I paid for a fitness app for years and barely used it. So I Vibe Coded my own — and shipped it to the Android app store in three months.
All the off-the-shelf apps I tried did the same thing: log my previous weights, set a timer, and bury everything under social feeds, premium plans, and overloaded menus. For years I told myself I'd actually use one. I didn't. There had to be a better way.
Vibe Coding is the practice of building working software primarily through natural-language prompts to AI, with little or no hand-written code — coined by AI researcher Andrej Karpathy in early 2025. See the Wikipedia entry on Vibe Coding for the origin story.
So I turned my frustration into a learning project:
I wanted to experience the full cycle of AI-assisted app development firsthand. From the first idea to feature conception, user journeys to App Store rules, distribution to marketing — all of it.
The result just entered the Android app store → SetLog: Simple Gym Tracker
A simple, focused, and completely ad-free gym tracker that does exactly what I actually needed: Track sets, reps, and weights. See my progress. Remember my last values automatically.
That's it.
No social feeds. No premium plans. No overloaded menus. Just training data, clearly structured.
What I learned along the way:
→ AI is a powerful co-pilot, but you still need to steer (otherwise you will create generic AI slop)
→ The AI-assisted design process is upside down — "let the AI start and iterate on that to only keep what you need" became the guiding principle
→ Combination of different tools is king — e.g. Lovable for the base, Bolt.new for everything that isn't React code, and Copilot/Claude for the fine-tuning
The app is free and ad-free. But honestly? The real value was in the process itself.
If Vibe Coding is on your mind, my deeper analysis of Vibe Coding risks and opportunities is the right next read.
Read next:
- Workshop meeting planner – another vibe-coded pet project, same workflow
- Designers code and developers start to design – the role-silo flip side
- Event management tool, vibe-coded – a portfolio case study of the same approach
FAQ
What is Vibe Coding?
Vibe Coding is the practice of building working software primarily through natural-language prompts to AI tools, with little or no hand-written code. The term was coined by AI researcher Andrej Karpathy in early 2025 to describe a workflow where the developer's intent and iteration drive the output, not manually typed code.
Can a non-developer Vibe Code their own app?
Technically yes — anyone with a clear idea and patience can ship a small, working app this way. The hard part isn't generation; it's the judgment to recognise whether the result is actually good, secure, usable, and solves the right problem. Without that evaluation skill, Vibe Coding produces fast output that often fails in real use.
Which tools are best for Vibe Coding?
How does Vibe Coding differ from no-code tools?
No-code tools give you fixed building blocks within clear limits. Vibe Coding with AI is more open-ended: you describe what you want in natural language and get working code back, which you can then keep iterating on. The ceiling is higher, but so is the risk of producing something that looks finished and isn't.
Is Vibe Coding only good for throwaway prototypes?
No — but it depends on what "good" means. For personal tools, internal prototypes, and small focused apps like my gym tracker, Vibe Coding is genuinely useful. For production systems that handle sensitive data or have to scale, you still need engineers who can read, review, and own the code. The tool amplifies your judgment; it doesn't replace the need for it.