AI & ML Paradigm Challenge

Stop giving your coding agents 'best practices' advice; it's actually making them worse at writing code.

April 14, 2026

Original Paper

Do Agent Rules Shape or Distort? Guardrails Beat Guidance in Coding Agents

Xing Zhang, Guanghui Wang, Yanwei Cui, Wei Qiu, Ziyuan Li, Bing Zhu, Peiyang He

arXiv · 2604.11088

The Takeaway

The study found that positive guidance (like 'follow code style') degrades performance, whereas only negative constraints (like 'do not refactor') are beneficial. It overturns the common intuition that more helpful guidance leads to better AI performance.

From the abstract

Developers increasingly guide AI coding agents through natural language instruction files (e.g.,this http URL, .cursorrules), yet no controlled study has measured whether these rules actually improve agent performance or which properties make a rule beneficial. We scrape 679 such files (25,532 rules) from GitHub and conduct the first large-scale empirical evaluation, running over 5,000 agent runs with a state-of-the-art coding agent on SWE-bench Verified. Rules improve performance by 7--14 perce