I tried coding with AI, it sucks
No-code is here to stay. As you were.
I keep reading the following phrase in various forms:
“AI code generators just killed no-code, now anyone can build an app from a prompt”
My take: Bullshit.
I spent the last few weeks testing various AI app builders (Lovable, Cursor, Replit, Bolt) and I’ve got some thoughts to share about the whole “AI will replace no-code” debate.
Don’t get me wrong, I desperately wanted these tools to work. The promise of typing a prompt and getting a fully functional app? Brilliant!
But after countless attempts and frustrations, I’ve realised something important.
The reality is… messy.
Every AI-generated app I tried to build turned into a game of “keep blindly clicking fix until the AI fixes it or gives up”.
When things went wrong (they always did), I felt completely powerless.
There’s so much content appearing of people apparently using AI to build apps in minutes, but I think they’re being highly selective about what they show.
- They’re not showing the abandoned projects when AI hits a bug it’s unable to fix (this happened to me many times).
- They’re not showing the part where they had to dive into the code to fix something AI couldn’t.

Coding with AI is great— if you’re a coder.
But for a no-coder, if I can’t understand the app I’ve built, it becomes impossible to maintain.
In fact, I felt like a product manager whose entire developer team had just resigned except for one over-confident junior dev with a compulsion for lying.
App owners need their apps to behave consistently. They need to understand how things work and know that making a change in one place won’t break something elsewhere.
I cannot see a future where I will feel confident asking an AI to update, refactor and fix code I don’t understand.
But no-code tools just generate code, what’s the difference?
The difference is no-code tools have put in place a framework which means consistent inputs get consistent outputs.
I know if I create a workflow in Bubble it will do what it’s supposed to. And if it doesn’t, it’s always my fault, and I can figure out how to fix it.
I don’t need to understand the underlying code just like people coding in React Native don’t need to understand the machine code and binary that lies beneath the code they’re writing.
If consistent inputs get consistent outputs, you can understand it and work with it. If it behaves differently every time, you can’t.
The problem with AI code is — if you feed an AI the same prompt over and over and you’ll get slightly different results every time. Vary the prompt a bit and… yikes.
But what about AI + NoCode?
I recently tried Bubble’s new AI developer feature, and although it was far from perfect, it was actually quite impressive.
It can scaffold out basic app structure and workflows in minutes, but – and this is crucial – in a way that I can then understand, modify, and maintain using visual development tools.
This is where I see the real future for no-coders: AI as a powerful assistant in the no-code space, not a replacement.
Think about it:
- AI can help us prototype faster
- AI can help us inject small code snippets where we need extra functionality
- Visual tools let us understand and control what we’re building
- We maintain ownership of our product’s destiny
For no-coders like us, this is actually fantastic news. We get the best of both worlds – AI’s speed and creativity, combined with no-code’s visibility and control.
🔑 Key Takeaway: Don’t get caught up in the AI hype. Focus on building products you can understand and maintain. Use AI as a helpful assistant, not a mysterious black box.