Download It and Make Something Small
Friends often ask how to begin with Claude Code. The command line makes many of them assume they need to know how to code. They do not.
My advice is simple: download it and try one small task. When you get stuck, work out why, then learn the piece you need. Explanations are hard to retain before you have touched the thing. Once there is a concrete question in your head, the material has somewhere to land.
There are many good ways into AI: YouTube, courses, bootcamps, and documentation. The useful path is the one that fits how you learn. Mine begins with making, then follows the questions that appear.
Begin With What You Already Know
Your first project should come from the area you know best, even if another direction sounds more impressive.
I had spent years working on national climate plans at WRI and BCG. When I chose my first tool, I went straight into that subject. The experience gave me more than content knowledge; it gave me a standard for quality. I could tell when an AI-generated analysis of a climate commitment was right and where it fell short.
You have the same advantage in your own field. You already know what good work looks like. Keep the first scope small enough to finish and learn from.
Let the Work Shape the Learning Path
I did not begin with a complete curriculum because I had no idea what the right order should be.
Instead, I built a small project and asked Claude what I needed to learn when a problem appeared. It suggested a course or concept. I studied that piece, returned to the project, and repeated the process. The learning path grew out of the work itself.
Each concept arrived with context and a reason to care. That made it easier to remember. AI is especially good at meeting you where you are and helping you learn the next necessary thing.
Good Ideas Rarely Arrive at My Desk
Much of my earlier work was analytical: identify the problem, run the analysis, make a recommendation. Building products made me pay closer attention to judgment.
Judgment grows out of experience, choices, and observation. It helps you notice whether a problem is worth solving and whether a polished answer still feels wrong. AI can execute, but it cannot make that call for you.
My best ideas often arrive while I am walking, resting, or fully away from work. I now make a point of spending time with things that have nothing to do with AI: art, nature, and experiences that belong only to my own life. That is where many of the ideas begin.
Share Before the Work Feels Finished
At first, I kept waiting for a version that would not embarrass me.
When I finally shared the work, messages arrived from former colleagues, friends, and people I had not spoken with in years. They told me what the projects stirred in them. Those conversations became new connections, feedback, and questions.
You cannot predict what sharing will bring. Letting real people encounter the work is usually more useful than continuing to guess alone.
Find a Few People You Trust
One person working with AI can now build a convincing first version. I ask AI to review product choices and find problems in the interface, and I often use the feedback.
I still keep a small circle of people around the work: an artist, a product manager, an AI engineer, a data scientist, and a visual designer. They do more than review the result. They bring different ways of seeing and give me people to think with.
When you build alone, a few trusted people become your team. The technical entry point is lower than it looks. The harder questions remain: what do you want to make, and who is it for?