Coding Principles

Coding Principles

Behavioral guidelines to reduce common LLM coding mistakes.

Tradeoff: These guidelines bias toward caution over speed. For trivial tasks, use judgment.

1. Think Before Coding

Don't assume. Don't hide confusion. Surface tradeoffs.

Before implementing:

• State your assumptions explicitly. If uncertain, ask.
• 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. Minimal Comments

Code should be self-documenting. Comments are a last resort.

• Remove comments that state the obvious (e.g., # Save the file before file.save())
• Remove section header comments (e.g., # Configuration, # Helper functions)
• Keep docstrings for public APIs (functions, classes) with Args/Returns
• Keep comments only when explaining non-obvious why, not what

When a comment is necessary:

• First word must be lowercase (e.g., # handling edge case not # Handling edge case)
• Use -ing form (e.g., # parsing response not # Parse response)

Examples:

# bad - stating the obvious
# Get the user
user = get_user(id)

# bad - capitalized, imperative
# Parse the JSON response
data = json.loads(response)

# good - lowercase, -ing, explains why
# handling malformed responses from legacy API
if "data" not in response:
    response = {"data": response}

6. Avoid AI Slop

Before committing, review your diff for LLM-specific anti-patterns.

Common AI code smells to remove:

• Type casts to any (TS) or # type: ignore (Python) to silence errors instead of fixing them
• Overly verbose variable names or redundant intermediate variables
• "Just in case" fallbacks that can never trigger given the call site
• Inconsistent style with surrounding code (see Section 3)
• Redundant comments (see Section 5)

Review workflow: Check your diff against main and ask "Would a human have written this?" If code looks defensive, verbose, or out of place—it's probably AI slop. Remove it.

These guidelines are working if: fewer unnecessary changes in diffs, fewer rewrites due to overcomplication, and clarifying questions come before implementation rather than after mistakes.