A designer's first week with Claude Code
A calm, step by step start for designers new to AI coding.

Zander Whitehurst
10 min read
Opening a terminal for the first time can feel like walking into the wrong room. Black screen, blinking cursor, no buttons. For many designers that moment is where the curiosity about AI coding ends.
I still remember my first time in a terminal. I typed something wrong, a wall of red text appeared and I closed the window like I had accidentally walked into someone else's house. If that sounds familiar, you are in good company. Almost every designer I have taught has a story like that, and almost all of them are now happily building in code.
It does not have to. Claude Code is one of the most approachable ways for a designer to start building real interfaces, and the first week is mostly about getting comfortable rather than learning to code. Here is a calm, day by day way to start.
What Claude Code is
Claude Code is an AI agent that works inside a project on your computer. You describe what you want in plain language, and it reads your files, writes code, runs commands and shows you what it changed.
The key difference from a chat window is that it works directly in your project. It can see your components, your styles and your structure, which means its suggestions fit what already exists instead of starting from nothing.
Day one: get set up and explore
Your only goal on the first day is to get Claude Code running and ask it questions. Do not build anything yet.
Open a project, either one your team already has or a simple starter, and ask things like:
- What does this project do?
- Where are the main components?
- How is styling handled here?
You will learn a surprising amount about how a codebase is organised just by asking. Treat it like a patient guide who never minds a basic question.
Day two: make one tiny change
Pick the smallest possible visual change. Change a button label. Adjust the spacing on a card. Swap a colour for a different token.
Ask Claude Code to make the change, then look at what it did. Most agents will show you exactly which lines changed. Read them, even if you do not understand everything. Over time those small reads build real understanding.
Then open the page in your browser and check the result. Seeing your change appear on screen is the moment it starts to feel real.
Day three: build a small component
Now try something slightly bigger. A simple card, a badge or a small list item. Describe it clearly, including what it should look like and which existing styles it should use.
A helpful prompt structure is:
- What the component is for
- How it should look, using your existing tokens and components
- Any states it needs, such as hover or disabled
If the first result is not right, do not start over. Ask for specific changes. "Make the padding match the other cards" works far better than "make it look better".
Day four: learn to review
Today, focus on reviewing rather than building. Ask Claude Code to explain the component it built yesterday. Ask why it made certain choices. Ask what would happen if you changed a particular value.
This is where designers often gain the most confidence. You start to see patterns: how components receive information, how styles are applied and how pieces fit together. You do not need to memorise any of it. Recognising it is enough for now.
Day five: prototype something real
End the week by prototyping a small piece of your actual product. A settings section, an empty state or a new card layout. Use what you have learned: clear prompts, small steps and careful review.
Share it with a colleague. Watch how they interact with it. Notice how different the feedback is compared with a static mockup.
Habits that help
A few habits make the whole process smoother:
- Work in small steps. One change at a time is easier to review and easier to undo.
- Commit often. Saving checkpoints with Git means you can always go back.
- Point at existing patterns. "Make it like the pricing card" gives the agent a clear reference.
- Stay in charge of taste. The agent is fast. You decide what is good.
What to do when things go wrong
Things will break. That is normal and fine. When something goes wrong:
- Describe exactly what you see, including any error messages
- Ask the agent to explain the problem before fixing it
- If you get lost, go back to your last working checkpoint and try a smaller step
Every error you work through teaches you something. Within a few weeks most of the early confusion fades.
One thing nobody tells you: breaking things is part of the fun. I once deleted an entire page by asking for "a cleaner layout". Git saved me, I learned what Git was for and I have never forgotten to commit since. Every disaster is a lesson in disguise, usually a fairly cheap one.
After the first week
By the end of a week you will not be an engineer, and that is not the goal. You will be comfortable opening a project, making changes, building small components and reviewing what an agent produces.
That is enough to start prototyping in code, collaborating more closely with engineers and bringing working ideas into reviews. The rest comes with practice, one small change at a time.
Memorisely