Caribbean AI expert Adrian Dunkley says the next Facebook is already being developed
I argue that the next major platform may begin as a small Caribbean project. Repeat use, trust and distribution will decide what happens next.
I believe the beginnings of the next major digital platform may already exist in a student project or a small founder's prototype. When I say “the next Facebook is already being developed”, I am asking Caribbean investors and educators to look earlier for that work.
This is an argument about where opportunity can begin, not a report that an identified Caribbean company will match Facebook's scale. The challenge is to help useful experiments reach users, then judge them by what those users do.
The opening could be a small community
A founder does not have to start by designing a service for everybody. A hypothetical tool for Caribbean school clubs could make it easier to arrange activities and share equipment. A service for small food sellers could help customers place repeat orders. Either would have to solve its first users' problem before there was a reason to expand.
My argument is that proximity can help a Caribbean founder ask better questions. A builder who understands how a local business takes orders or how a school organises activities can observe details that a distant product team may miss. That familiarity is a starting point for investigation, not proof of a competitive advantage.
The test is whether the product improves a task enough for somebody to choose it again. If users need repeated persuasion to return, adding features may be less useful than finding out why.
Building the first version leaves the harder work ahead
AI coding tools can assist with producing and changing code. Their output still needs inspection. GitHub's guidance on responsible use describes risks including inaccurate code and security weaknesses. A founder remains responsible for deciding what is safe to put in front of users.
A prototype can show the intended experience without having reliable access controls, recoverable data or a way to handle abuse. These gaps matter more as the product moves from a supervised demonstration to something people depend on. Mentorship should include the ability to recognise that transition.
For an early builder, a sensible first trial might involve a small group, made-up or low-sensitivity data and a clear way to report problems. Ask a competent reviewer to check the application before it handles payments or sensitive records. More users should follow stronger evidence, not substitute for it.
Distribution deserves as much attention as the demo
A product needs a way to reach people who might use it. That could begin with a school club, a trade group or an organisation willing to host a carefully supervised trial. The founder should explain why those people would try it and what would bring them back.
For products that connect people, the experience can depend on who else participates. An empty marketplace or community may offer little reason to return. Start with a group whose members have a clear reason to interact, then check whether the product helps them do it better than their current method.
Measure that behaviour in a way that fits the service. For an ordering tool, completed and repeat orders are more informative than account registrations alone. For a coordination tool, look at whether groups complete the activity they came to organise. These are proposed measures for a trial, not results from an existing study.
Trust belongs in the business model
A new platform also makes choices about whose interests it serves. If the business earns more when people spend longer looking at a screen, founders should ask whether that goal conflicts with what users came to accomplish. A service can instead define success around helping people finish a task.
My wider concern about AI and autonomy makes this an early design question. A product that makes suggestions should let people understand their options and retain meaningful control. Before adding an AI agent that acts for a user, decide what it may do, which actions require confirmation and how a person can stop or undo a mistake.
Those choices are part of the opportunity for a Caribbean founder. Competing on trust requires concrete behaviour: clear permissions, understandable prices, accessible support and a product that respects its users' time.
Look for builders who learn from evidence
Investors and mentors can ask three useful questions: what has the team made, what did users do and what changed as a result? A candid explanation of a failed test can reveal more about a founder's judgement than a presentation full of market-size claims.
Y Combinator's public application guidance stresses plain descriptions and specific founder achievements. The lesson for regional programmes is to make it possible to assess the work, even when the founder lacks polished business language or an established network.
The 2027 14West AI cohort is open for applications. Founders can submit their product, current stage, user evidence and next milestone. A demo link is useful when there is a demo; an idea-stage applicant can explain the problem and the first test instead.
The next major platform will need more than an impressive first version. The useful action now is to find people building something worth trying and give them a fair opportunity to discover whether anyone wants to keep using it.