A practical walkthrough on using AI coding assistants effectively for software development, based on 6+ months of experience building a C++ game engine with a team of senior engineers. Key takeaways: AI works best as a tool for skilled engineers, not a replacement. Practical tips include asking Claude directly how to use it better, setting up a CLAUDE.md file with project conventions, using Claude Code in the CLI, generating codebase reports, creating reusable skills (e.g., /cr for code review), being highly specific with prompts, and always reviewing AI-generated code. The author emphasizes that results vary based on engineering skill level and project complexity, and that blindly trusting AI output is not advisable for high-performance systems.