Why I’m (Finally) Migrating off Bubble. AMA

Mobile Beta
1

I built my MVP on Bubble last year. It took months, but it did exactly what it needed to do: it got a web app into the world. When the mobile beta dropped, I wrapped it and kept pushing, but the friction started to outweigh the speed.

In a world where Claude Code, Cursor, and Codex are moving at the speed of light, “no-code” is no longer an advantage. It’s a massive drag.

I’m nearly finished with my migration to a custom stack on Cursor, but honestly, I’m grateful for the Bubble foundation. It taught me the fundamentals of workflows, database schema, and front-end logic. That “Bubble logic” makes “vibe coding” so effortless.

My takeaway for the Bubble team: I hate that I had to migrate. I wish the platform could have iterated as fast as the AI dev ecosystem is moving right now. I’m moving off because I simply to ramp up my velocity, but I’ll always appreciate the start you gave me.

I’m deep into the final stages of the migration now—ask me anything about the transition, the database mapping, or why I decided to make the jump if you’re considering it.

Good luck to everyone still in the trenches.

P.S. if you’re a golfer, my app is called Course Vaults - check it out!

2

Company is stagnant on 4 core, AI, Mobile, Core, Web. Bubble is stagnant not on actual output but on roadmap too, like 2nd derivative type of stagnancy can’t argue. But why you necessarily need to vibe code?

3

Good luck with everything man, the app is definitely very cool.

I’d love to hear more on why you are moving to code. Sounds like the main reason is that you need to develop features for the app faster than you currently can within Bubble?

4

thanks fede.

yes, thats the main reason - speed. speed to create. speed to connect external tools & libraries with a simple prompt.

5

Bubble is low code btw. It and other no code tools are not going anywhere. AI has been in the dev space for long enough. Nothing has changed other than the development processes themselves.

Before AI it was No Code > Low Code > Code

With AI it’s still No Code > Low Code > Code. There’s just LLMs added.

The skill gaps still remains. AI is just a tool. Like any tool, how effectively it is used and the results produced depend on the user.

If you were never able to grasp software development in the first place, AI is not going to make you better. It’s only going to amplify it.

With AI, I’ve coded lots of backend services and small full stack web apps at this point. So I’ve been evaluating the pros and cons of moving out of Bubble for the last 6 months. No matter how I look at it, Bubble as my main stack is still the most cost effective. For time, labour and system complexity.

Code has always been the easiest part of software development.

All the best to you.

6

“If you were never able to grasp software development in the first place, AI is not going to make you better. It’s only going to amplify it.”

maybe. speaking from personal experience i’ve found that learning bubble is enough to be dangerous with these new tools. and i’m not convinced you need to understand code anymore either.

“It and other no code tools are not going anywhere”

I disagree. I don’t see a place for for nocode/low code in the future. I’d be willing to bet Bubble realizes this too.

7

Could be true, but you’re also overlooking a huge point right now…

you went from no code to vibing.

Anyone can take your app and build one just like it in a couple of days. So basically, you don’t have anything special that anyone else can’t sit down and quickly build. A whole lot of disposable apps are being built right now.

Everyone is excited about ‘vibe coding’. Yes, it is fun. Yes, it also has some huge problems with it that most aren’t considering.

Plus, Agentic UX is on the rise right now and will only grow.

There are ways these days to still come out ahead with an app, but that’s another long post, and it’s almost 9 at night here, and it’s been a long day :slightly_smiling_face:

Best of luck with your project

8

“So basically, you don’t have anything special that anyone else can’t sit down and quickly build”

moats can still be built with vibe coded apps.

9

You can’t vibe code a technical moat.

I, or anyone, can replicate your app in about 48 hours.

Moats, as you call them, using old terms, can only be done outside a vibe-coded app…

and there are only certain things that would be considered a moat.

My whole point wasn’t to put you down for vibe coding an app… the point I was making is that I’m always seeing people who are so excited they can vibe code an app now. I think Loveable says they now have like 10,000 apps made a day…

and the point being, yes, you can quickly build an app, but anyone else can take and rebuild your app just as quickly.

So, the point that caught my attention was that you said you thought Bubble probably knew they were becoming irrelevant, and it made me wonder if you actually realized how vibe coding was making apps irrelevant in general?

10

Which is why I like the analogy of a tool in a toolbox. Anyone can hammer a nail with a hammer. Some might break a finger or make a hole in the wall in the process.

I have been developing pretty complex systems for a government sector for the last few years and code has always been the easiest part.

11

Sorry to hear you’ll no longer be using my Mapbox plugin @zico.gaeffke, but just wanted to say the very best of luck with Course Vaults! It’s a great app.

Will be following your progress with interest.

12

Sounds like Dunning-Kruger. Let us know when you hit a wall and come back to Bubble.

13

It’s slow. Making and shipping features now is slower than doing it with AI. In our team, we are maintaining only the existing app, but all new features are shipped with AI.

U need to nail agenting development asap.

14

These are really some of the problems that have been popping up because of using AI to code…

quite a few stories to try and go into them all…

Lovable had a massive leak and tried to cover it up.

