What Is Vibe Coding? 7 Real Projects That Prove Anyone Can Build Software Now
Problem
I had an idea for a ghost job detector - a tool that identifies fake job postings that companies never intend to fill. But I don’t know how to build web scrapers, I’ve never written a browser extension, and I have zero experience with pattern analysis algorithms.
Six months ago, this idea would have stayed an idea. I would have needed to:
- Learn JavaScript for the scraper
- Understand browser extension APIs
- Figure out how to analyze job posting patterns
- Debug everything myself
Then I discovered vibe coding. Two weeks later, I had a working prototype.
What Is Vibe Coding?
Vibe coding is a software development approach where you describe what you want in plain English, and AI transforms your intentions into working code. Andrej Karpathy coined the term in February 2025, defining it as:
“Writing code using natural language descriptions where you just go with the vibe, iterate, and let the AI handle the implementation details.”
The key difference from traditional development:
┌─────────────────────────────────────────────────────────────────────┐│ TRADITIONAL DEVELOPMENT │├─────────────────────────────────────────────────────────────────────┤│ Idea → Learn Syntax → Write Code → Debug → Iterate ││ (months) (hours) (days) (more days) │└─────────────────────────────────────────────────────────────────────┘
┌─────────────────────────────────────────────────────────────────────┐│ VIBE CODING │├─────────────────────────────────────────────────────────────────────┤│ Idea → Describe Intent → AI Generates → Refine Prompt ││ (minutes) (seconds) (minutes) │└─────────────────────────────────────────────────────────────────────┘The Three Pillars of Vibe Coding
When I started vibe coding, I realized it’s not just about “talking to AI.” It requires three mindset shifts:
1. Natural Language Interface
Instead of learning syntax, I express my intent:
Traditional approach:- Learn React hooks for state management- Study localStorage API- Implement drag-and-drop libraries- Write sorting algorithms
Vibe coding approach:"I want a shopping list app that remembers the order I check items offand learns my walking route through the store."The AI handles all the implementation details. I focus on what I want, not how to build it.
2. Rapid Iteration Mindset
In traditional development, I plan extensively because changes are expensive. With vibe coding, I prototype first:
┌─────────────────────────────────────────────────────────┐│ ││ ┌──────────┐ ┌──────────┐ ┌──────────┐ ││ │ Prompt │───▶│ Result │───▶│ Adjust │ ││ └──────────┘ └──────────┘ └──────────┘ ││ ▲ │ ││ └─────────────────────────────────┘ ││ ││ Each cycle: 30 seconds to 2 minutes ││ Traditional: Each cycle: hours to days ││ │└─────────────────────────────────────────────────────────┘3. Problem-Centric Focus
I stopped worrying about React vs Vue, PostgreSQL vs MongoDB, or any technical decisions. The AI makes reasonable choices, and I can always say “use Vue instead” if needed.
The Tools I Use
Here’s my toolkit for vibe coding:
| Tool | Best For | My Experience |
|---|---|---|
| Cursor AI | Full-stack apps | ”Tab” completion to full agent mode |
| v0.dev | Web interfaces | One prompt, deployed in minutes |
| Claude | Complex logic, debugging | Best at explaining why code fails |
| Replit | Quick prototypes | Browser-based, instant deployment |
I start with v0.dev for UI, then move to Cursor for backend work. If I hit a bug I can’t explain, I paste the code into Claude and ask “why doesn’t this work?“
7 Real Projects Built With Vibe Coding
I found a Reddit thread where 89 people shared their vibe-coded projects. These aren’t toy examples - they’re production applications:
1. Offline AI Application (Most Impressive)
Project: A fully offline AI app combining Gemma 3, EmbeddingGemma, SD 1.5, Kokoro TTS, and Whisper.
Why it matters: This combines five different AI models into one cohesive application. The creator (Fear_ltself) reported performance “at or above proprietary SOTA of early 2023” - all running locally.
What I learned: Vibe coding isn’t just for simple apps. Complex AI pipelines are possible.
2. Security Tool for Vibe Coders
Project: Stageclear.dev - a CLI tool that acts as “first layer of defense against vulnerabilities.”
Why it matters: Someone vibe-coded a security tool for other vibe coders. It scans AI-generated code for common vulnerabilities.
What I learned: The ecosystem is already self-reinforcing. Vibe-coded tools help other vibe coders.
3. Grocery Flow - Shopping Route Learning
Project: A shopping list app that learns your walking route through the store.
Why it matters: This would require machine learning, route optimization, and mobile development traditionally. Vibe coding made it accessible.
What I learned: The app combines several complex features that I wouldn’t attempt manually.
4. Deghost - Ghost Job Detector
Project: Detects job postings that companies never intend to fill.
Why it matters: Ghost jobs waste time for millions of job seekers. This tool identifies them by analyzing patterns:
- Posted more than 30 days
- Generic descriptions
- No actual contact information
What I learned: Real-world problems can be solved quickly. The barrier between “I have an idea” and “I built it” has collapsed.
5. Transcrisper - Browser Audio Transcription
Project: Browser-based audio transcription using Web Speech API.
Why it matters: No server needed, no API costs, works offline.
What I learned: Vibe coding excels at leveraging browser APIs that developers might not know exist.
6. YaCV - YAML Resume Generator
Project: CV generator that takes YAML input and outputs beautifully designed PDFs.
