Vibe Coding: Intent-First Development and Its Limits
Vibe coding — Karpathy's February 2025 term for steering AI agents almost entirely by intent, accepting changes without reading the code — is a genuine workflow and a genuine risk. The useful move is treating it as a dial: prototype-grade freedom on one end, production-grade discipline (specs, tests, review, provenance) on the other.
01.The Problem: Code Nobody Read Is Now Running
It is Friday afternoon. Someone on your team typed two sentences into an AI tool.
By Friday evening there is a working app: database, login page, admin panel, live URL.
Nobody read the code. Not really. The prompts felt right, the demo looked right, the button worked.
So the question becomes
Is "it does what I asked" enough of a relationship with software to call it development?
In February 2025, researcher Andrej Karpathy (former OpenAI, ex-Tesla AI) gave the practice its name in a now-famous post:
"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."
In plain words:
Vibe coding = you say what you want, the AI writes the code, and you judge only the result.
He meant it as a description of a new, legitimate activity — using LLMs (with tools like Cursor, Topic 269) where you accept nearly all changes without reading them, steering by outcome, not by code.
The practice exploded. And so did a second story: exposed admin panels, plaintext secrets, paywall bypasses. Both stories are true. This topic is about holding both at once.
The Vibe-Coding Spectrum and Its Safety Gates
The Vibe-Coding Spectrum and Its Safety Gates
Vibe coding is a dial, not a switch: as human code-contact drops, automated tests, security review, observability, and a maintained mental model become the load-bearing safety gates.
Unlock Topic #277: Vibe Coding: Intent-First Development and Its Limits
You are viewing a preview. The full in-depth technical walkthrough, worked derivations, and code notebooks for this concept, along with self-assessment quizzes, are available with Pro or Lifetime Access.
Failure modes, high-throughput bottlenecks, and real FAANG implementation decisions.
Interactive system topology diagrams, live parameter simulators, and downloadable SVG charts.
Staff-level multiple-choice quiz questions with instant feedback and answer explanations.
Firebase Google authentication automatically syncs your completed topics and quiz scores.
How clear and actionable was this distributed systems breakdown?