Beginner · Systems
Vibe Coding
Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.
TL;DR. Building software by describing what you want to an AI and accepting its output with minimal review — steering by feel rather than reading the code.
Technical Definition
Building software by describing what you want to an AI and accepting its output with minimal review — steering by feel rather than reading the code.
How it works
Coined by Andrej Karpathy in early 2025 and added to Merriam-Webster's dictionary in September 2026, vibe coding names the style of programming where you describe intent, run what the AI writes, and iterate on the error messages — without closely reading the diff. It is genuinely productive for prototypes, personal tools, and throwaway scripts, where speed matters more than correctness. It becomes dangerous when the same posture is applied to production systems: code nobody has read can carry security holes, license problems, and subtle logic errors that only surface under load. The mature 2026 position treats vibe coding as a mode with a blast radius — fine behind a feature flag or in a sandbox, paired with review, tests, and evals anywhere real users or data are involved.
Related Concepts
- Agent Harness — Everything surrounding a model that makes it useful in real workflows — the tools, permissions, memory, context, and orchestration around the model itself.
- Loop Engineering — The practice of designing repeatable systems around agents, instead of manually prompting them for one task at a time.