Prevent cognitive debt by manually retyping LLM-generated code

The author uses coding assistants on personal projects but manually retypes all LLM-generated code rather than copying it directly, preventing cognitive debt and maintaining deep understanding of the codebase. While this approach is slower than fully automating with AI (roughly 2x faster rather than 10x), it forces deliberate engagement with the code that helps catch errors and build mental models of how the solution works. The strategy balances productivity gains from AI assistance with the satisfaction and learning that comes from understanding every line of code in personal projects.