Caribbean AI expert Adrian Dunkley says the next trillion-dollar founder is vibe coding at lunchtime
I argue that Caribbean students building with AI need a route from a lunchtime prototype to customers, technical judgement and patient capital.
A student in the Caribbean could be building the beginnings of a major technology company between classes. I want parents, schools and investors to take that possibility seriously before the student has a company name, a polished pitch or permission to call themselves a founder.
My suggestion that the next trillion-dollar founder is “vibe coding at lunchtime” is a challenge to the region's approach to finding talent. It is a statement of possibility, not a valuation of an identified business. The opportunity is to notice young people who can turn a problem into something another person wants to use.
Pay attention to what the student can demonstrate
Vibe coding describes building software by describing what you want to an AI tool, trying its output and asking for changes. It can help a beginner get from an idea to a working screen. A working screen, however, leaves plenty of questions unanswered: whether the system handles mistakes, protects information and does anything useful enough for people to return.
My point is that a student's early experiment deserves a route to those next questions. Waiting for a finished company before providing help can exclude the people who most need an experienced engineer, a customer introduction or a small testing budget.
Consider a hypothetical student who makes an app to organise school-club equipment. The useful evidence is whether the club can find its equipment and keep using the app. A flashy demonstration with no repeat users tells a different story. Both outcomes teach the builder something, provided somebody helps them ask why.
Parents can make room without making a gamble
The practical response at home need not be an expensive course or a decision to abandon school. Ask your child to show you what they made, describe who it helps and explain one thing that went wrong. Help them find a trustworthy adult who understands the problem and can give specific feedback.
Agree a modest budget for subscriptions and hosting, keep adult control of payments where needed, and protect time for schoolwork, sleep and activities away from a screen. A parent's encouragement should make experimentation easier without turning every hobby into a demand to earn money.
Young builders also need boundaries. Use made-up records when testing. Get permission before using another person's information, photograph or work. An app that appears helpful can still expose a classmate's details or give strangers access to an account.
Schools should reward explanation as well as output
A school can give students a regular opportunity to demonstrate an unfinished project. Ask them to identify the user, show a task from beginning to end, explain what AI produced and describe how they checked it. Students who can explain a failure should receive useful feedback rather than feel pressure to hide it.
That approach gives teachers a way to distinguish a student's understanding from the tool's output. It also makes room for students whose strength is spotting a problem, interviewing users or designing a clear interface. They can work alongside classmates developing stronger programming skills.
Technical review remains necessary. GitHub's own guidance warns that AI-generated code can be inaccurate or insecure and calls for review and testing. A mentor should help the student understand the code, manage access and recognise when a project is too consequential for unsupervised deployment.
Investors should fund the next piece of evidence
For an early project, an introduction to a potential user may be more useful than a pitch competition. Ask what the founder has learned, what people do with the product and what test would change their mind. Keep the size and terms of any support proportionate to that stage.
This has a useful parallel in Y Combinator's application guidance: explain the product plainly and give specific evidence of what the founders have built. Caribbean programmes can apply that principle without copying another accelerator's investment terms or assuming that every good builder needs venture capital.
For a minor, any commercial arrangement needs appropriate adult involvement and qualified advice on the agreement. A pathway should protect a young person's education, ownership and ability to say no.
The next step is to show the work
My argument is ultimately about access. A student should not need the right surname or an overseas introduction for someone to examine a useful piece of work. The region can create more opportunities for that examination through school showcases, technical mentors and supervised trials with local organisations.
Applications for the 2027 14West AI cohort are open. Applicants can describe an idea or show an early product, explain what they have learned and identify the support they need. Submitting an application is a request for consideration; it does not establish a company's future value.
For the student building at lunchtime, the immediate task is smaller than becoming a trillion-dollar founder: choose one person with a problem, make something they can try and learn enough to improve it.