The plain-English version
Vibe coding is describing what you want to build in ordinary language and letting an AI coding agent — Claude Code, Cursor, Codex, or similar — write, run, and fix the actual code. You're steering by outcome ("a booking page that reads my Google Calendar and lets someone pick a free slot") rather than by syntax. The term caught on because it captures something real: for a huge range of small, well-scoped tools, you no longer need to know how to code to get one built. You need to know what you want clearly enough to describe it, and enough judgment to tell when the result is actually right.
What it's good at
Vibe coding is strongest on bounded, well-understood problems: a landing page, a CRUD app, a booking flow, a Stripe checkout, a Chrome extension. These are things thousands of developers have built before, so an agent has a deep, well-trodden pattern to draw from. That's exactly the category this site is built to evaluate — every app in the register got scored on precisely this: how bounded and well-understood is the actual build, component by component, not just "is this technically possible."
What it's bad at
It's weaker exactly where the underlying engineering is genuinely novel or where the value was never really the code in the first place. A multiplayer canvas renderer, a licensed music catalog, a payroll-tax filing system — these aren't hard because nobody's tried to prompt their way there, they're hard because the actual bottleneck is a rendering engine, a legal agreement, or a compliance surface that no amount of good prompting shortcuts. Knowing the difference between "an agent hasn't built this exact thing" and "this can't be vibe-coded" is most of the judgment call, and it's why some pages on this site end in a stamp that says JUST PAY instead of BUILD IT.
A reasonable way to start
Pick something you're already paying for that's mostly a form, a list, and a database — a link-in-bio page, a URL shortener, a personal task tracker. Those are the apps on this register that score highest, and for good reason: the pattern is genuinely common, so an agent gets it right the first time far more often. Build one of those before you attempt something with real-time collaboration or a payment marketplace baked in. Confidence compounds; a working weekend project teaches you more about what your agent is actually capable of than reading about it does.