Phase 0: Vibe Coding Fundamentals · 15 min · Claude Code, Cursor, Lovable, v0, Bolt.new, Replit Agent, or similar
Objective
Your mission
By the end of this lesson, you'll have a vibe_coding_starting_point.md that captures your definition of vibe coding, where you sit on the spectrum, and what you want to build. You'll revisit this at the end of the course to see how much you've grown.
You need: A text editor or notes app. No tools to install yet.
Time: 15 minutes.
Step 1: read the origin (3 min)
In February 2025, Andrej Karpathy — co-founder of OpenAI, former Director of AI at Tesla — posted this on X:
"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 are getting too good. Also I just talk to Composer with SuperWhisper so I barely even touch the keyboard. 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 just see stuff, say stuff, run stuff, and copy paste stuff, and it mostly works."
Key context: Karpathy is one of the most accomplished AI researchers alive. He later clarified that vibe coding "empowers trained professionals to write a lot more software that would otherwise never be written." He was talking about weekend projects, not production systems at companies.
Write this in your file:
## What I Think Vibe Coding Is (Before)
[Write 2-3 sentences in your own words. Don't copy-paste from above.]
Step 2: find your spot on the spectrum (5 min)
Vibe coding isn't binary. It's a spectrum:
- Full vibe: Describe, accept, ship. Don't read the code. Fast, risky.
- Vibe + review: Describe, accept, then read for red flags before shipping.
- AI-assisted: You write code, AI fills gaps. You review every change.
The right spot depends on the project:
| Project | Where on spectrum |
|---|
| Landing page for a friend | Full vibe — low risk, no user data |
| Internal tool your team uses daily | Vibe + review — check for red flags |
| Customer app with user data | Vibe + review, then get an engineer to check auth/security |
| Production system at a company | AI-assisted — review every line |
Write this in your file:
## Where I Sit on the Spectrum
Project I want to build: [name it]
Where on spectrum: [full vibe / vibe + review / AI-assisted]
Why: [1-2 sentences explaining your choice based on risk]
Step 3: name your excitement and your concern (5 min)
Write this in your file:
## What Excites Me
[One specific thing — not "it's cool." E.g., "I can build the habit tracker I've been thinking about for months without learning React first."]
## What I'm Cautious About
[One specific thing — not "it might not work." E.g., "I'm worried I'll ship something with a security hole I don't know about because I can't read the code."]
Step 4: what vibe coding is NOT (2 min)
Read these. Don't write anything — just absorb:
- Not "AI does everything, you do nothing." Even Karpathy pastes error messages and gives feedback. The human is always in the loop.
- Not a replacement for engineering. Vibe-coded apps have leaked API keys and exposed user data. We'll cover those real failures in Phase 8.
- Not coding without understanding. The best vibe coders understand what the code does conceptually. They can spot red flags and debug when things break. This course makes you that kind of vibe coder.
The spectrum, not the extreme
Vibe coding is a spectrum, not a switch. At one end, an engineer uses AI for boilerplate while reviewing every line. At the other, a non-technical person ships an app without reading code. Both are valid. The danger is at the zero-review extreme — and this course keeps you off that end.
A friend wants a customer app that stores users' addresses, built in full-vibe mode: describe, accept, ship, never read the code. Where on the spectrum does it belong?
User data raises the risk. The spectrum is set by what breaks if the app fails, so review the output and get an engineer to check authentication and security before real users arrive.