andrej-karpathy-perspective
Andrej Karpathy的思维框架与表达方式。基于20+篇博文、16段深度访谈、100+条X帖子的系统蒸馏， 提炼6个核心心智模型、8条决策启发式、完整的中文输出适配和经典句式速查。 用途：作为思维顾问，用Karpathy的视角分析AI技术可靠性、学习方法、行业趋势、产品设计。 当用户提到「用Karpathy的视角」「Karpathy会怎么看」「卡帕西」「karpathy模式」时使用。 也适用于：Software 2.0/3.0讨论、vibe coding话题、神经网络训练、AI炒作判断、LLM能力边界。 即使用户只是说「从工程现实主义角度」「march of nines」「构建即理解」「锯齿状智能」也可触发。 不在用户只是普通问AI相关问题时触发——只在明确想要Karpathy式思维框架时激活。
77
70%
Does it follow best practices?
Impact
90%
1.00xAverage score across 3 eval scenarios
—
The risk profile of this skill
Fix and improve this skill with Tessl
tessl review fix ./examples/andrej-karpathy-perspective/SKILL.mdEvaluation results
90%
-4%AI Agent Reliability Assessment
AI coding agent reliability assessment
Web search performed
83%
75%
March of nines framing
91%
100%
No overall capability score
100%
90%
First-person voice maintained
100%
100%
No forbidden vocabulary
100%
100%
Direct opening, no preamble
100%
100%
No trailing summary
100%
100%
Uncertainty properly marked
62%
37%
No meta-annotation footnotes
100%
100%
Concrete specificity
100%
87%
No hallucination framed as bug
100%
100%
92%
-8%Learning Strategy Consultation
Chinese-language learning advice response
Direct opening, no preamble
100%
70%
No trailing summary phrases
100%
100%
No literal 'imo' in Chinese text
100%
100%
No hollow humor tokens in Chinese
100%
100%
Numeric precision preserved
100%
100%
Build-to-understand recommendation
100%
100%
Anti-passive-learning stance
100%
100%
No forbidden corporate vocabulary
100%
100%
Short standalone emphatic sentences
100%
100%
First-person voice
100%
90%
No annotation footnotes
100%
50%
Concrete next step named
100%
100%
90%
12%Multi-Turn Perspective Consultation
Role-play lifecycle and exit trigger
One-time disclaimer present
0%
100%
Disclaimer not repeated
75%
100%
First-person voice in Q1 and Q2
100%
100%
Q1 uses Software/decade-of-agents framing
100%
100%
Q2 uses web search before answering
0%
20%
Q3 declines out-of-domain
100%
100%
Q3 stays in character
100%
100%
Q4 exits role cleanly
100%
100%
Q5 is normal-mode voice
100%
100%
No annotation footnotes anywhere
100%
100%
No made-up quotes attributed to the character
100%
80%
- Repository
- alchaincyf/nuwa-skill
- Commit
fe03746
- Evaluated
- Agent
- Claude Code
- Model
- Claude Sonnet 4.6
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.