Getting the Most Out of AI Coding

Wes Bos and Scott Tolinski share actionable techniques for maximizing productivity with AI-powered coding tools—including Cursor, Claude, ChatGPT, and Windsurf. They begin by advising developers to scaffold applications manually to maintain context and recharge ownership over generated code.

The hosts cover advanced prompt crafting: being clear, breaking down tasks, adding <XML> or <tags> to define context more precisely, and building reusable rule sets (e.g., Cursor rules or Copilot rules). They also explain how to automate rule generation from existing codebases and store rules in llm.txt files for consistency across projects.

Further insights include methods to maintain AI chat quality over time: tagging relevant files/functions, feeding execution or error logs back into AI, and avoiding overly long chat sessions by starting fresh threads for new features. These practices help remedy AI drift and improve feedback loops.