Why You Need to Start Prototyping with Real Code

Why codebase access might be the most underused tool in a product designer’s workflow

Most designers are using AI in the dark.

You open a chat window, paste in a screenshot, describe your problem, and hope the model knows enough about your product to give something useful back. Sometimes it does.

More often, you get something that sounds right but has nothing to do with how your product actually works.

The problem isn’t the AI. It’s the context.

About four months ago, I asked my engineering team if I could get a copy of our codebase.

My goal wasn’t to write code; it was to understand the product properly, beyond Jira tickets and design files that may or may not reflect what actually shipped. Setting it up required some developer help, but getting it running locally was genuinely thrilling. Seeing the live product on my machine, knowing I could push it into any state without breaking things, was a massive shift.

Here is what actually changed in my workflow:

Your prototypes are ignoring reality

Every designer is using AI to prototype now. The question is, what are you starting from?

Most prototypes start from nothing:

• Generic components.
• Invented rules about how the product…