Concept 14

Vibe coding

Vibe coding is building software by describing what you want to an AI and letting it write the code, often without reading the code yourself. It is the everyday name for how people who do not program now make apps, sites and small tools, in builders such as Lovable, Replit and Google AI Studio and in ChatGPT and Claude themselves. Developers who review every change sometimes call their version agentic engineering; the difference is how much gets checked, not whether AI writes it.

Standing: Established since 2025-11
Stars, week: +980 across 4 repositories
Stars, month: +7,618, through 2026-11-08
Listings: 4 chosen by hand

The standing

Where it stands

Among the 14 concepts. Bars: GitHub stars gained this week.

Established · 7

Rising · 7

How to use it today

In ChatGPT or Claude, as a person doing the work, not as a developer.

1Describe the thing in plain words: who uses it, what they do on each screen, what it must remember. Ask for a first version, try it, then ask for one change at a time.
2Use a builder that hosts the app for you (Lovable, Replit, Google AI Studio), or ask Claude or ChatGPT to build it in the chat. You do not need to read the code to get a working first version.
3Before anyone else uses it, ask the assistant to explain in plain words what it built, where the data is stored, who can see it, and what it does not handle. Keep that list next to the app.
4Run the builder's security check before you publish, then ask a fresh chat to look for exposed data, open sign up and leaked keys.
5Keep payments, passwords and other people's personal data out until someone who can read code has checked it, or let an established service (a payment provider, a sign in service) handle that part.
6When people start to rely on it, write down what it must do and how you check it (see Spec-driven work), and keep a copy of every working version so you can go back.

What changes day to day

An idea becomes something you can click through in an afternoon, without waiting for engineering. Your work moves to describing clearly, trying it as a user would, and deciding what has to be checked before anyone depends on it.

Where it goes wrong

Code nobody read has bugs and security holes nobody looked for.

Tests of over 100 models in 2025 found security flaws in 45% of generated code, and a scan of one popular builder found 170 of 1,645 apps exposing personal data. The prototype that works quietly becomes the product, and nobody owns it.

Listings that use it well

Chosen by hand after reading them, each with the reason. Install from the listing's own page.

Listing	Kind	Why it is here	Stars
CloudBase-AI-Toolkit	Agent skills	cloudbase-sites vibe codes a full stack web app with database and hosting, Lovable style, inside a chat harness.	1.2k
vibe-security-skill	Agent skills	Audits vibe coded apps for the vulnerabilities AI assistants introduce.	1.2k
thinkrail	Agent skills	JetBrains' lightweight IDE for vibe coding with pi, with workflow and spec skills for people who do not read the code.	8k
VibesOS	Plugin	A GUI for Claude Code that makes it easy to vibe code simple, un-hackable apps.	135

Elsewhere

Repositories that embody the idea and are not in the catalog.

stackblitz-labs/bolt.diy ↗Open-source Bolt: prompt, run, edit and deploy full-stack apps with any LLM.
dyad-sh/dyad ↗Local, open-source app builder in the Lovable and v0 style.
cline/cline ↗Autonomous coding agent for VS Code; the editor-side route to the same build-by-prompt flow.

Standing

Our judgment, with the reason and the sources it rests on.

Established since 2025-11. Coined by Andrej Karpathy on 2025-02-02 for casual building where you "forget that the code even exists", it became the general word for building software with AI: Collins named it Word of the Year in November 2025, and Google, Lovable, Replit and Microsoft use it for their own products. On its first anniversary Karpathy proposed agentic engineering as a separate name for professional work with careful review; the casual practice and its name carried on. Settled practice: the methods are known, the tools are mature, and the question is how well you use it.