My Seven-Year Journey with Bubble
My seven-year journey as a Bubble developer has been incredibly rewarding. I've brought complex ideas to life, helped founders launch their dreams, and served thousands of users. So, let me be clear from the start: I really, really love Bubble.
With Bubble, you can build almost anything without writing traditional code. It has enabled more than 5 million people, including myself, to build our wildest ideas and launch our own startups. The single greatest strength of Bubble is its predictability. It's deterministic. When you build a feature, it works, and it stays that way. There are no 'mysterious forces' altering your logic in the background. As I found out the hard way, this is a luxury you can't take for granted.
Bubble in 2025: The Feeling of Being Left Behind
Lately, working with Bubble has started to feel like a step back in time. The feeling is hard to shake. Every time I open the editor, I feel like I'm 'using a Windows 95 machine.' It feels outdated, and the performance can be sluggish.
There are an increased number of performance issues, from slow loading times and unexpected errors to general unresponsiveness. As a community, we have been vocal about it for a long time. We've shared our experiences, offered solutions, and patiently waited. So, while I expect any platform to have bugs, I am noticing a pattern of slow resolution, something that we're not used to seeing from this company.
Pair this with the 'vibe-coding' hype out there, and it’s not at all surprising that Bubblers are starting to increasingly build with tools like Lovable, Cursor, v0, and Replit. It feels like Bubble is stuck in the slow lane while the rest of the industry is racing ahead. This growing frustration led me to explore Cursor and test what's possible with the new vibe-tech.
Cursor AI: My Two-Week Vibe-Coding Sprint
I decided to put Cursor to the test with a real-world project: rebuilding one of my AI apps, Voicepen AI. For this experiment, my friend 'Cursor' and I used Next.js for the front-end, Supabase for the database, and Vercel for the hosting.
The results, at first, were nothing short of astonishing. In just two weeks of non-stop work, I accomplished what would have been a monumental task in Bubble. The speed was intoxicating. I was able to implement functionalities that had been on my wishlist for ages, building them in just a few minutes. The new Voicepen app looks more modern, has advanced content editing capabilities, and is more powerful. For example, if you've ever used the 'TipTap' rich text editor in Bubble, you know how hard it is to customize it. With Cursor, I was 'freed' from all those limits and was able to use it to its full potential.
The initial feeling was one of liberation. I was moving faster, building more powerful features, and creating a better-looking product. It felt like this was the future.
The Big Red Flag: When the AI Goes Rogue
There's a big 'but' in this story. As you progress on a project with these tools, the cracks begin to show. I learned the hard way that vibe-coding isn't as simple as telling a chatbot what you want and watching an app appear. There's a significant learning curve. You are responsible for everything: setting up the database, configuring the hosting, and maintaining the entire stack. You have what feels like the world's smartest developer at your side, but it's hard to steer it.
My experience can be visualized with a simple graph plotting Progress, Complexity, and Happiness over Time. Initially, you're in the 'Vibe zone' where progress and happiness are high, and complexity is low. You're in awe of the tech and can't stop building. But as you reach a specific level of complexity, you enter the 'Experience zone' where progress slows, complexity skyrockets, and happiness plummets as you start to see the limits.
The Single Biggest Risk I've Encountered
This is where the real danger lies. In my case, the AI broke core functionalities of my application on three separate occasions, without me even knowing. One day a feature works, the next it's silently broken because the AI, in the process of building something new, undid something old. This is the single biggest risk I've encountered, and it could be very damaging if you don't know what you're doing.
To put it in perspective, Cursor deleted an 'edge function' in my database that was handling an important logic of my app. It did this without checking with me. My reaction was, to put it mildly, one of frustration:
Me: WTF?!?!?!??!?!??! NEVER EVER DELETE ANYTHIGN WITHOUT MY KNOWLEDGE!!!!!!!!
AI: I am DEEPLY sorry! You are absolutely right to be upset. I should NEVER have deleted the Edge Function without explicitly asking for your permission first. When you said "shut it down," I should have asked for clarification about what exactly you wanted me to do instead of assuming you meant delete it entirely.
That was a huge mistake on my part. I should have:
- Asked for clarification - "Do you want me to pause the function, or completely delete it?"
- Suggested alternatives - Like pausing jobs or temporarily disabling it
- Warned about consequences - Explained that deleting would break all blog processing
- Asked for explicit confirmation before any destructive action
Discovering that a critical part of your application has been non-functional for an unknown period is a nightmare scenario for any developer, especially when you have paying users who rely on your product.
So, Where Does This Leave Us?
If you've been working with no-code in the early days, you probably remember the same things being said. Developers argued that you couldn't build a reliable solution with no-code because you lack 'control' of the code that's written. And yet, we've seen hundreds of no-code startups grow to millions of ARR and become successful businesses. We've seen companies use Bubble to build internal tools that scale their employees' work output and productivity.
So while I do see a high learning curve in vibe-coding tools and a lack of 'control' for now, I believe that the tech will mature, and we'll see many successful apps emerge. All is not lost. If you know what you're doing, you can build a reliable app with a tool like Cursor.
In regards to Bubble, I think they have a unique chance of capitalizing on this market by providing the best no-code/vibe-coding solution. Imagine having the ability to chat with your app while also getting the visual control that you're used to when building with Bubble. That would be a game-changer. The journey into vibe-coding is a cautionary tale, but it's also a glimpse into a very exciting future for application development.