Don't you think Claude is literally pretends to be stupid? It doesn't listen to you properly, and many people ignore this so they're burning more tokens.
I've been scrolling on the Github and found the repo that fully changes the game.
One file, 44k+ new stars in seven days, and most people using Claude Code have never heard of it.
Here is what it is, why it exists, and how to install it in two minutes.
🚨 Before we start, me and @noisyb0y1 are launching a completely free AI course, everything you need to go from zero to building real projects with AI in a weekend.
Leaving all details at the end of this article, let's get started:
Andrej Karpathy, co-founder of OpenAI, former Tesla AI director, shifted to roughly 80% agent-driven development.
Then he posted an honest breakdown of everything that was wrong with how AI coding tools actually behave.
Three quotes read them slowly:
"The models make wrong assumptions on your behalf and just run along with them without checking. They don't manage their confusion, don't seek clarifications, don't surface inconsistencies, don't present tradeoffs, don't push back when they should"
"They really like to overcomplicate code and APIs, bloat abstractions, don't clean up dead code. Implement a bloated construction over 1000 lines when 100 would do"
"They still sometimes change and remove comments and code they don't sufficiently understand as side effects, even if orthogonal to the task"
Every developer using Claude Code recognized this instantly.
A developer named Forrest Chang read the post and built a fix, one *CLAUDE.md** *file, four principles, it rewires how Claude Code behaves on every task and 45,000 people installed it this week.
What it actually does
A *CLAUDE.md* file is a behavior configuration that Claude Code reads automatically at the start of every session, whatever is in it becomes part of how Claude thinks about your project.
This file adds four principles derived directly from Karpathy's complaints. Each one fixes a specific habit.
Here is the full file, copy it right now.
That is the whole thing, four principles, one file.
What each principle actually changes
Principle 1: Think Before Coding, before this file:
After this file:
The confusion surfaces before the work happens, not after.
Principle 2: Simplicity First, before this file:
After this file:
Would a senior engineer say the first version is overcomplicated? Yes. So Claude rewrites it.
Principle 3: Surgical Changes
Before this file, you ask Claude to fix one bug in a 200-line file. You get back:
You asked it to fix a null check. It added type hints, renamed variables, rewrote comments, and refactored the function signature.
After this file:
One change. The one you asked for. Nothing else.
Principle 4: Goal-Driven Execution
This is the one Karpathy called the key insight.
"LLMs are exceptionally good at looping until they meet specific goals. Don't tell it what to do. Give it success criteria and watch it go."
Before this file:
After this file:
You stop babysitting -> you start reviewing.
The most important part, How to install it:
Two options, pick one.
**Option A: Plugin (installs across all your projects)
**Open Claude Code and run these two commands one after the other:
Active immediately. Works in every project from this session forward.
Option B: Per-project (drop it into one repo) you can find it in the repo on github link that I left at the start of this article.
How to Add Your Own Rules On Top
The file is designed to be merged with project-specific instructions. After the four principles, add a section like this:
Your project rules sit on top of the four principles, both apply
How you know it is working:
Three things change that you will notice immediately.
Your diffs get cleaner. Only what you asked for shows up. No reformatted functions. No renamed variables. No improved comments you never mentioned.
Clarifying questions come before implementation, claude stops guessing and starts asking. You spend less time throwing away wrong work.
Code is simpler the first time. No rewrites because it overengineered the solution. No abstractions that made sense to Claude but nobody asked for.
Cherry on top for the AI frens:
if you made it this far, you deserve it: a completely free AI course, built for people who actually want to learn.
> https://www.skool.com/ai-builderss/classroom**
More is coming, and it's going to be bigger than anything we've dropped so far stay tuned, grab your spot early, and get ready to fall deep into the AI rabbit hole.
- Leo**