Why it matters: Developers love YAML for its simplicity. This tool bridges developer preferences with professional output.
What I learned: Vibe coding can produce polished, professional tools.
7. SAKFinder - Developer Comparison Tool
Project: A “Swiss Army Knife” comparison tool for developers.
Why it matters: Comparing frameworks, libraries, or approaches usually requires building multiple prototypes. SAKFinder automates this.
What I learned: Tools for developers are a natural fit for vibe coding.
Common Misconceptions I Had
When I first heard about vibe coding, I was skeptical. Here’s what I got wrong:
Misconception 1: “No Skill Required”
Reality: Vibe coding requires clear thinking. I still need to:
- Break problems into steps
- Describe what I want precisely
- Understand whether the result is correct
- Know when to ask for changes
The AI is a powerful tool, not a magic wand. My first prompts were too vague:
BAD: "Make me a todo app" Result: Generic todo app I didn't want
GOOD: "Create a todo app where: - Each task has a priority (1-5) - Tasks auto-sort by priority - I can drag tasks between 'Today' and 'Later' - Data persists in localStorage - Dark mode by default" Result: Exactly what I neededMisconception 2: “The Code Will Be Buggy”
Reality: AI-generated code isn’t perfect, but neither is human code. The difference:
Traditional: Find bug → Google → StackOverflow → Try fixes → Debug Time: 30 minutes to several hours
Vibe Coding: "This function throws an error when input is empty. Fix it and add a test." Time: 30 secondsI still review the code. But I catch issues faster because I can ask “what does this function do?” and get instant explanations.
Misconception 3: “Only For Simple Projects”
Reality: The Reddit thread proves otherwise. Projects include:
- Multi-model AI systems (Gemma + Whisper + TTS)
- Browser extensions with complex permissions
- Full SaaS platforms with payment integration
- iOS games with App Store deployment
Misconception 4: “Replaces Developers”
Reality: Vibe coding augments developers. At companies using Cursor:
Before Cursor:- 60% coding time- 20% debugging time- 20% thinking/planning
After Cursor:- 30% coding time (AI does 50%+)- 10% debugging time (fewer bugs)- 60% thinking/planning (higher value work)Developers become architects. They think about what to build, not how to type it.
How to Start Vibe Coding Today
If you have an idea, here’s my recommended path:
Step 1: Pick Your Tool
┌─────────────────────────────────────────────────────────────┐│ What do you want to build? │├─────────────────────────────────────────────────────────────┤│ ││ Web page or simple UI ───────────▶ v0.dev ││ ││ Full web application ───────────▶ Cursor AI ││ ││ Quick prototype ───────────▶ Replit ││ ││ Complex logic/debug ───────────▶ Claude (API/chat) ││ ││ Mobile app ───────────▶ Cursor + React Native ││ │└─────────────────────────────────────────────────────────────┘All of these have free tiers.
Step 2: Write Your First Prompt
Start with a problem, not a solution:
Level 1 (Too vague):"Build me an app"
Level 2 (Better):"Build a todo app"
Level 3 (Good):"Build a todo app that syncs across devices"
Level 4 (Best):"Build a todo app where:- Tasks have title, due date, and priority- Tasks sync across devices via localStorage + manual export- I can filter by 'Today', 'This Week', 'Overdue'- The UI is clean and minimal, similar to Things app- Include a dark mode toggle"Step 3: Iterate
The first result won’t be perfect. That’s expected. Each iteration refines:
Prompt 1: "Add a shopping list that learns my store route"Result: Basic list with route memory
Prompt 2: "The route learning doesn't work. I want it to remember which aisles I visit in order."Result: Better route tracking
Prompt 3: "Add a button to manually reorder aisles if the learned order is wrong."Result: Full control over routeStep 4: Ship It
Vibe coding removes the barrier to starting. You still need to ship:
- Deploy to Vercel (one click from v0.dev)
- Export from Cursor and deploy anywhere
- Use Replit’s instant hosting
What I Built With Vibe Coding
After reading that Reddit thread, I decided to try vibe coding myself. I built the ghost job detector I mentioned at the start:
My process:
- Opened v0.dev and described the core feature
- Copied the code to Cursor for backend work
- Asked Claude to help with the scraper logic
- Iterated through 12 versions over 2 weeks
- Deployed to Vercel
What I learned:
- The hardest part was describing what I wanted, not the coding
- Each bug fix took 1-2 prompts instead of 1-2 hours
- I understood the code because the AI explained each decision
Summary
Vibe coding represents a fundamental shift in software development. Instead of learning syntax for months, you describe intent in minutes. The 89 projects in that Reddit thread prove it works for everything from iOS games to offline AI systems.
The key insights I gained:
- Natural language becomes code - Express intent, not implementation
- Iteration speed increases 100x - What took days now takes hours
- Problem focus over syntax focus - Think about what to build, not how to type it
- Learning accelerates - AI explains code as it generates it
The barrier between “I have an idea” and “I built it” has never been lower. If you’re reading this and have an idea, open Cursor or v0.dev and start describing it.
Final Words + More Resources
My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me
Here are also the most important links from this article along with some further resources that will help you in this scope:
- 👨💻 IBM Think: What is Vibe Coding
- 👨💻 Cursor AI - AI-First Code Editor
- 👨💻 v0.dev by Vercel - AI-Powered Web Development
- 👨💻 Andrej Karpathy's Vibe Coding Definition
Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!
Comments