, and defend it in a three-minute demo in front of judges.

This is not a hypothetical exercise dressed up as a hackathon. The instructions are specific about what does not count: impressive beats real is the wrong order. "Real beats impressive," the challenge page tells entrants directly, and step two of the six-step build process is "check your data," a discipline most professional AI teams skip until it costs them a project. A sixteen-year-old in Portmore or a twelve-year-old in Castries working through that same discipline, with a parent as co-builder if they are in the Pre-Teen track, is doing something closer to real applied machine learning practice than most corporate AI pilots manage in their first quarter.

Why a Free Caribbean AI Course Matters More in September 2026 Than It Did in June

Two numbers from this year explain why the timing of this challenge is not incidental. The Statistical Institute of Jamaica's Labour Force Survey put youth unemployment at 11.7% as of April 2026, with young women carrying a heavier share of that burden than young men. Zoom out regionally and the International Labour Organization's most recent analysis puts Caribbean youth unemployment at 17.6% for 2024, excluding Haiti, against just 4.7% for adults 25 and over, a gap of nearly four to one. Employers across the region are not short of applicants. They are short of applicants with the specific technical skills current job openings require, a mismatch documented repeatedly by regional labour economists and echoed in the World Bank's 2025 commentary on closing the Caribbean's skills gap.

The second number is CXC's. The Caribbean Examinations Council's May-June 2026 CSEC and CAPE results, released August 18, logged 128 exam irregularity cases region-wide, up 60% from 80 the year before, with AI misuse tracked as its own violation category for the first time. Dr Nicole Manning, CXC's director of operations, called it "the new one for us." Put those two facts side by side and a pattern appears: Caribbean youth are already using AI, in classrooms and in job searches, faster than most of them are being taught to use it well. A free course that teaches disclosure, testing, and honest demoing, not just prompting, is a direct answer to both numbers at once.

The region's policy institutions are converging on the same point from the top down. UNESCO's Caribbean Artificial Intelligence Policy Roadmap, built on consultations with more than 1,000 stakeholders across the English- and Dutch-speaking Caribbean, was formally endorsed by CARICOM's COTED-ICT ministers in 2026, with Education and Upskilling named as one of its four core pillars. The Caribbean Artificial Intelligence Forum brought regional policymakers, educators, and youth representatives together in Trinidad and Tobago in late July 2026 to work through exactly this question: how does a region with 1 in 10 students still lacking a computer at school, and 2 in 10 lacking internet access, per UNESCO's own figures, build AI literacy fast enough to matter. Ground-level programmes are where that policy either becomes real or stays a PDF.

What the AI Solution Challenge Actually Asks Students to Do

The build process has six steps, and each one strips out a way students commonly fake competence. Step one is picking a real problem: a family shop's stock levels, guessing hurricane prep needs, a household's light bill. Step two is checking what data actually exists, past bills, weather history, prices, photographs, past exam papers, and reshaping the idea if the data is not there. Step three is choosing the right AI approach and explaining why: regression or a small neural net for a number to predict, a classifier for sorting into groups, a well-engineered prompt on Gemini or Claude for something conversational or generative.

Step four is building it, in a Colab notebook for a model, or as a carefully designed AI assistant, or as a game, with the instructions explicit that small and working beats big and broken. Step five is the one judges are told to weight heavily: put the solution in front of someone who actually has the problem, watch them use it, and write down what worked and what got fixed afterward. Step six is a structured three-minute demo: the problem in 30 seconds, the solution running live in 90 seconds, what the AI is doing under the hood in 30 seconds, and what comes next in 30 seconds.

Judging mirrors that structure exactly: 25 points for whether the problem is real, 25 for whether the AI genuinely works and the team can explain it, 25 for evidence the solution was tested on an actual person, and 25 for the demo itself. The challenge's own rules push honesty over polish, telling entrants to say what their solution cannot do yet, because "judges respect honesty." For students used to exams that punish any admitted gap, that instruction alone is a different kind of lesson.

