You've heard about vibe coding and you're wondering what it actually looks like once the novelty wears off. The question isn't whether AI can write code: it does that every day. The real question is what someone with no technical training can actually put online, how long it takes, and what it costs.
Here are 12 projects built by non-developers: internal tools that save hours, micro-products that bring in money, personal and nonprofit apps. For each one, you'll find the original problem, the tools used, the time spent, and the real cost. You'll see that none of them requires a computer science degree — but none of them happens in thirty minutes either.
By the end, you'll know which type of project matches your current level, what these builds have in common, and where to start this week.
What vibe coding means, without the jargon
Vibe coding means describing in plain English what you want, and letting an AI assistant write the code for you, step by step, while you check the result on screen.
The term was popularized by Andrej Karpathy, former head of AI at Tesla, in early 2025. The idea: you no longer type the lines of code, you steer. You say I want a page that shows today's orders with a button to mark them as delivered, the assistant writes the files, you look at it, and you correct it out loud or in writing.
Two words to keep in mind before reading on:
- A prompt: your written instruction to the AI. The more specific it is, the better the result.
- A deployment: the step that takes your project from your computer to a real web address anyone can visit.
The most widely used tool for this kind of work is Claude Code, an assistant that reads and edits your project files directly. Its official documentation explains how it works. Nearly all the projects below were built with a tool like this, sometimes paired with a visual editor.
One honest note before we continue: the profiles are anonymized and the projects are grouped by type. The figures quoted are ballpark numbers observed on projects of this size, not market averages.
4 projects that save time at work
The most successful beginner projects are almost never consumer products: they're small internal tools that replace a spreadsheet that got out of hand.
1. An order dashboard for a florist
- Profile: shop owner, no technical experience
- Built with: Next.js and Supabase (a ready-to-use database)
- Time: one weekend, about 14 hours
- Cost: €0 for hosting, €12/year for the domain name
Orders came in by phone, Instagram, and email, then got copied into a notebook. The app shows the day's deliveries, the address, the time slot, and a delivered button. The real gain: about thirty minutes a day, and above all zero forgotten orders during the holiday rush.
2. A PDF quote generator for a self-employed plumber
- Profile: independent plumber, 46 years old
- Built with: a simple web app and a PDF generation library
- Time: three evenings
- Cost: €0
He enters the job, the quantities, and the tax rate, and the tool produces a clean quote on his letterhead, ready to send. Before, each quote took twenty minutes in a word processor. Today it's two minutes from his phone, on site.
3. A team schedule for a hair salon
- Profile: owner, five employees
- Built with: an interactive drag-and-drop board and a shared database
- Time: about 20 hours spread over two weeks
- Cost: €0
The paper schedule pinned up in the salon was causing time-off conflicts. The app shows who works when, calculates monthly hours, and sends a summary by email on Sunday evening. The project took three complete rewrites of the hour-calculation logic.
4. A supplier invoice reader
- Profile: operations assistant at an 18-person company
- Built with: a web interface wired to a document-reading model
- Time: one week, part-time
- Cost: about €8/month in AI usage
She drops in a PDF, the tool pulls out the supplier, the date, the pre-tax amount, and the tax, then exports a file ready for the accountant. Human review stays mandatory: roughly one invoice in ten needs a manual fix.
4 projects that make money
A non-developer can sell what they build, as long as they aim at a specific niche rather than a universal product.
5. A subscription-tracking micro-SaaS
- Profile: communications manager, 31 years old
- Built with: Next.js, Supabase, and Stripe for payments
- Time: six weeks, evenings
- Result: about sixty paying subscribers at €4/month after eight months
The app lists your subscriptions, warns you before each charge, and calculates what you spend per year. The longest part wasn't the code: it was setting up payments, refunds, and automated emails.
6. A browser extension for research
- Profile: independent consultant
- Built with: a Chrome extension in JavaScript
- Time: four days
- Result: €19 lifetime license, a few hundred sales to date
The extension saves an article in one click, summarizes it, and files it by topic. Publishing it on the extension store took as much effort as the development itself: privacy policy, screenshots, review process.
7. A booking site for a personal trainer
- Profile: personal trainer, 29 years old
- Built with: a slot calendar, online payments, and SMS reminders
- Time: two weeks
- Result: roughly €40/month saved on booking tools, and far fewer wasted slots
He first wanted to add messaging, nutrition tracking, and a store. He cut all of it and kept only the booking. That's what got the project shipped.
