Building with Claude: A developer's perspective

Six months ago I started using Claude daily for development work. Not as a novelty, but as an actual part of my workflow. Here’s what stuck and what didn’t.

What works

Rubber duck debugging, but the duck talks back. Explaining a problem to Claude often helps me find the solution myself. When it doesn’t, Claude usually points me in a useful direction.

Boilerplate and repetitive code. Writing the fifth API endpoint of the day? Let Claude handle the scaffolding while you focus on the business logic.

Learning new codebases. Paste in unfamiliar code and ask what it does. Faster than reading docs, more reliable than guessing.

What doesn’t

Complex architectural decisions. Claude can discuss tradeoffs, but it doesn’t know your specific constraints. You still need to make the call.

Blindly trusting output. Code that looks right can be subtly wrong. Review everything like you’d review a junior developer’s PR.

The habit that helped most

Being specific about what I want. “Fix this bug” rarely works. “This function returns null when the array is empty, but it should return an empty array” gets results.

Good prompting is just good communication. That skill transfers everywhere.