Vibe coding an app works well for getting a prototype running in a weekend. It does not get that app through store review, in front of users, or into a revenue model on its own. This guide covers what to hand to the AI, what to check yourself, and the order I would work in for 2027.
On 2 February 2025, Andrej Karpathy posted that he had found "a new kind of coding" he called vibe coding, where you "fully give in to the vibes, embrace exponentials, and forget that the code even exists" (Andrej Karpathy on X, 2025). Within months the phrase was everywhere. Collins Dictionary later named it its Word of the Year for 2025 (Collins Dictionary, 2025).
I read that post as someone who has spent 20+ years in growth marketing across 300+ brands. My reaction was split. A cheaper first draft is a gift to any founder. A first draft that nobody can secure, maintain or market is a liability with a nice interface. I now think about every app in three steps: build it, get it found, get it paid. Vibe coding only touches the first one.
Key takeaways
- AI-written code is now normal: Y Combinator said 25% of its Winter 2025 startups had codebases that were 95% AI-generated (TechCrunch, 2025).
- Normal does not mean safe: 45% of AI-generated code samples failed security tests in Veracode's study (Veracode, 2025).
- New personal Google Play accounts must run a closed test with at least 12 testers for 14 days before production access (Google Play Console Help, 2026). No prompt skips that clock.
- Distribution decides the outcome: in ApsteQ's LunarLinc case study, 113 video ads drove 41,420 installs versus 468 from 83 image ads.
Can you really vibe code an app and ship it to the stores?
Yes, you can ship a simple app this way, and plenty of people have. The honest limit is that vibe coding gets you a working draft fast, while store approval, security, analytics and growth still need a person who reads the code and knows the rules.
Vibe coding is building software by describing what you want in plain language to an AI tool (Cursor, Replit, Lovable, Bolt or Claude Code, for example) and accepting what it writes with little or no line-by-line review. Karpathy's own post called it fine for "throwaway weekend projects." That scope matters.
Adoption is not the issue. The Stack Overflow survey found 84% of developers use or plan to use AI tools, yet 46% say they distrust the accuracy of what those tools produce (Stack Overflow Developer Survey, 2025). People use it daily and still check it. A founder with no coding background cannot do that checking, which is the gap.
Here is where I would trust vibe coding today, and where I would not:
- Good fit: clickable prototypes for investor or user tests, internal tools, single-purpose utility apps with no payments and no sensitive data.
- Risky: anything with user accounts, health or financial data, subscriptions, or a backend that stores other people's information.
- Poor fit alone: apps that need custom native features, offline sync, or real scale.
For illustration: imagine a founder builds a habit tracker in a weekend. It looks great and runs on her phone. Then Apple rejects it under Guideline 4.2 (minimum functionality) because it behaves like a thin web wrapper, and her login flow has no account deletion option, which Guideline 5.1.1 requires (Apple App Store Review Guidelines, 2026). The AI never mentioned either rule. It does what you ask, not what the store demands.
How do you vibe code an app step by step?
Treat vibe coding as the first of three stages. Write a one-page spec, build in small slices with version control, review security before you invite testers, then plan discovery and monetization before launch day. This is the order I would follow.
Build it
- Write the spec first. One job for the app, the 3 to 5 screens, what data it stores, and what happens with no network. Paste this into every session.
- Prompt in slices. Ask for one screen or one feature at a time. Huge prompts produce tangled code you cannot debug.
- Commit to Git after every working step. When the AI breaks
Want a second pair of eyes on your growth?
Book a free 30-minute strategy call. Bring your numbers, leave with two or three moves worth making. No pitch, no deck.
Book a Free Strategy Call