I used AI to code a personal trainer app in one weekend. I see why big apps fear the competition.

• Business Insider's Lee Chong Ming vibe-coded a custom trainer app in a weekend.
• It replaced his fitness stack and made him question why he should pay for or use multiple apps.
• Vibe coding can make personalization easy and fast — and that should worry big apps.

As a former group fitness coach, I was trained to track my goals in numbers: calories in, hours trained, kilograms lifted. Numbers don't lie.

However, tracking them was messy. I bounced between MyFitnessPal for meals, Apple Watch for workout duration, and the Notes app for lifting logs. When I needed help with the form, I'd scroll through videos online between sets. It took the joy out of the gym.

At a vibe-coding workshop in Singapore in February, I tested a simple idea: What if all of this lived in one place?

I set out to build a personal trainer app that could generate a workout program, log my lifts, and surface exercise cues when I needed them. By the end of the weekend, I had a working version — and I saw why big apps should be worried.

Building the app was simple

I was given about 9,000 credits on the AI tool Manus — worth roughly $40 to $50 — to build an app.

Manus, which Meta acquired only for the Chinese government to block the deal earlier this week, isn't positioned as a dedicated coding tool like Cursor or Lovable. It's a general-purpose AI agent: one that can write code, but isn't built solely for it.

At the workshop, I learned how to kick things off with a simple, plain-English prompt. I told the AI what the app should do, who it was for, and the key features, like recommending exercises based on a user's goals.

Anyone with a clear idea could build something just by describing what they want — no coding experience required. The key was scope: not too broad, not too narrow.

Once I locked that in, Manus got to work. About 30 minutes later, my app appeared. I didn't touch a line of code. I just watched it build.

The result was a web app called "TrainerPro." It came with a grungy black-and-orange interface, which Manus described as an "Iron Forge" industrial brutalist aesthetic.

The app came loaded with a full exercise library of about 200 movements, complete with GIF demos, coaching cues, and filters by muscle group, equipment, or environment.

It could also generate structured eight- or 12-week training programs based on a popular coaching framework — adjusting for goals, fitness level, and starting weights, while factoring in de-load weeks and progressive overload.

There were a few early bugs, like some exercises not loading properly. But fixing them was simple. I just told the AI what was wrong, and it fixed it. By the end of the weekend, I had something usable.

The satisfaction of using something I built

When I stepped into the gym and opened my app, I felt a small jolt of pride.

I'd built this in a few hours. Now I was using it.

I generated an eight-week training plan using the app, mapping out exactly what I should lift each week, complete with cues for every exercise.

On the surface, it wasn't radically different from how I already trained: bench presses, split squats, deadlifts.

The difference was in the structure. The app told me exactly what to lift and for how long on any given day. No more flipping between apps or checking my notes mid-workout.

I could also rely on it for exercise cues, which meant I no longer had to browse YouTube or TikTok to fix my form between sets.

Of course, there were moments I ignored the intensity and dialed it down, especially after a long day.

There were limitations. I couldn't tweak the program to swap in the exercises I preferred, as I had run out of credits to build that feature. I also forgot to integrate a meal tracker, so I still had to rely on the MyFitnessPal nutrition app.

But that was the thing — it all felt fixable. I just needed more credits.

With a few more prompts, a bit more time, I could shape it into something that fit me exactly.

There were also days that I wanted to train with friends and to follow their lead instead of my screen. Sometimes, you just need to unplug.

Vibe coding is a game changer

In the fitness world, giants like Apple Fitness+, Strava, and MyFitnessPal dominate.

They're widely used by fitness enthusiasts and are no longer just tools. Platforms like Strava have become social networks in their own right, where workouts double as content and community.

Here's what I think: Apps that have evolved into full ecosystems aren't going away anytime soon. But the rest of the stack — the more functional, isolated tools — look far more vulnerable.

If a non-technical user like me can build a working, personalized trainer in a weekend, why pay for something off-the-shelf that can't adapt?

To be fair, building an app still costs time and money — but a one-time $50 to $100 could be more economical than paying monthly.

With my vibe-coded app, I also didn't need separate tools to log workouts, generate programs, or even buy training plans from Instagram coaches. It was all there, tailored to me. If something didn't fit, I could change it.

Fitness apps have long been built around scale — one product for millions of users. Vibe coding flips that. It makes hyper-personalization cheap, fast, and accessible.

It also opens up something else: distribution. If you solve a common pain point, your app can become a side hustle or a business, something you share and even sell.

After all, users often understand their own frustrations better than any product team.

There's also a more intangible pull. Using something you built — with your own design choices and logic — feels different.

In a world where everyone is trying to stand out, that kind of customization has value.

Do you have a story to share about vibe coding? Contact this reporter at cmlee@businessinsider.com.