Skip to content
Get packs

Karpathy Perspective Thinking Skill

Makes the agent reason and speak in the style of Andrej Karpathy's public views on AI, learning and engineering

Source Download skill
Repository
alchaincyf/nuwa-skill
Stars
37,170
Licence
MIT
Files
8 files

What it does

Once installed, the agent can take on a thinking framework built from Andrej Karpathy's public blog posts, interviews and posts. It answers questions about AI reliability, learning by building, LLM limits and AI hype in his voice, using his mental models and phrasing, and it notes that these are inferences from public statements rather than his actual opinions. The skill tells the agent to research current facts with web search before answering factual questions, so it can use live sources.

Great for

  • Developers or students who want to discuss AI ideas through the lens of a well-known engineer's public thinking
  • Writers or educators looking for a structured way to frame AI trends and learning advice
  • People who want a critical, engineering-focused view of AI product claims

Skip it if

  • Anyone who needs an authentic statement of what Karpathy actually believes, since the output is a model-based inference from public material and may be wrong or outdated (the file says its research cutoff is 2026-04-05)
  • Business strategy, marketing or policy decisions, which the skill itself says fall outside its scope
  • Anyone who would be misled by a roleplay persona presenting invented-sounding first-person opinions as a real person's views

Try asking

  • “Use the Karpathy perspective to assess whether this AI product demo is reliable enough to deploy”
  • “What would Karpathy say about vibe coding and learning to build neural networks from scratch?”
  • “Karpathy would look at the march of nines for this agent workflow, how would he frame the reliability gap?”
  • “Give me a Software 3.0 view of how English as a programming language changes my team's work”
Step-by-step guide: how to use Karpathy Perspective Thinking SkillInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 8 original files, the MIT licence and an attribution note.

    andrej-karpathy-perspective.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip andrej-karpathy-perspective.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip andrej-karpathy-perspective.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Ask for a framework-style answer on a specific question rather than general chat, since the skill is meant to activate only when you explicitly ask for a Karpathy-style view.
  • Say 退出 or 不用扮演了 to leave the persona and get normal answers.
  • Check any cited quotes or facts against the original sources before relying on them, because the output blends research with inference.

What's inside

  • FIDELITY.md
  • references/research/01-writings.md
  • references/research/02-conversations.md
  • references/research/03-expression-dna.md
  • references/research/04-external-views.md
  • references/research/05-decisions.md
  • references/research/06-timeline.md
  • SKILL.md

Redistributed unmodified from alchaincyf/nuwa-skill under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.