perspective-andrej-karpathy

Software 3.0 frames AI deployment around natural language prompting of LLMs, whereas Software 2.0 focuses on data-driven neural network training. This Skill uses Karpathy's framework to analyze how AI software evolves and is deployed across engineering and product contexts.

Jagged intelligence describes uneven AI capabilities across different tasks. You can use this Skill to apply the concept and evaluate where AI workflows excel or fail, helping to critique product design, assess safety risks, and identify deployment limitations.

Yes, this Skill applies Karpathy’s education-centric approach to design teaching materials and tutorials. It encodes his pragmatic engineering guidance and mental models to shape learning workflows for AI education strategies.

The LLM-as-OS concept frames large language models as operating systems managing resources and applications. This Skill applies that mental model to guide your analysis of AI trends, evaluating product design and user workflows from that operating system perspective.

Jagged intelligence highlights capability mismatches but does not provide quantitative metrics. This Skill uses the concept to shape qualitative analysis of user workflows and deployment risks, meaning it guides reasoning rather than generating automated evaluation scores.