For the past several months, my social media feed has been flooded with people bragging about spinning up apps and websites over a weekend without any engineering help or coding — with just vibes.

Vibe coding is simply using regular language to describe your intentions to an artificial intelligence bot, which then generates the code.

I was intrigued, but I brushed it aside as just another fleeting internet moment like NFTs and metaverse farmlands. Also, trying vibe coding felt like self-mockery because I was the kid who learned coding in school, only to abandon it for writing stories about people who were building businesses by writing code. I learned just enough coding to pass the exams. I never built anything and hadn’t even thought about coding in about a decade.

As a tech reporter, watching random strangers build “products” overnight felt unfair because I have seen many entrepreneurs go through the grind of Africa’s broken infrastructure, scarce capital, and endless bureaucratic headaches to achieve even a little success.

But then a friend broke me.

Watching random strangers build “products” overnight felt unfair.

Over lunch, she casually mentioned she had built a Trello-like tool to manage a team of 10. The tool had improved her productivity as a communication and partnerships manager for a Kenyan fintech startup, she said, and that she had built it by simply chatting with a Lovable AI bot — the same way she would with a cousin over WhatsApp. The tool basically helps create a custom workflow for her team and gives visibility on tickets for the full team.

It wasn’t just the tool that surprised me — it was also the ease with which she built something that improved her day-to-day work life.

Was vibe coding for real? If she could build something so productive for her entire team, I wanted to give it a try, too.

But what could a journalist build? What problem could I solve for myself?

My initial ideas were the obvious ones: an app that tracks breaking news on my beat, or a bot that suggests unique framing around a breaking news event (don’t judge me!).

Eventually, I landed on something that would save me several hours each day: a web app that trawls the internet for daily updates on African macroeconomic data, company announcements, and digital economy stats. I wanted the tool to track the varied industries that shape African economies, but whose data lives in fragmented silos. If I could build something that makes all this information flow in a clean and faster manner, it would make my work easier and sharpen my reporting.

Lovable created a Bloomberg terminal-style interface for my web app. https://lovable.dev

What followed was chaos. Creative chaos. The kind that feels like building a pillow fort, which surprisingly becomes a real house.

My friend recommended using the tool from Lovable, a Swedish AI company that Forbes called the fastest-growing software startup ever. The first prompt I gave Lovable was “build a tool that tracks and reports the latest big tech news in the global south.” Loveable took just a few minutes to generate the images and functions it needed to build my product and then spun up a tool immediately. I couldn’t wait to start tinkering further.

Within the first few days, I realized my imagination mattered as much as my instructions. I tried to model the interface of my platform after Bloomberg’s terminal, but the results weren’t visually appealing. It was all dark and intimidating. When I asked the bot to add more visuals and make it less stiff, it gave me a cluttered explosion of visuals that looked like Times Square.

Now, I have spent more than five weeks fine-tuning my app. Every other day, I put in more prompts: refine this, shrink that, maybe … blue? And that, I believe, is the biggest disadvantage of this strong tool in the hands of novices like me. The fact that it obeys my every instruction makes me never want to stop. I am not a trained user-interface and user-experience designer, which means I don’t know what the best version is.

The Lovable AI bot responds to a prompt.

AI obeys too well, and that has become its own problem. It never says no. I wish it would, sometimes.

AI obeys too well, and that has become its own problem.

Surprisingly, though, the process has made me feel the highs and lows that African entrepreneurs — ones I have interviewed for my job over the years — must go through. When an interface breaks because the underlying structure isn’t stable, I feel the frustration of startup founders. When a feature works after 10 iterations, I feel their joy. I understand the fragility of building in ecosystems where nothing is standardized.

Will this project go anywhere? I don’t know. Maybe it’ll remain a personal tool to streamline my early reporting process. Perhaps I’ll open it up to other journalists. Or it’ll evolve into a broader platform for tracking African markets in real time.

For now, I spend several hours every weekend on Lovable and Replit, two of the most widely used vibe-coding tools, to get a better hang of using them smartly. The experiment has made me go from someone who ignored vibe coding to someone who buys books on AI. (I’m currently reading Principles of Building AI Agents by Sam Bhagwat, co-founder and CEO at Mistra.ai.)