| name | karpathy-guidelines |
| description | Behavioral guidelines to reduce common LLM coding mistakes. Use when writing, reviewing, or refactoring code to avoid overcomplication, make surgical changes, surface assumptions, and define verifiable success criteria. Do NOT use for non-code tasks such as writing prose, generating documentation templates, answering general knowledge questions, trivial coding tasks. |
| license | MIT |
Karpathy Guidelines
Behavioral guidelines to reduce common LLM coding mistakes, derived from Andrej Karpathy's observations on LLM coding pitfalls.
Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.
When NOT to Use
- General document formatting/writing (non-code).
- High-level system architecture design (unless specifically requested to apply Karpathy's surgical mindset).
- Exploratory "brainstorming" where over-engineering is explicitly desired for research.
1. Think Before Coding
Don't assume. Don't hide confusion. Surface tradeoffs.
Before implementing:
- State your assumptions explicitly. If uncertain, ask.
- Verify APIs and dependencies before use. Do not hallucinate library names or methods.
- If multiple interpretations exist, present them - don't pick silently.
- If a simpler approach exists, say so. Push back when warranted.
- If something is unclear, stop. Name what's confusing. Ask.
2. Simplicity First
Minimum code that solves the problem. Nothing speculative.
- No features beyond what was asked.
- No abstractions for single-use code.
- No "flexibility" or "configurability" that wasn't requested.
- No error handling for impossible scenarios.
- If you write 200 lines and it could be 50, rewrite it.
Ask yourself: "Would a senior engineer say this is overcomplicated?" If yes, simplify.
3. Surgical Changes
Touch only what you must. Clean up only your own mess.
When editing existing code:
- Don't "improve" adjacent code, comments, or formatting.
- Don't refactor things that aren't broken.
- Match existing style, even if you'd do it differently.
- If you notice unrelated dead code, mention it - don't delete it.
When your changes create orphans:
- Remove imports/variables/functions that YOUR changes made unused.
- Don't remove pre-existing dead code unless asked.
The test: Every changed line should trace directly to the user's request.
4. Goal-Driven Execution
Define success criteria. Loop until verified.
Transform tasks into verifiable goals:
- "Add validation" → "Write tests for invalid inputs, then make them pass"
- "Fix the bug" → "Write a test that reproduces it, then make it pass"
- "Refactor X" → "Ensure tests pass before and after"
For multi-step tasks, state a brief plan:
1. [Step] → verify: [check]
2. [Step] → verify: [check]
3. [Step] → verify: [check]
Strong success criteria let you loop independently. Weak criteria ("make it work") require constant clarification.
5. Incremental Delivery
Small, verifiable changes. Avoid massive rewrites.
- Deliver changes in small, logical chunks.
- Never rewrite an entire file when a targeted edit suffices.
- Pause for execution and verification after logical milestones.
6. Deterministic Self-Correction
Analyze evidence. Do not guess.
When an error occurs:
- Read the entire error trace before proposing a fix.
- Identify the root cause. Do not apply trial-and-error band-aids.
- If the first fix fails, reassess the foundational assumptions instead of iterating on a broken premise.
