My First steps with Claude AI
Just before Christmas, I hit a wall. I was deep in a client project, wrestling with a technical bug that refused to budge. For years, my routine was simple: tap a developer on the shoulder, explain the mess, and ask for a hand. But this time, the response was different: “Fire up Claude and have it fix the problem for you.”
I’ll be honest—I was irritated. It felt dismissive, like being told to go ask a calculator to write a poem. I’d played with LLMs in a browser before, but I’d never let an AI touch my source code directly. Reluctantly, I grabbed an API key, plugged it into my terminal, and described the bug in plain English. Five minutes later, I was submitting a pull request. I wasn’t just unstuck; I was already dreaming up the next feature.
Breaking the Implementation Barrier
I love being a designer, but I’ve always had a love-hate relationship with the gritty side of implementation. I’m essentially a junior developer: I can write code well enough to get what I want working but it’s clunky and needs revising. Once that client project wrapped, I decided to run an experiment: Could I build an entire app by using AI to bypass the parts of the job I usually struggle with?
I started by telling Claude to spin up a blank React app with Tailwind CSS. Usually, that’s an hour of configuration, dependency conflicts, and npm headaches. This time? Done in seconds. The real test was getting what I had envisioned in my head into the browser. Without wireframes or a formal spec I gave Claude a single paragraph describing the features and layout I had in my head. I expected a cycle of endless corrections. Instead, after a few moments of processing, it spat out a functional UI that was almost exactly what I envisioned. A few months ago that would have taken days of wrestling with component props, broken click events, and database socket errors. However now, within an hour I had a working prototype I would’ve felt confident demoing to stakeholders.
The Magic Wand vs. The Reality Check
It felt like I’d been handed a magic wand. The code-based headaches had vanished, leaving me free to focus on what I actually enjoy: crafting a great user experience and building a beautiful interface. However, this superpower comes with an asterisk.
I have enough experience to know that "functioning" is not the same as "production-ready." My AI-built app lacked tests, deployment pipelines, and robust security protocols. If I told Claude to handle security, I wouldn't actually know if it did it correctly. Is user data encrypted? Are there API keys leaking in the repo? Is there a silent memory leak waiting to crash the server? An experienced developer understands these fundamentals; a hobbyist with a prompt might not even know they exist.
The Rise of the Creative Director
This leads to a looming concern: the "Intern with a Prompt" scenario. Marketing teams love showing commercials where anyone can build software but what does that codebase look like in six months? If a mission-critical bug hits flight control software or a city’s traffic grid, and the person coding it doesn't understand the underlying logic, we’re in trouble. AI is a co-pilot, not an autopilot. We’re already seeing stories of AI deleting entire databases or hallucinating libraries that don't exist. Without a fundamental understanding of software architecture, we aren't just building apps - we’re walking into a minefield.
That said, my workflow has fundamentally shifted. I don’t feel like a Designer anymore—I feel like a Creative Director with superpowers. I have a junior developer at my command right in my terminal. I’m no longer limited by what I can manually code; I’m only limited by the clarity of my vision.
What the Future May Bring
AI adoption is moving at an extraordinary pace. I don’t believe anyone understands where the industry will be in 5 years let alone 5 months. I don’t believe it will kill design jobs. Instead, it will let us focus on the parts that machines cannot and will not be able to do - finding the correct problems to solve, determining what priorities we should be focused on, being creative and delivering a unique experience that isn't the same as 1000s of other AI generated sides, and most importantly being empathetic to our users needs
The "Magic Wand" is here. Now, we just have to make sure we're building the right things and solving the right problems, and importantly we need to know the magic works before we start casting spells on production servers.