From Idea to Launched App in 48 Hours — This Is How Vibe Coding Works

We built a working SaaS product over a weekend using nothing but natural language. Here's an honest account of what happened.

It was a Friday evening in Oslo. We had a client brief, a vague idea, and exactly zero lines of code. By Sunday night, we had a live product with real users. I'm still not entirely sure how.

That's not hyperbole — that's vibe coding. And it's worth understanding what that actually means before you dismiss it or over-romanticize it.

What "Vibe Coding" Actually Means

The term was coined by Andrej Karpathy in early 2025 to describe a style of software development where you don't write code so much as you direct it. You describe what you want — in plain language — and an AI system produces the implementation. You review, iterate, redirect. The code is almost beside the point.

This sounds simple. It isn't. The skill isn't in typing; it's in thinking. You need to know enough about how software works to describe it precisely, catch mistakes, and push back when the AI goes sideways. We've seen people with no coding background struggle because they couldn't articulate the problem. And we've seen experienced developers fly because they could describe system architecture in one paragraph and have it scaffolded in seconds.

The Weekend Build

Friday at 9pm: the client needed a B2B lead qualification tool. Form on one end, scored leads on the other, with email notifications and a simple admin dashboard. A month-long project by traditional estimates.

We opened Claude, described the data model in a single prompt, and had a working schema within minutes. Not a rough sketch — an actual Postgres schema with indexes, foreign keys, and sensible naming conventions. We described the scoring logic in plain English: "weight company size 40%, budget fit 35%, timeline urgency 25%." The function appeared, correct, tested, edge cases handled.

This surprised us. Not the output — we'd seen good AI code before — but the pace. We weren't writing code at all. We were making decisions. Architecture decisions, UX decisions, business logic decisions. The implementation was largely automatic.

Where It Gets Hard

Saturday afternoon, we hit the first real wall. The dashboard needed a real-time update when new leads came in. We described what we wanted three different ways. The AI kept giving us polling solutions when we needed WebSockets. The gap between what we said and what we meant was the problem — not the AI.

This is the part no one tells you about vibe coding: precision matters enormously. Vague descriptions produce vague results. "Make it update in real time" is not a specification. "Use a WebSocket connection to push new lead objects to the dashboard immediately upon insertion, without a page refresh" — that works.

A GitHub study on AI-assisted development found that developers using AI coding tools completed tasks 55% faster on average. Our experience tracked with this — but only after we learned to communicate clearly. The productivity gains weren't automatic. They came from adapting how we thought about problems.

Sunday: Launch

By Sunday morning we had a product. Edge cases cleaned up, styling done, deployed to a Cloudflare Worker with a custom domain. The client was testing it by noon.

48 hours. A lead qualification SaaS, live, with real users, built by two people. The client's previous vendor had estimated six weeks.

We're not saying six weeks of work happened in a weekend. We're saying the gap between idea and shipped product has collapsed in a way that changes what's economically viable to build. Projects that would have required a funded team can now be explored by two people over a long weekend. That's not a marginal improvement. That's a different kind of possibility.

What This Doesn't Replace

Vibe coding doesn't make technical judgment irrelevant — it makes it more important. You're no longer the one writing every line, which means you need to understand what good looks like when you see it. Security, performance, scalability: the AI will miss things. Your job is to catch them.

What's changed is the leverage. One person who knows what they're doing now moves faster than a team used to. And that's just the beginning.