Using AI Coding Assistants the Right Way

The Productivity Trap

AI coding assistants can make you faster, but speed without understanding is technical debt in disguise. The goal isn't to accept every suggestion — it's to use AI as a thinking partner.

What AI Is Good At

• Boilerplate generation — Config files, test scaffolding, repetitive CRUD
• Pattern completion — When you've established a pattern, AI extends it reliably
• Documentation — Generating JSDoc, README sections, inline comments
• Exploration — Quickly prototyping approaches you're considering

What AI Is Bad At

• Architecture decisions — It optimizes locally, not globally
• Security — It will happily generate SQL injection vulnerabilities
• Novel algorithms — It recombines training data, it doesn't innovate
• Context beyond the file — It doesn't understand your system's invariants

My Workflow

1. Write the function signature and a comment describing intent
2. Let AI generate the implementation
3. Read every line critically — would I write this?
4. Run the tests. If there are none, write them first next time.

The Growth Question

If you can't write the code without AI, you don't understand it well enough to review what AI writes. Use AI to go faster on things you already know. Learn new things the hard way first.