karpathy-guidelines

To prevent overcomplicating code during refactoring, apply behavioral guidelines that enforce simplicity, encourage surgical changes to only necessary sections, and surface assumptions to avoid unintended side effects.

Best practices for LLM-assisted coding include thinking before coding to surface tradeoffs, writing minimum viable code without speculation, and defining verifiable success criteria to maintain software development quality.

You can make AI write surgical code changes by applying guidelines that restrict modifications to only necessary code sections, preventing overcomplication and preserving existing software maintainability during development.

To define verifiable success criteria, establish goal-driven execution guidelines that emphasize cautious development, explicit assumptions, and clear problem-solving objectives before writing or reviewing any code with an LLM.

You should use coding guidelines for LLM code generation and review whenever you need to improve code quality, prevent unintended side effects during refactoring, and ensure cautious, goal-oriented problem-solving in software development.

Applying simplicity principles to LLM code generation improves maintainability by promoting minimum viable code without speculation, ensuring surgical changes, and surfacing assumptions to prevent overcomplication in software development.