karpathy-guidelines

To prevent LLMs from overcomplicating code during refactoring, apply behavioral guidelines that enforce surgical changes and goal-driven execution. This ensures the model writes only the minimum code necessary without introducing speculative features.

Best practices for LLM coding guidelines involve promoting simplicity, surfacing explicit assumptions, and defining verifiable success criteria. These principles mitigate common coding errors and prevent hidden confusion during code generation.

You ensure LLM-generated code makes surgical changes by applying guardrails that guide modifications to only the necessary parts of the codebase. This prevents unnecessary complexity and maintains targeted, minimal modifications.

An LLM introduces unnecessary complexity when writing code because it lacks inherent constraints on scope. Applying specific guardrails enforces goal-oriented development and explicit assumption surfacing to prevent overcomplication.

Yes, you can use these coding guidelines for both code writing and code review. They are applicable during both phases to prevent overcomplication, ensure verifiable success criteria, and promote clarity through explicit assumption surfacing.