8. A wedding vendor directory
- Profile: former wedding planner, career changer
- Built with: a site generated from a data table, one page per city and per trade
- Time: three weeks
- Result: ad revenue and paid featured listings for vendors
The principle: 400 pages created automatically from a single spreadsheet. The risk with this format is producing pages with no real content, so she wrote specific copy for each city before publishing.
4 personal, nonprofit, or family projects
Projects with no commercial stakes make the best learning ground, because a mistake has no consequences.
9. A shared family shopping list
- Profile: father of three, no technical background
- Built with: a web app you can install on your phone
- Time: one Sunday
- Cost: €0
Everyone adds an item from their phone, and the list updates in real time for the whole family. A tiny project, but one that makes it concretely clear what a shared database actually is.
10. A Telegram bot for a neighborhood association
- Profile: volunteer, retired
- Built with: a script connected to the Telegram API
- Time: two evenings
- Cost: €0
The bot answers common questions (hours, address, next meeting) and relays announcements to the group. It replaced an email inbox nobody was checking.
11. A real-time fundraising tracker
- Profile: treasurer of a sports club
- Built with: a public page connected to a spreadsheet
- Time: one day
- Cost: €0
A gauge shows the amount raised and the goal, updated automatically. Simple, visible, easy to share on social media: donations picked up noticeably in the final days.
12. A review quiz for a ninth-grade class
- Profile: history and geography teacher
- Built with: a quiz app with scoring and explained answers
- Time: one week
- Cost: €0
She writes the questions in a file, the app shuffles them and explains every wrong answer. Students access it through a simple link, with nothing to install.
What these 12 projects have in common
None of these projects started from a desire to code: they all started from a specific problem the person ran into every single week.
Four things come up every time:
- A tiny scope at the start. One feature, one page, one user. The projects that fail are the ones that start with a platform that does everything.
- Real data from day one. Testing with your actual orders or your actual students immediately reveals what doesn't work.
- Lots of back and forth. Nobody gets the right result on the first prompt. Expect ten to thirty exchanges with the AI for a feature that actually holds up.
- Time spent outside the code. Payments, emails, privacy, going live: often half the project.
The pitfalls repeat themselves too: accepting code you don't understand at all, leaving an API key visible in the browser, never backing up your work. They're covered in detail in our article on the 10 mistakes to avoid when you're starting out with vibe coding.
Where to start building your own
The fastest way to get from idea to live project is to follow a guided path that imposes an order of operations, because the beginner's obstacle isn't the AI — it's knowing what to ask it, and in what order.
Method 1: a guided path on Skilzy
Skilzy is a French platform where you learn AI and use it directly. The program for building an app with Claude Code follows exactly the same logic as the projects above: frame the need, write a short spec, have the code written, test, deploy. The built-in AI Lab gives you access to the tools with credits included, which saves you from stacking up four subscriptions before you've even started.
The platform has more than 15 programs (images, video, music, automation with n8n, prompt writing, AI-assisted development) and two state-recognized, fundable certifications: RS7439 in AI content marketing and RS6792 in AI and sales. Subscriptions start at €29.90/month, no commitment.
If you just want to see what it looks like, the trial demo runs with no credit card: 7 days with 1 image, 1 video, 1 music track, and 10 messages. You do need to create an account, but no card is requested.
Method 2: the official documentation, for free
The Claude Code documentation and the hosting providers' docs are public and free. Vercel's Hobby plan, for example, lets you put a personal project online at no cost. This is perfect if you're self-directed and comfortable sorting through information on your own. The weak spot: nothing tells you what order to work in, and a lot of beginners give up at the deployment stage.
Method 3: video tutorials and communities
YouTube, Discord, and forums are full of demos. It's free and motivating, but the content ages fast: an eighteen-month-old tutorial often describes an interface that no longer exists. Use it as a supplement, to get unstuck on a specific point.
Method 4: in-person training
Some organizations offer on-site sessions. It's the most structured and most expensive format, mainly worth it when a company is training a whole team.
Whatever method you choose, pick your project first. If you're drawing a blank, our list of 30 vibe coding project ideas will give you a starting point, and the complete guide to vibe coding walks you through a full project from start to finish.
The real common thread
None of the people here became a developer. They simply stopped putting up with a problem they ran into every week. The first project is almost always small, ugly, and useful. That's exactly what you should be aiming for.
Pick the task that annoys you most in your week, write five lines describing what you'd want instead, and start there.
