Building with AI
Building with AI
Engineering is evolving — and vibe coding is here.
Vibe coding is the process of building software through natural language and iterative feedback. With AI as a creative and technical partner, we can now write code, design interfaces, and shape applications in entirely new ways. It’s transforming not just what we build, but how we build — and who gets to build.
In this series of experiments, I’ve been exploring how AI can accelerate product development: turning high-level concepts into prototypes, speeding up front-end builds, and enabling faster, tighter feedback loops. What once took expert teams months can now be done solo, in hours — for hundreds, not thousands.
Simple games
One of the quickest ways to get started with vibe coding is by building simple, playful games. Games small in scope, fast to prototype, and low-stakes — making them ideal for exploring tools, testing ideas, and getting comfortable with the process. This section showcases a few of the lightweight games I’ve been building lately.
Honey
A Wordle-inspired, dog breed guessing game.
Wordle Poetry Club
Turn your Wordle guess words into poems.
More complex builds
Once you get more comfortable with vibe coding, it’s exciting to push into more complex builds — exploring deeper functionality, more refined interfaces, and smarter AI-powered features. This section showcases some of my latest experiments, where I’ve been pushing the limits of what I can build — and what becomes possible when AI is your new product partner.
Honey
An AI-powered dog breed analyser
Alexandria
A bookshelf scanning and sharing app.
Mixtape Mate
An chat-based playlist maker.
What’s going on
An AI-powered news summariser
🧰 The Tools I’ve Been Using
To build fast and with minimal cost, I’ve been relying on a lean but powerful stack:
From name to features, target users to positioning — I use ChatGPT to get ideas out of my head and into shape fast. It helps me clarify the concept, test different angles, and quickly move from abstract to actionable.
🧠 Step 1 — Flesh out the concept
Tool: ChatGPT
Once the concept is defined, I get ChatGPT to generate high-fidelity UI screens — using the context we’ve built together in the chat. It usually takes a few iterations, but the results are surprisingly solid.
🎨 Step 2 — Rough designs
Tool: ChatGPT (again)
Lovable is where the concept really comes to life. I provide a short summary of the product and a few ChatGPT-generated screens, and we take it from there — turning the concept into a working product, fast.
🔨 Step 3 — Build the thing
Tool: Lovable