Thirty-two vibe coding examples, every one a single prompt and a single attempt, every one live. Four frontier AI models each built the same eight projects, from a landing page to a top-down MOBA, and nobody edited a line. Pick a project below to see all four versions at once, then open the one you like.
What are the 32 examples?
Eight projects, each built by Claude Opus 5.5, Claude Fable 5.1, GPT-6 Astra and GPT-6.1 Sol.
| Project | What the prompt asked for | See all four |
|---|---|---|
| Marketing site | A full landing page for a scheduling app, with a pricing switch | Landing page |
| Admin dashboard | KPIs, a chart, a filterable table and dark mode from fixed data | Dashboard |
| Physics toy | Twenty balls in a spinning hexagon, physics from scratch | Hexagon |
| 3D shooter | Waves, reloading, HUD and game over in three.js | Shooter |
| Phone app | A finance app with a four-step send-money flow | Finance app |
| Strategy game | A one-lane MOBA with minions, towers and Q W E R abilities | MOBA |
| Invented brand | Any product the model likes, with a memorable site | Open brief |
| Team tool | A kanban board with drag and drop, undo and filters | Kanban |
What do the examples show about vibe coding?
Three things. First, it works: every one of the 32 builds runs, and the recordings show every required interaction succeeding. Second, the prompt decides how much personality you get: on tightly specified briefs the models converged on a shared style, while the open brief produced a night train, a sky lamp, a bird whistle and an off switch. Third, the models have habits you can learn: the Claude builds are larger and richer, the two GPT builds of a brief often look like twins.
How were they built?
The same way for all 32. Each model ran in its vendor's own coding agent, Claude Code or Codex, at maximum reasoning effort, with the prompt passed word for word, a 150-minute limit and no follow-up. Each project had to be a file you can open from disk. The methodology page lists the exact commands, and each brief page has the prompt with a copy button, so any of these can be your own starting point.
Four vibe-coded finance apps, names hidden. Which would you actually use?
Take the blind test
Frequently asked questions
What are good vibe coding examples?
Complete projects built from a single description, with no hand edits. The 32 here span a marketing site, an analytics dashboard, a phone finance app, a kanban board, a physics toy, a 3D shooter, a MOBA and a site for an invented product, each built by four models.
Were these vibe coding examples edited?
No. Each is the model's first and only attempt, published exactly as it was written. The only retries allowed were for infrastructure failures, and those are listed on each build.
Can I use these prompts for my own vibe coding projects?
Yes. Every prompt is published on its brief page with a copy button. Swap in your own product, data or game idea and keep the structure.
Which AI made the best vibe coding examples?
It depends on the project. The Claude models made the richest builds, the GPT models the fastest and cheapest. Every brief has a blind test so you can judge without the labels.