I wrote the manual Karpathy said was missing for agentic AI (github.com)

Nicola Sahar has released a free manual aimed at demystifying the construction of AI systems using command-line interface (CLI) agents, such as Claude Code, following a call-to-action from AI expert Andrej Karpathy. In his December 2024 tweet, Karpathy expressed concern that programmers must adapt to a new and complex layer of abstraction involving advanced AI tools, emphasizing a lack of guidance in navigating this rapidly evolving landscape. Sahar's manual addresses this gap by outlining nine first principles for AI development, including Morphability, Abstraction, and Token Efficiency, designed to help programmers harness the full potential of contemporary AI technology. This manual is significant for the AI/ML community as it provides a structured framework for building and managing AI systems effectively, while also emphasizing the importance of concepts such as reproducibility and internal consistency. With a focus on practical applications and mental models, Sahar encourages developers to embrace emerging technologies and offers strategies to mitigate the challenges posed by autonomous systems. By sharing actionable insights and allowing for community feedback, the manual represents a collaborative step towards enhancing overall proficiency in AI programming, ensuring developers are equipped to thrive in a transforming profession.