Back to all trends
techFeatured

Vibe Coding

The controversial practice of building software by describing what you want to an AI

What is Vibe Coding?

Vibe coding is a software development approach where programmers describe what they want in natural language and let AI write the actual code. The term was coined by Andrej Karpathy (former Tesla AI director and OpenAI researcher) in a February 2025 tweet where he described 'just vibing' with an AI coding assistant and accepting whatever code it produced.

The practice ranges from casual prototyping to production development. At its most basic, vibe coding means telling an AI 'build me a to-do app with authentication' and iterating on the result through conversation rather than writing code manually. At its most extreme, developers ship AI-generated code they haven't fully read or understood.

Vibe coding became a genuine movement in 2026 as AI coding tools matured. Tools like Cursor, Claude Code, GitHub Copilot Workspace, and Replit Agent made it possible to build functional applications through conversation alone. Y Combinator's Winter 2026 batch included multiple startups where the entire codebase was 'vibe coded' by non-technical founders.

The backlash has been equally intense. Senior engineers argue that vibe coding produces fragile, insecure code and that developers who can't read what they're shipping are a liability. Security researchers have demonstrated vulnerabilities in vibe-coded applications. The term itself has become polarizing -- a badge of honor for some and an insult for others.

Regardless of opinion, vibe coding reflects a real shift: the barrier to building software has dropped dramatically, and the definition of 'programmer' is expanding.

Origin

Andrej Karpathy coined the term in a viral tweet on February 2, 2025: 'There's a new kind of coding I call "vibe coding" where you fully give in to the vibes, embrace exponentials, and forget that the code even exists.' The tweet resonated because it described something many developers were already doing privately but hadn't named. By late 2025, 'vibe coding' appeared in tech conference talks, job listings, and even a Y Combinator application question.

Timeline

2025-02-02
Andrej Karpathy coins 'vibe coding' in viral tweet
2025-06-15
Term enters mainstream tech discourse; conference talks proliferate
2025-11-01
Y Combinator applications begin mentioning vibe-coded MVPs
2026-01-20
Claude Code and Cursor AI Ship major agent-mode updates
2026-03-05
Viral HN thread: solo dev builds profitable SaaS via vibe coding in 2 weeks

Why Is This Trending Now?

Vibe coding surged again in March 2026 for several reasons. Claude 4 and GPT-5's improved coding capabilities made the practice more viable for production use. A viral Hacker News thread documented a solo developer who built and launched a profitable SaaS product entirely through vibe coding in two weeks. Meanwhile, a widely-shared blog post from a Stripe engineer argued that vibe coding was 'the most dangerous trend in software,' creating a fierce debate that spilled across Twitter, Reddit, and LinkedIn.

Frequently Asked Questions

What is vibe coding?
Vibe coding is building software by describing what you want to an AI in natural language rather than writing code manually. The developer guides the AI through conversation, iterating on results by explaining changes rather than editing code directly. The term was coined by Andrej Karpathy in February 2025.
Is vibe coding safe for production?
This is heavily debated. Proponents argue that modern AI coding tools produce code comparable to junior developers, which can then be reviewed. Critics point out that many vibe coders ship code they haven't read or understood, creating security and reliability risks. The consensus among experienced engineers is that vibe coding works for prototyping but production code still needs human review.
Can you get a job as a vibe coder?
Some startups and agencies are hiring for roles that are essentially vibe coding -- product builders who use AI tools to ship fast rather than writing code from scratch. However, most traditional software engineering roles still require the ability to read, debug, and modify code. The skill set is shifting from 'write code' to 'evaluate and direct AI-generated code.'

Sources

  1. Andrej Karpathy - Original Tweet
  2. Simon Willison - Not All AI-Assisted Programming is Vibe Coding
  3. Vibe Coding - Wikipedia

Tags

codingaiprogrammingkarpathydeveloper-tools
More tech trends