The origin story
What is vibe coding?
Vibe coding is building software by talking to an AI agent: you describe, it writes, you steer. The term has a specific birthplace, and Superwhisper was in the room.
February 2025
Where the term comes from
Andrej Karpathy, OpenAI co-founder and former director of AI at Tesla, posted in early 2025 about a new way he was building software. He gave it a name, and the name stuck hard enough that it now has a Wikipedia page and a spot in dictionary watchlists.
Andrej Karpathy
@karpathy · 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. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard.
I ask for the dumbest things like "decrease the padding on the sidebar by half" because I'm too lazy to find it. I "Accept All" always, I don't read the diffs anymore. When I get error messages I just copy paste them in with no comment, usually that fixes it. The code grows beyond my usual comprehension, I'd have to really read through it for a while. Sometimes the LLMs can't fix a bug so I just work around it or ask for random changes until it goes away.
It's not too bad for throwaway weekend projects, but still quite amusing. I'm building a project or webapp, but it's not really coding - I just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works.
Read the post again and notice the mechanics. Karpathy isn't typing prompts into Cursor. He's talking to it, through Superwhisper. The original description of vibe coding is a description of coding by voice.
Most explainers drop that detail, but it matters. Vibe coding works because you hand the agent intent as fast as you can describe it, then judge the output. Typing your prompts puts a 40-words-per-minute bottleneck in the middle of a workflow whose whole point is speed.
The loop
How vibe coding actually works
The loop is short. Say what you want. The agent writes code, runs it, and reports back. You look at the result, not the diff, and say what's wrong or what's next. Repeat until the thing works.
What changes compared to normal programming is where your judgment goes. You stop sweating syntax and start directing: scope the next step, catch the agent drifting, decide when something is good enough. The skill ceiling moved, it didn't disappear.
A practical setup needs two pieces. An agent that can edit and run code: Cursor, Claude Code, Codex, or OpenCode. And a way to talk to it: Superwhisper turns your speech into clean prompts in any of them, with a hotkey, on Mac, Windows, and iPhone.
In good company
Who codes like this
Pieter Levels
Tried @superwhisperapp today. Very nice. Lets me talk to Cursor and then it codes for me, just gets it right.
Related
Keep exploring
Voice coding
The mechanics of writing code by talking.
Superwhisper + Claude Code
Drive Claude Code sessions by voice.
Superwhisper + Codex
Dictate to OpenAI's coding agent.
Words per minute test
Measure how much faster you speak than type.
Support