10 Real Products Built by Vibe Coding

Why Does This List of the Best Vibe Coding Examples Matter?

This list matters because vibe coding results are easy to claim and hard to check. It ranks ten real examples by one criterion: best known first, then products an outside outlet confirmed, then single-source products. Willison's tools come last, included for their prompts, not customers.

The best vibe coding examples start with two famous vibe-coded apps, fly.pieter.com and MenuGen. Lumoo and GenAIPI follow because independent press covers them. Plinq, Nursa Study, PrintPigeon, Russ and Spell Strike rest on one source each.

The answer is fly.pieter.com, a browser flight simulator from Pieter Levels. In his own write-up, published in February 2025, Levels says it is his first vibe coded game and that it reached $1M ARR in just 17 days, with the build done mostly live on X. That is his claim, not an audit.

404 Media covered fly.pieter.com as an example of vibe coding and treated the revenue as the creator's own claim.

Best for: Readers who want proof a solo builder can ship something people play.

What it is: A browser flight simulator, his first vibe coded game per its builder.

Why it ranks here: Best known here, with independent press.

• Timeline: Built mostly live on X. No build time stated.
• Team effort: One builder, Pieter Levels.
• Maintenance: Not stated.

Limitations: - Revenue is self-reported. - A game says little about accounts or payments.

Choose this if: - You want a famous, checkable solo launch.

Andrej Karpathy, who coined the term vibe coding, built MenuGen, one of the most famous vibe-coded apps. It turns a photo of a restaurant menu into images of each dish. In his April 2025 write-up, he says he wrote no code directly and that 100% of the code was written by Cursor+Claude.

Karpathy is candid about the gap. He felt 80% done after the first local version, and says in his post that feeling was wrong. The hard part was wiring Vercel, Clerk, Stripe and API keys, not the code.

Joylo generates the frontend, backend, database, login and payments in the same build, and its AI Confidence Score includes an integrations audit.

Best for: Readers who want to see what the person who named the practice shipped.

What it is: A menu-to-images web app.

Why it ranks here: Second best known and most candid, but no outside outlet covers it.

• Timeline: No duration stated.
• Team effort: Karpathy alone, directing Cursor and Claude.
• Maintenance: Not stated.

Limitations: - The source is the builder's own blog. - An expert's experience may not match a first-timer's.

Choose this if: - You plan to add sign-in and payments and want the honest account of that step.

Lumoo, an AI content platform for fashion brands, reached €700k ARR in 9 months, according to Lovable's October 2025 customer story. Swedish founders Henrik Skagerlind Fasth and Peter Thörngren built it entirely on Lovable. Disclosure: Lovable's CEO is an investor in Lumoo, so this is a vendor with a stake reporting a customer's growth.

Tech.eu's December 2025 report independently confirms Lumoo exists, is built on Lovable and has signed brands including Gant, and reports an angel round of around €550,000 that included Lovable's CEO.

A product that starts on Lovable can move later: Joylo imports from Lovable, Replit, Bolt and v0 free on every plan and delivers the code to a GitHub repository you own.

Best for: Readers weighing a venture-backed route, not a side project.

What it is: An AI content platform for fashion brands, with virtual try-on.

Why it ranks here: Independent press confirms the company, but the ARR figure traces to Lovable.

• Timeline: Build time not stated.
• Team effort: Two named founders.
• Maintenance: Not stated.

Limitations: - The ARR figure comes from a vendor whose CEO invested. - It shows a startup, not a hobby build.

Choose this if: - You are weighing outside investment and want to see how a vendor-backed claim is disclosed.

According to Replit's customer story, as of September 2026, Jon Cheney, who says he does not know how to code, received a $105,000 quote to build GenAIPI and built it on Replit over a weekend instead. The story reports he spent less than $400 and landed a first customer within five days. Cheney reports these figures.

Utah Business independently covers Cheney building GenAIPI over a weekend with AI and landing a first customer soon after. It does not verify the Replit figures.

Best for: Readers weighing a quote against building a first version.

What it is: Business software with a website and admin area, built on Replit.

