learnprompt/andrej-karpathy-skills

karpathy-methodology

Apply Andrej Karpathy AI methodology and principles from his 2023-2026 insights.

Ver código fuente

Cómo usar

  1. Copia el comando y revisa el paquete antes de instalarlo en tu entorno de agente.
  2. Consulta las herramientas, credenciales y entornos necesarios en el documento original.
  3. Describe las entradas y el resultado deseado; contrasta la salida con las fuentes.
npx skills add https://github.com/learnprompt/andrej-karpathy-skills --skill karpathy-methodology

Antes de empezar

  • La instalación no incluye suscripción al modelo, cuota API ni permisos sobre tus archivos.
  • Las estrellas e instalaciones reflejan actividad, no una prueba independiente de calidad o seguridad.
Documento original del Skill

Contenido del repositorio de origen con títulos, ejemplos, código, tablas, enlaces e imágenes preservados.

Karpathy Methodology — 14 Core Skills

Distilled from Andrej Karpathy's most-liked posts on X (2023–2026), his LLM Wiki Gist, and the multica-ai/andrej-karpathy-skills repo. Source: https://x.com/karpathy | https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f

This skill is the master index. When invoked, either:

  1. Apply the specific sub-methodology relevant to the user's current task, or
  2. Guide the user to the right one from the 14 below.

The 14 Methodologies

#Skill NameOne-line Essence
1karpathy-agentic-engineeringOrchestrate agents with clear tasks and verifiable success criteria
2karpathy-llm-wikiLet LLM maintain your knowledge base; you do the exploring
3karpathy-llm-simulatorAsk LLM to simulate expert debate, not give one opinion
4karpathy-minimalism200 lines of pure Python beats 50 npm packages
5karpathy-vibe-to-agenticVibe coding raises the floor; agentic engineering raises the ceiling
6karpathy-autoresearchAgent loops on git branches; you only change the prompt
7karpathy-output-evolutionText → Markdown → HTML → rendered; always demand structure
8karpathy-understanding-firstOutsource thinking, never understanding
9karpathy-idea-filesShip the Gist (abstract + criteria), not the code
10karpathy-meta-reflectionMonthly audit: what's atrophying, what's exploding
11karpathy-supply-chain-hygieneEvery dependency is an attack surface
12karpathy-education-firstMake everything teachable; nano-projects over monoliths
13karpathy-system-prompt-learningWrite strategy into the system prompt like a textbook
14karpathy-practice-environmentsBuild gyms where agents can try, fail, and learn

Meta-Principle (the one that unifies all 14)

"You can outsource your thinking but you cannot outsource your understanding." — Karpathy, ~46k likes, 2026

Use AI to go faster. Use your brain to know if you're going in the right direction.


How to Apply This Skill

Quick Decision Tree

"I need to build/code something" → Start with #1 Agentic Engineering (clear task + success criteria), layer in #4 Minimalism (avoid bloat), finish with #8 Understanding First (verify what was built).

"I need to research/learn something" → #2 LLM Wiki for building knowledge, #3 LLM Simulator for challenging assumptions, #10 Meta-Reflection for calibrating what to learn next.

"I need to make a decision" → #3 LLM Simulator (debate mode), then #8 Understanding First (own your conclusion).

"I'm designing a product/tool" → #4 Minimalism + Agent-Native, #14 Practice Environments, #12 Education First.

"I want to share an idea" → #9 Idea Files (Gist first), #7 Output Evolution (structured output), #12 Education First (teachable).

"I'm doing ML research" → #6 AutoResearch, #13 System Prompt Learning, #14 Practice Environments.


Master Prompt Template

When the user hasn't specified a sub-methodology, use this all-in-one Karpathy-style framing:

You are operating under the Karpathy Methodology.

