The Vibe Coding Compromise
In February 2025, Andrej Karpathy, co-founder of OpenAI, coined the term "vibe coding" on X.
https://x.com/karpathy/status/1886192184808149383
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.
Karpathy himself limits this to "throwaway weekend projects." It's not a production-grade methodology. It's a compromise: if it runs, it's good enough.
What happens when you bring this compromise to writing?
Student Papers Today
As a faculty member, I encounter two types of AI-assisted term papers.
First, the desperate type. Hallucinated references to books that don't exist, listed as sources. Point out "this book doesn't exist" and it's over—along with all the student's credits for the semester.
Second, the savvy type. They leverage research functions, throw in a topic, and produce something that looks like a paper. The prose smells like AI, but there's no proof. You sense the gap from their usual ability, but you have no way to verify it. This is the crucial difference from the copy-paste plagiarism of the past.
For now, we can still detect the delta between pre-AI and post-AI work for individual students. But in a few years, generations raised on AI-generated text from elementary school onward will enter universities. The baseline for authorship will vanish. The question "who wrote this?" will lose its meaning.
This isn't just a student problem. It's a question for everyone whose work involves writing.
What Is Vibe Writing?
Let's flip the meaning of "vibe." Not vibes as in atmosphere. Vibe as in the tremor of the soul.
Vibe writing follows this process:
The human produces the esquisse—the foundational design, plot, idea fragments, models, and formulations. The seeds.
AI assembles a draft.
The human examines and revises paragraph by paragraph.
Iterate until the manuscript is complete.
Think of an electric-assist bicycle. You pedal. The motor just helps on the hills.
The bottleneck for writers is output. The image of what you want to write is there. You know what you want to say. But linearizing it, making it logical, making it concise—that's where you get stuck. AI clears this bottleneck.
But there are things AI cannot do. Fact-checking. Cross-referencing with your own experience. Adjusting the texture of the prose. Only humans can do these. If you let the draft drag you along and ignore your discomfort, it's no longer your writing.
Ownership and Responsibility
The question isn't "who wrote it?" It's "who takes responsibility?"
AI generates the draft. The human signs it. If you sign it, you must make every sentence your own. If something feels off, fix it. If you can't fix it, you have no right to sign.
Back to student papers. The issue isn't "did you use AI?" It's "did you think for yourself?" Throwing in a topic and submitting whatever comes out—there's no soul in that. It's vibe coding at its worst.
Hold the Soul
Don't let AI write for you. Use AI to write.
The question is the same for students and writers alike. Is the esquisse at the core your own? Can you take responsibility for every sentence?
Write with your soul trembling. That's vibe writing.