The Genius Project's Numbers, and Why They Are Worth Naming

Since Adrian Dunkley founded The Genius Project in 2024, the organisation has empowered more than 450 Caribbean youth, directed more than $14 million in cash and aid to youth development programmes, and launched 12 AI projects across 14 parishes in Jamaica alone, according to the organisation's own published impact figures. The 2026 cohort's earlier bootcamp track, reported in August, drew over 200 participants aged 5 to 18 and awarded roughly $1 million in cash and prizes across a single month. None of that is a substitute for a national curriculum. It is what a Jamaican non-profit, sponsored by StarApple AI, Maestro AI Labs, the Caribbean AI Association, 14West, AI Trinidad and Tobago, and Orbital Brand Science, has managed to build tuition-free while national curricula catch up.

Dunkley is also the founder of StarApple AI, the first AI company launched in the Caribbean, and is widely regarded across the region as one of its leading voices on how AI should actually be taught to young people, not just permitted in a classroom or policed on an exam script. That dual role, building a commercial AI company and a free youth programme out of the same conviction, is why The Genius Project's courses read less like a coding bootcamp and more like an apprenticeship in how AI gets built responsibly. Several Genius Project alumni have gone on to 14West, the entrepreneurship pipeline built to take Caribbean youth from a finished student project to an actual early-stage venture, which is the intended next step for a graduate who finishes the AI Solution Challenge and wants to keep building.

What This Means If You Are a Student, a Parent, or a Teacher

If you are a student who just finished the course, or are thinking about the next cohort, the practical path is short. Pick a problem you can describe in one sentence to someone who has never heard of it. If you cannot name the data behind it, that is the sign to change the idea, not push through anyway. Build the smallest version that runs, not the most impressive one that does not. Then find one real person with the problem, hand them your solution, and write down exactly what broke.

Parents backing a Pre-Teen entrant are named co-builders in the rules, not spectators. That is deliberate. The habit The Genius Project is teaching, disclosing exactly where AI helped and being honest about what a solution cannot do yet, is easier to absorb with an adult modelling it in the room than from a slide weeks before a deadline. Teachers watching this from outside the programme have a template worth borrowing regardless of whether their students are enrolled: CXC's own guidance already requires a Disclosure Form and an originality report for AI use on a School-Based Assessment. A classroom exercise structured like the Solution Challenge, real problem, tested build, honest demo, teaches that disclosure habit before an exam deadline forces it.

The wider Caribbean AI network Dunkley has helped build gives that habit somewhere to go after the demo. The Caribbean AI Association works the regional policy side of the same problem UNESCO's roadmap and CXC's new rules are both circling: shared standards for how AI gets taught and disclosed, rather than each territory improvising alone. A student who can already explain what their AI is doing and where it might fail is arriving at that policy conversation with something most adults in the room do not have yet: proof of work.

What Comes After Demo Day

The AI Solution Challenge is built as a finale, but it is not designed as an ending. A working predictor for a family shop's stock, a classifier for sorting ripe fruit from a phone photo, a generative study buddy tuned to a specific CSEC subject: each of those is a portfolio piece a sixteen-year-old can point to that almost no other applicant their age will have. Against an 11.7% youth unemployment rate and a job market that keeps citing a skills mismatch rather than a shortage of applicants, a demoed, tested, explainable AI project is a more concrete answer than another certificate of attendance.

Sign-up for the challenge is open now at beagenius.org/ai-solution-challenge, free, for any Genius Project 2026 graduate, teen or preteen, anywhere in the Caribbean. Two weeks is not long. It is long enough to prove a rule the whole region is still learning: the students who build AI, disclose it honestly, and test it on a real person are the ones the next exam cycle, and the next job market, will treat differently from the ones who only learned to prompt.

Supported by StarApple AI, the Caribbean's first AI company.