Why it ranks here: Independent press corroborates the build, but the costs are the founder's account.

• Timeline: A weekend, per Cheney.
• Team effort: One founder with no coding background.
• Maintenance: Not stated.

Limitations: - Cost and customer figures are founder-reported. - One quote is not a market rate.

Choose this if: - You hold a quote for an app and want to test the idea first.

Yes, according to Lovable's September 2025 customer story. Growth marketer Sabrine Matos, with no technical coding background, built Plinq, a women's safety app for Brazil that offers consumer criminal-record checks, entirely on Lovable. The story says she built the platform in 45 days.

Lovable's Plinq story also reports 10k+ users in 3 months, a vendor-reported figure for a product built on the vendor's own platform. For a fuller story, see 5 Real Apps Built by People With No Coding Experience.

A product that handles people's records is where security matters most. Joylo scores every build across five domains, security among them, and flags uncertain code before it ships. A human engineer check is a separate step you add.

Best for: Non-technical founders building something people must trust.

What it is: A consumer safety app that runs criminal-record checks.

Why it ranks here: Commercial traction, but every figure is Lovable's, so it ranks below the press-confirmed entries.

• Timeline: A vendor-reported build period.
• Team effort: One growth marketer with no coding background.
• Maintenance: Not stated.

Limitations: - Traction figures come from Lovable alone. - A trust-sensitive product carries more risk than a game.

Choose this if: - You expect to handle personal data and will want an engineering review.

Yes, according to Lovable's July 2026 customer story. Nursa's VP of Product built Nursa Study, a product for nursing students and colleges, on Lovable over a single weekend, and the story says it now has paying enterprise customers. This is a vendor-reported account, not an independent audit.

Best for: Professionals digitising part of their own company's work.

What it is: A new product line built on Lovable by a company's VP of Product.

Why it ranks here: An in-house build, but the vendor's story is the only source.

• Timeline: A single weekend, per Lovable.
• Team effort: One product leader.
• Maintenance: Not stated.

Limitations: - No user or revenue figures are given. - A weekend prototype and an enterprise-ready product differ.

Choose this if: - You own a process you could turn into software.

It looks like PrintPigeon, a service that prints and posts documents for people without a printer. Digital marketer Yannis, who is not a developer, says in a Lovable interview that the whole process took about three days and cost $38 in credits on Lovable, including learning as he went.

Yannis wired together Supabase, Resend and Stripe. The $38 is what one builder reports for one project, not a general price for vibe coding.

Joylo's AI Confidence Score includes an integrations audit on every build. Email and payment connections are the first thing to test in a paid app.

Best for: Readers who want to charge for one small, single-purpose service.

What it is: A print-and-post service built by a non-developer.

Why it ranks here: The smallest paid product here, with the cost from the builder.

• Timeline: About three days including learning, per the builder.
• Team effort: One non-developer connecting three services.
• Maintenance: Not stated.

Limitations: - The cost is builder-reported and covers credits only. - The interview reports no users or revenue.

Choose this if: - You plan to charge for one narrow service and need a cost reference from a first-time builder.

Yes, according to Zapier's June 2026 article on non-developer builds. Real estate broker Marcus Rush of Rush Home vibe coded an AI agent called Russ with Claude and Zapier MCP. Russ scores leads across a database of 11,000+ contacts and writes follow-up plans for his agents every morning.

Zapier's article also describes its growth marketing manager building an influencer ROI dashboard with Claude Code to track over a hundred influencer partnerships.

An agent that reads a contact database needs clear access rules and a home for its code. Joylo delivers the code to a GitHub repository you own, on a conventional React, Node and PostgreSQL stack.

Best for: Business owners wanting a tool for their own follow-up, not a product for sale.

What it is: An AI agent that scores leads and plans follow-ups for a brokerage.

Why it ranks here: An internal tool with one editorial source, but it shows what a non-developer can automate.

• Timeline: Not stated.
• Team effort: One broker, using Claude and Zapier MCP.
• Maintenance: Not detailed.

Limitations: - The source is an editorial piece quoting the builder. - It says nothing on access control for contact data.

