karpathy-guidelines

To prevent overcomplication during refactoring, apply behavioral guidelines that enforce surgical edits and minimum viable code. This ensures the LLM touches only necessary code, matches existing styles, and avoids speculative features or unnecessary side effects.

Best practices for writing code with LLMs involve stating explicit assumptions, surfacing tradeoffs, and defining verifiable success criteria. This goal-driven execution approach ensures the LLM focuses on clarity, efficiency, and maintainable software development.

Unnecessary features and abstractions occur when LLMs lack simplicity constraints. Enforcing guidelines that promote minimum viable code and explicit assumption stating prevents the LLM from introducing speculative complexity during code generation.

To ensure surgical and minimal code modifications, use guidelines that restrict the LLM to touching only necessary code and cleaning up only its own mess. This prevents unwanted side effects and maintains existing code style during refactoring.

You need coding guidelines for LLM-generated code during code writing, review, and refactoring. These guidelines mitigate common LLM coding errors by establishing verifiable success criteria and ensuring goal-oriented execution throughout the software development process.