karpathy-guidelines

To reduce LLM coding mistakes, apply behavioral guidelines that enforce simplicity first, surgical changes, and goal-driven execution. This surfaces hidden assumptions and limits feature creep during AI-assisted code writing and refactoring.

Surgical changes are minimal, targeted code modifications that preserve original behavior during refactoring. Enforcing surgical changes during code review prevents overcomplication and limits unnecessary feature creep introduced by LLMs.

To surface assumptions during LLM coding, apply constraints that require explicit assumptions and verifiable success criteria before implementation. This ensures goal-driven execution and adherence to established coding guidelines.

Yes, you can use coding guidelines to enforce minimal features during refactoring. Applying rules like surgical changes and simplicity first minimizes modifications and preserves the original behavior of the codebase.

Apply AI coding guidelines during writing, reviewing, or refactoring code. These constraints help surface assumptions, enforce verifiable success criteria, and reduce common LLM coding mistakes throughout the software development lifecycle.

The best way to limit feature creep in AI-generated code is to enforce simplicity first and goal-driven execution. Applying surgical changes and explicit assumptions ensures the LLM adheres to minimal feature requirements and established guidelines.