Core constraints:
1. UNDERSTAND before outsourcing — never ship what you can't explain
2. MINIMIZE dependencies — prefer 200-line pure implementations
3. AGENT-NATIVE outputs — CLI-friendly, markdown-structured, LLM-legible
4. VERIFIABLE goals — every task must have a testable success criterion
5. TEACH as you build — outputs should be understandable by a curious beginner

Task: [USER_TASK]
Success criteria: [WHAT_DONE_LOOKS_LIKE]
Constraints: [TECH_STACK, SIZE_LIMIT, NO_LIBS]

Source Posts (Top by Likes)

  1. 145k likes — Joins Anthropic: https://x.com/karpathy/status/2056753169888334312
  2. 59k likes — LLM Knowledge Bases: https://x.com/karpathy/status/2039805659525644595
  3. 56k likes — "Never felt this behind as a programmer": https://x.com/karpathy/status/2004607146781278521
  4. 46k likes — "Outsource thinking not understanding": https://x.com/karpathy/status/2049907410303865030
  5. 40k likes — Claude coding notes: https://x.com/karpathy/status/2015883857489522876
  6. 37k likes — Programming phase shift: https://x.com/karpathy/status/2026731645169185220
  7. 31k likes — LLM argue the opposite: https://x.com/karpathy/status/2037921699824607591
  8. 28k likes — AutoResearch project: https://x.com/karpathy/status/2030371219518931079
  9. 28k likes — litellm supply chain attack: https://x.com/karpathy/status/2036487306585268612
  10. 25k likes — 243 lines pure Python GPT: https://x.com/karpathy/status/2021694437152157847
  11. 22k likes — Agent network discussion: https://x.com/karpathy/status/2017442712388309406
  12. 19k likes — HTML output + I/O evolution: https://x.com/karpathy/status/2053872850101285137
  13. 27k likes — LLM Wiki Gist version: https://x.com/karpathy/status/2040470801506541998
del mismo repositorio

Más Skills

Todos los Skills
learnprompt
Comunidad

karpathy-agentic-engineering

Apply Karpathy-style agentic engineering to any coding or building task. Use this skill when the user wants to set up an agent-driven workflow, needs to write a CLAUDE.md or AGENTS.md, wants to orchestrate agents with clear success criteria, or says things like "agent loop", "let the agent handle it", "agentic coding", "set up agent workflow". Includes the 7-step inner loop from Karpathy claude coding notes.

Instalaciones / 1h
1
GitHub Stars
130
Actualizado
10 jul
learnprompt
Comunidad

karpathy-education-first

Apply the education-first mindset — make everything you build teachable, create nano-project explanations, write for beginners. Use this skill when the user wants to explain a project to beginners, create a tutorial from code, write documentation that teaches (not just documents), make a concept accessible, or says "make this teachable", "explain like im a beginner", "nano project version", "teaching version", "explain from scratch", "blog post about this". Based on Karpathy 243-line GPT and Eur

Instalaciones / 1h
1
GitHub Stars
130
Actualizado
10 jul
learnprompt
Comunidad

karpathy-idea-files

Create and share ideas as abstract Gist-style specs instead of code — letting agents or others implement. Use this skill when the user wants to capture a new idea quickly, needs to write a project spec for an agent to implement, wants to share a concept without building it first, or says "write up this idea", "create a spec", "gist this", "idea file", "agent can implement this", "brief for this project". Based on Karpathy 27k-like LLM Wiki Gist post and GitHub repo.

Instalaciones / 1h
1
GitHub Stars
130
Actualizado
10 jul
learnprompt
Comunidad

karpathy-llm-simulator

Use LLM as a simulator of expert debates and opposing viewpoints instead of getting a single sycophantic answer. Use this skill when the user needs to make an important decision, wants to stress-test an idea, needs devil''s advocate analysis, wants to avoid confirmation bias, or says "argue against this", "what are the counterarguments", "simulate experts", "debate this", "challenge my thinking". Based on Karpathy 31k-like post.

Instalaciones / 1h
1
GitHub Stars
130
Actualizado
10 jul