Vibe Coding SG Tutors: Web App + iOS App in Under 5 Hours
Summary
We built SG Tutors — a full-stack tutor marketplace and a native iOS app — from an empty folder to a live site and an App Store submission in under five hours with Claude Fable. Here's what shipped, the workflow, and how to learn it.
We gave ourselves an afternoon and one rule: build a real product, not a demo. Starting from an empty folder, we used Claude's Fable model to design, build and deploy SG Tutors — a full-stack tutor marketplace — to a live URL, and to build a native iOS app for it and submit that app to the App Store for review. Total elapsed time: under five hours. This is a case study of what AI-assisted vibe coding actually ships when you treat the model as a build partner, not a chatbot. If you want your team building at this speed, book a vibe-coding training consultation.
What we shipped, and the clock
SG Tutors is a working Singapore tutor marketplace: parents and students search verified tutors by subject and level, read reviews, and enquire with a tutor directly. It is not a landing page with a waitlist — it is a deployed full-stack application with a database behind it, live on the public internet. Alongside it we built a native iOS client and pushed it into App Store review the same afternoon. Here is roughly how the five hours broke down.
| Phase | What happened | ≈ Time |
|---|---|---|
| Scope & data model | Define the marketplace, tutors/subjects/reviews schema, pages | 30 min |
| Full-stack web build | Frontend, API routes, database, search and enquiry flow | ~2 hrs |
| Deploy live | Containerise, ship to a live URL, smoke-test | 30 min |
| Native iOS app | SwiftUI client against the same API | ~1.5 hrs |
| App Store submission | Archive, upload to App Store Connect, submit for review | 30 min |
The headline is not "AI wrote code fast". The headline is that a single builder, directing a capable model, closed the whole loop from empty folder to a live web product and a submitted iOS app in an afternoon — a scope that traditionally books a small team for a fortnight.
The stack and the workflow
The full-stack website
The web app is a modern full-stack build: a server-rendered frontend, typed API routes, and a Postgres database holding tutors, subjects, levels and reviews. Search filters by subject and level; each tutor has a profile and a direct enquiry path. Everything a real marketplace needs — data model, server logic, UI — was built and wired in one sitting. This is the same shape as any marketplace app development engagement; vibe coding just compresses the timeline.
The native iOS app
The iOS client is a genuine native app in SwiftUI, talking to the same API as the website rather than wrapping it in a web view. Building it immediately after the web app is where AI assistance compounds: the data model and endpoints already existed, so the model could scaffold the screens, networking and models against a known contract. The result is the kind of native mobile app that earns a place in the App Store, produced in an hour and a half.
The App Store submission
Shipping code is not the finish line — a real product has to reach users. We archived the build, uploaded it to App Store Connect and submitted it for review the same afternoon. The submission mechanics (signing, metadata, review notes) are exactly the kind of repetitive, error-prone work an AI copilot removes friction from, so the human can focus on the parts that need judgement.
How five hours is even possible
Vibe coding is not "type a prompt, paste the code". It is a disciplined workflow where you hold the intent and the model holds the typing — you describe the outcome, review what comes back, and steer. The speed comes from removing the slow parts of software work that were never the interesting parts. The contrast with the traditional path is stark.
| Step | Traditional | AI-assisted vibe coding |
|---|---|---|
| Boilerplate & setup | Hours of scaffolding | Generated in minutes |
| Data model → API → UI | Hand-written, layer by layer | Scaffolded against one contract |
| Second client (iOS) | A fresh project from zero | Built against the existing API |
| Your job | Typing every line | Directing, reviewing, deciding |
The skill that makes this work is not prompt trivia — it is knowing the product you want, reviewing critically, and steering the model when it drifts. It pairs naturally with a structured, self-correcting approach to agents; if you want the deeper version of that discipline, read our companion piece on loop engineering. The same judgement underpins our AI solutions and app builds — and if you would rather have us ship it, ask us to build your product.
This is a learnable, fundable skill
The most common reaction to a five-hour build is "that only works if you're already an expert". Partly true — judgement is what makes it fast — but the workflow itself is teachable, and it is the fastest way to make a competent developer dramatically more productive. That is exactly what WSQ - Build Professional Web Apps Quickly with AI-Assisted Vibe Coding teaches in one day: setting up the AI development environment, prompting practices that actually work, and building frontend, backend and APIs with AI tools. It is SkillsFuture-funded — Singapore citizens and PRs pay from S$156–S$236 nett after subsidy — so upskilling a whole team is affordable.
For organisations that want a tailored programme rather than a public class, we design custom curricula and assessments through our WSQ course development practice, and you can browse the wider catalogue of AI courses at Tertiary Courses Singapore to build a path for your team. If you are curious how far this goes, our take on how AI is transforming software development and why domain knowledge is the real moat are good next reads.
FAQ
Is a five-hour build production-ready or just a prototype?
SG Tutors is deployed and functional — a real database, real search, real enquiry flow, live on the internet. As with any product, hardening, edge cases and scale work continue after launch. The point of the exercise is that AI-assisted development collapses the distance from idea to a live, testable product from weeks to hours, so you validate with something real instead of a mockup.
Do I need to be a senior developer to vibe code like this?
You need enough engineering literacy to review what the model produces and steer it — the workflow amplifies judgement, it does not replace it. A competent developer sees the biggest gains, but the practices are learnable, which is precisely what the vibe-coding course is designed to build.
Which AI model did you use?
We used Claude's Fable model for this build. The workflow, not the specific model, is what transfers — the discipline of describing intent, reviewing output and steering applies across capable coding models, and it is what we teach.
Can you build our product this way?
Yes. The same approach that produced SG Tutors — web plus native mobile on a shared API, shipped fast — is how we run bespoke builds. Tell us the product and we will scope it, or train your team to do it in-house.
What to do next
- See the result for yourself — open the live SG Tutors marketplace and our related read on vibe coding as a moat.
- Learn the workflow at WSQ AI-Assisted Vibe Coding, from S$156 nett with SkillsFuture funding.
- Want us to build or co-build your product? request a build quote and we will scope it with you.