Skip to main content

Envisioning is an emerging technology research institute and advisory.

LinkedInInstagramGitHub

Since 2010

research
  • Observatory
  • Adaptive capacity
  • Newsletter
  • Methodology
  • Origins
  • Vocab
  • RSS feeds
services
  • Signals Session
  • Bespoke Projects
  • Build Sessions
  • Pricing
  • Use cases
  • Signals
  • Signal Scan↗free
impact
  • ANBIMAFuture of Brazilian Capital Markets
  • IEEECharting the Energy Transition
  • Horizon 2045Future of Human and Planetary Security
  • WKOTechnology Scanning for Austria
solutions
  • Innovation
  • Strategy
  • Consultants
  • Foresight
  • Associations
  • Governments
  • L&D
resources
  • Partners
  • Coding for Non-Coders
  • How we work
  • Data visualization
  • Multi-Model Convergence
  • FAQ
  • Security and privacy
  • Public sector
about
  • Manifesto
  • Community
  • Events
  • Support
  • Contact
ResearchCapabilityServicesSignalsAbout
ResearchCapabilityServicesSignalsAbout
  1. Home
  2. Vocab
  3. Vibe Coding

Vibe Coding

Building software by describing intent to AI rather than writing code manually, developer as director

Year: 2025Generality: 615
Back to Vocab

Vibe coding is a paradigm shift in software development introduced into mainstream discourse by Andrej Karpathy in 2025, where developers describe what they want to build in natural language and let AI models generate, refine, and complete the code. Rather than manually typing every line, the developer becomes a director and reviewer—articulating intent, validating outputs, and steering the implementation through conversation. Tools like Claude Code, Cursor with AI autocomplete, and Codex enable this workflow by providing real-time code synthesis and suggestion in response to English descriptions.

The practice inverts traditional coding hierarchy. Instead of the human writing logic and the compiler checking syntax, the AI proposes logical structures and implementations while the human validates correctness and aligns output with intent. A developer might describe: "Create a function that takes a CSV file and returns a sorted list of unique email domains" and receive working code immediately, then ask for refinements like "make it case-insensitive" or "add error handling for malformed rows." This reduces context switching between thinking and typing, and dramatically accelerates prototyping and iteration.

Vibe coding scales across skill levels—junior developers learn by reading AI-generated code and understanding why certain choices work, while seniors spend cognitive load on architecture rather than boilerplate. The tradeoff is trust: developers must read and validate every generated block, understand its correctness, and take responsibility for bugs. It's not abdication; it's amplification. The quality of the vibe—clarity and specificity of intent—directly determines code quality. As models improve, this becomes the default development practice, shifting the craft from keystroke precision to communication precision.

Research this in Signals

Scan Vibe Coding for yourself.

Signals turns a topic into a sourced research record you can inspect and rerun. Your first scan is free, and this one starts with Vibe Coding already loaded, so edit it or scan as is.

Related

Related

Text-to-Code Model
Text-to-Code Model

AI models that translate natural language descriptions into executable programming code.

2021Generality: 620
Dialectical Autocoding
Dialectical Autocoding

An iterative code generation method using opposing model perspectives to refine output.

2023Generality: 43
Flow Engineering
Flow Engineering

A structured, iterative methodology for guiding AI models through multi-phase problem-solving workflows.

2024Generality: 339
Program Induction
Program Induction

Automatically generating programs from data and desired input-output behavior.

1988Generality: 579
Generative Workflow
Generative Workflow

An end-to-end AI pipeline that produces original content by learning from data.

2021Generality: 694
Co-Pilot
Co-Pilot

An AI system that assists humans by suggesting actions and automating routine tasks.

2021Generality: 678