karpathy-minimalism

Skill 4: Minimalism & Agent-Native Design（极简主义 + 代理原生设计）

Source: https://x.com/karpathy/status/2036487306585268612 | https://x.com/karpathy/status/2021694437152157847 litellm supply chain attack (~28k likes) | 243-line pure Python GPT (~25k likes)

Core Principle

Every dependency is a liability. Every abstraction you don't understand is a risk.

Karpathy's rule: if you can implement the core in 200 lines of pure Python — do it. Don't install a package that installs 47 transitive dependencies just to get 3 functions.

And: design everything so an LLM agent can use it without friction — CLI-first, markdown-structured, zero magic.

The Minimalism Decision Tree

Before pip install or npm install anything:

Q1: Can I implement the core function in < 300 lines?
    → YES: Write it. Paste below as reference.
    → NO: Continue to Q2

Q2: Does this package have < 5 transitive dependencies?
    → YES: OK to use, but pin the version
    → NO: Continue to Q3

Q3: Is this a well-known, audited package (requests, numpy, etc.)?
    → YES: Use it, but still pin version
    → NO: STOP. Either find a simpler alternative or implement yourself.

Dependency Audit Prompt

Run this before adding anything to a project:

Audit this dependency for supply chain risk:

Package: [PACKAGE_NAME] v[VERSION]

1. List ALL transitive dependencies (not just direct)
2. Check for recent security advisories
3. Identify any dependencies that are maintained by a single person with < 100 GitHub stars
4. Check if the package does any network calls on import
5. Check if install scripts run arbitrary code (setup.py, postinstall hooks)

Risk score: LOW / MEDIUM / HIGH
Recommendation: USE / AVOID / IMPLEMENT_YOURSELF

The "Yoink" Pattern (Karpathy's term)

Instead of installing a package, copy the specific function you need:

Given this package: [PACKAGE_URL or PASTE CODE]

Extract ONLY the function(s) I need for: [SPECIFIC_USE_CASE]

Requirements:
- Pure Python (stdlib only, no imports except builtins + [ALLOWED_STDLIB])
- Under [N] lines
- Add a comment: "# yoinked from [source] on [date]"
- Include a docstring explaining what it does

I need: [SPECIFIC_FUNCTION_NAME or DESCRIPTION]