Choose this if: - You have a repeatable process such as lead follow-up.

Yes. Homeschooling mother Kelly Vanderbeek, with no technical background, described what she wanted in her own words and built Spell Strike, a spelling game, on Replit. Replit's customer story, as of September 2026, headlines the result as two grade levels of progress in three months, which is Replit's claim.

Spell Strike is a personal-need build, shaped by Vanderbeek's son and his feedback, and the source attaches no revenue figure to it. Joylo builds from the same kind of plain-language description.

Best for: Parents, teachers and anyone building for a person they know.

What it is: A custom spelling game, shaped by a child's feedback.

Why it ranks here: A personal build with no commercial claim and one vendor headline.

• Timeline: Not stated.
• Team effort: One parent, describing the game in her own words.
• Maintenance: Changes driven by her son's feedback.

Limitations: - The progress claim is Replit's headline, not a study. - Little is said about how the game was tested.

Choose this if: - You are building for one person you know well.

They read like a brief, then a nudge. Simon Willison, co-creator of Django, publishes the prompts behind his browser tools on his colophon page, which says they were mostly built with AI-assisted programming and lists 237 tools as read on 30 September 2026.

One tool, photo-scrubber, started with this prompt: "Build an HTML page that loads necessary dependencies from a CDN and implements a tool which lets the user select a photo and then both strips EXIF data from that photo and also finds and blurs any faces in that photo - discuss implementation options first". A short follow-up came next: "Ditch the eyebrow and the marketing copy".

The pattern is to state the end result in plain language, ask for options before code, then refine in short follow-ups. Joylo's builder works from the same kind of prompt.

Best for: Readers who want to see real prompts before writing their own.

What it is: A public collection of small browser tools, with prompts in commit messages.

Why it ranks here: No customers, but the only source here that publishes its prompts, so it comes last.

• Timeline: Small tools with public commit histories.
• Team effort: One experienced developer.
• Maintenance: The page grows as tools are added.

Limitations: - Willison is an experienced developer. - The page changes over time.

Choose this if: - You want a prompt shape to copy for a single-purpose browser tool.

What Kinds of Things Can You Vibe Code, Going by These Ten?

Going by these ten, you can vibe code several kinds of product: personal utilities (MenuGen, Willison's tools), paid micro-SaaS (PrintPigeon, Plinq, GenAIPI), internal business tools (Russ, Nursa Study), a family learning game (Spell Strike), a viral game (fly.pieter.com) and a startup platform (Lumoo). The common thread is single-purpose tools that ship fast.

For the wider map, see What Can You Actually Build With Vibe Coding? and 11 Apps You Can Actually Build With AI.

When Do Lower-Ranked Examples Teach More Than the Top Ones?

Lower-ranked examples teach more when your situation matches theirs and the top entries do not. Spell Strike suits a parent building for one person, Willison's tools suit someone who wants prompts to copy, and PrintPigeon suits someone pricing a small paid service. Rank reflects how well known and confirmed an entry is, not how useful it is to you.

fly.pieter.com is a game and MenuGen an expert's app, so PrintPigeon or GenAIPI say more about a first paid product.

What Do Real-World Decision Scenarios Look Like?

They look like matching your situation to the closest example, not the biggest result. If you hold a quote for an app, the GenAIPI shape fits. If you are a parent building for one person, Spell Strike fits, though it ranks lower. If your app has already broken, the fit is engineer help, not another prompt.

Karpathy's account shows the demo-to-production gap: in his post, the hard part of MenuGen was connecting hosting, sign-in and payments services, not writing code. In our analysis of the repair market, The Rescue Economy, one pattern repeats. Builders describe the same failure points across independent threads: authentication, database and payment integrations breaking, security holes shipped to production, and apps that held up in a demo but failed at first real traffic.

When an engineer reviews an app before launch, we check sign-in, backups, expected load and a tested rollback path. On self-serve plans that review is fixed-price engineer help you add, not part of every build.

Start free on Joylo and build the first version. If it has to survive real users, fixed-price engineer help is one click away. From Prototype to Launch: What Still Has to Be Built covers the gap.