Amazon lost millions because a developer used AI-written code. Now, all the AI code has to be reviewed by a senior engineer.

However, in my mind, a lot of this talk about vibe coding for most is really a moot point, and I’m starting to wonder why I even find the conversations somewhat interesting, :upside_down_face: because only about .05% of apps succeed. Most never even get 100 paying customers.

Right now, there is app saturation.

A few years ago, 2,000 apps being introduced a month was normal… now I think the last estimate was 15,000 a month.

Where I work, we’re seeing problems with AI. They’re never really noticeable for most builders because again… a lot of apps are hobby apps, and it’s lucky if they get 1000 dedicated users.

I say if you want to vibe code, vibe code and have fun. I’ve tried them all myself.

But, if you have a serious app that is bringing in good revenue and has some serious users… unless you have a full team working for you, it’s still best to use Bubble’s team and let them do all the work for you. I mean, even if you need the dedicated plan ($5000, I think?), it’s still a lot cheaper than hiring your own in-house team.

Or, you can even hire an agency to keep things running. I mean, losing $50k in an afternoon because your app went down usually makes you think you should, at the very least, hire someone and have an SLA to fix it.

Each their own, of course.

Now, I’m off to work to fix some AI issues for a couple of customers :grinning_face:

15

Okay, here are some hard facts

  1. Vibe code is garbage, and now we are on garbage out phase, apple now made it hard for mobile app approvals, google search gets 2 core updates, 1 for spam, 1 for quality content. I don’t think we are far away from “made with AI” type of stuff on youtube and other platforms, maybe even on grocery shops now a days.

  2. Bubble has an attitude like “Its okay to use framer for landing page, then bind it to bubble app”, so generally, native mobile and AI in your case will be 70% product, not a full product and I don’t think bubble is no code it is low code. (Note: core is 70%, web is 70% IMO)

  3. Bubble is very slow on iteration

  4. Bubble doesn’t fulfill its promises. (You can check Bubblecon, and the prior roadmaps )

  5. Bubble is not beginner friendly and doesn’t care about feedback, about a year ago I was pushing for the users that coming from AI and they will churn out due to core and beginner friendliness just not here ( This is my opinion, but you can check ideaboard to see they don’t care at all)

  6. I think AI peaked at gpt5, and technology is not there, I think maybe it won’t be there, we will see when funding stops.

So, you are at tuff spot, if I were you I would choose Bubble, but I wouldn’t expect Bubble to get better soon.

16

I immediately close, delete, scroll past, ignore, anyone talking about vibe coding or pushing ai made anything. Unfortunately ALL bubble youtubers and influencers have moved to this and it is really just so unpopular. None of their view counts have increased and they abandoned their base fans.

17

This is an incredibly arrogant response.

But i’ll be candid with you in the hope that it opens your eyes and a few others reading.

I’m not suggesting people move to things like lovable or replit - what I consider rudimentary “vibe-coding” tools.

My point is that there is a whole new suite of tools that allows builders who understand the basics to move incredibly fast. I’d bet Bubble is even using a lot of these new tools to build their own set of new features.

And sure, there are learning curves. But if you can set up the basics - your building speed will ramp up substantially. Features that took me weeks on bubble now take me a few minutes.

Shake your head at it if you will, but speaking from my own experience, i’m incredibly delighted at the efficiency and output i’ve achieved so far.

best of luck in your own projects.

18

Yes, slop is fast. Let us know when you hit a wall, OK?

19

Saying that Vibe Coding or building with AI won’t be a thing in the Future is ignoring the trend in Evolution as its always been.
Let’s not be too defensive and come true about reality, AI is Good across ALL DOMAINS right now!!! (Talkless of how it’ll be in the next 2 years). People investing in it the right way and using the right resources are getting great outcomes. Systems are rapidly being automated, new workflows are being created, and more.

I stick with bubble because i believe it is scaling too, just in a silent/dispropotionate way to the flow of things online and it will be more capable to support robust achitectures. i also think ways of building with Bubble will be alot easier too (All we need is Godspeed).

If you’re sticking to bubble, My advice will be to also get an understanding of these “new” AI systems. Create shit okay! create slop, vibe code.. don’t be afraid to pickup new stuffs..
Try them out and then Merge these new skills to move alot faster with Bubble.

Hate on Zico if you want but he’s true. And the best of wishes!
the game is about making the right moves.

Staying with Bubble or going with Vibe Coding are both Wins.
Those that are losing are the ones that are giving up.

20

Totally false

There’s a lot of already scientific publications about all the problem with current AI response, in all domains. One of them said AI goes from 30% to 70% of wrong, inaccurate or inadequate response depending of the domain or the AI used.

AI sometimes help, sometimes not. You need to be careful about what you receive from AI. Validate, inspect before you use it. This apply for ALL domains.

It’s already slowing down and some new version of AI can lead to worse response than the previous version in some case. This is normal. It’s always easy to make big step at the beginning but each future step will just take more and more time.

Absolutely. This need to be a case analysis each time.