@mitchbaylis Lol I felt this. My software company is dealing with this issue now – We use Codex, but with high levels of supervision.. Bugs that we don’t charge for, it’s amazing! Saving lots of time that I wasn’t getting paid for. For new features.. well, they can also be built in less time.. but that means less hours.. which means.. well, less revenue. Fun little business problem to solve!
This was something I was high lighting on this forum couple of years ago, trying to help bubble developers get ahead of the inevitable and let them understand how they will need to 4X their client base to maintain same income when AI helps speed them up by 4X.
I suppose one way to alleviate that issue is to use a subscription model like @georgecollier does instead of hourly rates. I’m personally looking into the idea.
Well, any business where you charge hourly will naturally only scale with the amount of time you put in. Detaching earnings from time is the only way to grow favourably.
That’s exactly correct.
Strong
Around 25% of startups in the most recent YC batch used AI to generate 95% or more of their code, a practice referred to as “vibe coding”.
and it’s growing
I think almost everyone’s right about “vibe coding” and AI builders. It totally depends on who your clients are.
Our clients are multi-million-dollar companies. If we tried to convince them that an AI builder could guarantee security, scalability, and reliability, we’d be laughed out of the market, and they wouldn’t be wrong.
But if you’re working with small businesses, startups, or hobbyists, then yeah, AI builders and vibe coding might make sense.
The real question isn’t whether these tools are good or bad, it’s about who you’re building for.
Overall, I think, and I believe, Bubble has a ton of advantages over any AI builder. But then again, it’s from experience and looking at a ton of factors
I believe the exact opposite of your core argument is true.
What multi-million dollar business is going to willingly expose itself to massive platform risk by building on a system that fundamentally rejects enterprise standards?
-
No Export Function: You are permanently, irreversibly locked in.
-
No Code Ownership: The core application code is proprietary and completely inaccessible.
-
Very limited logic: Lack of if/else, switch, useful debugging, find and replace, confusing order execution, minimal optimization functions, archaic APIs/plugins, cumbersome SSO, barebones native mobile etc
-
Unpredictable Costs: Scaling is expensive, dictated by proprietary and opaque Workload Units (WU).
-
High Labor Overhead: The platform’s lack of true AI assistance drives up development costs significantly.
Bubble has always been a tough sell for big clients who mandate full control and ownership. Now, even smaller users are nervous; recent downtime has hammered home the truth: they are locked in. Moving off the platform demands nothing less than a complete rebuild.
Market Segmentation Reality
-
Bubble is for the Hobbyist: It succeeds because DIYers need the guardrails and necessary abstraction it provides.
-
Serious Applications require: ownership, robust scaling, top-tier security, a large and established library of bolt-on functions, complete flexibility, and full control.
The Inevitable AI Shift
The notion that AI cannot handle enterprise-grade security and scalability is obsolete. AI is already far superior at codebase and security audits, and its capabilities are accelerating- fast.
The Mag 7 is pouring billions into training these models specifically to replace their own massive tech workforces. This investment guarantees an incredibly high velocity of code quality improvement.
Just look at the progression: a year ago, “vibe coding” was laughably bad—like those terrible early AI videos. Now, it’s super realistic. Six months ago, I was solidly in the camp of calling vibe coding rubbish, but that is absolutely not the case today—it is now far superior at coding than any human.
Dismissing AI builders is a profound mistake—it’s the modern equivalent of being laughed out of the office for suggesting a business needed a website in the 1990s.
Bubble still retains a place in the future, but it is not for production-grade applications; it is now only for the hobbyist market.
Previously, the high costs associated with traditional coding and specialized development teams gave platforms like Bubble a defensible competitive moat and a unique selling proposition—but that era is definitively over. Properly leveraged vibe coding tools with correct oversight (ie a senior dev) now deliver a far superior application, in significantly less time, and at a far lower total cost than Bubble can offer.
Leaving bubble with only the hobbyists since serious/enterprise apps now can use enterprise grade frameworks at a fraction of the cost.
I’m not overly experienced with vibe coding (other than making plugins for bubble using Claude) so not sure on all the ins and outs… but unless you have your own dedicated server (and are very technical) are you still not reliant on a company or service (or multiple) that provides some kind of infrastructure?
For example, scheduling a workflow in bubble to run on the 1st of each month at 2am which sends out some emails, is a 2 minute job and super simple to set up, leveraging bubbles security, time management & dedicated infrastructure. Scheduled workflows can be managed & deleted easily all in bubble.
How do you do this in the vibe coding? Surely there has to be some kind of ‘always on’ server running? So unless you’re running that yourself in your own house then you’re reliant on other services? Which probably themselves rely on other services … in which case what is this ‘full control’ being spoken of?
This whole thread made me think of DHH’s take on rewriting (talk, related article). I wonder what the Bubble engineers would do if commissioned to rewrite Bubble from scratch, taking all their experience over the years.
I think the fact people call AI coding vibe coding tells me that there is a lack of understanding of what AI coding is and how to use AI in coding. If you work for an enterprise and you are not using AI in general and especially AI in coding, then I think the enterprise organization should and would be disappointed. Of course, You need to have some coding skills and choose the right framework, and use Ai as another dev.
I’m all for people defending bubble, it’s a great platform with backend and front end built into one platform, and almost everything can be done even if it needs multiple workarounds. But trying to say AI is bad at debugging or is only for hobby or small business use just means AI is not being used wisely.
And it’s never going to be a winning argument to defend bubble, especially as AI models and agents keep improving. Also, I agree enterprises wouldn’t want to use bubble for serious applications.
I see a lot of cognitive dissonance in many of the responses, which is fair enough and likely how I would have responded had I chosen to remain using Bubble. It seems like unless an app can be built in a single prompt, it is considered to be of no concern. Can a Bubble app be fully built, just they way you want it with a single prompt? I think not.
I was a big bubble user for 5 years. I wanted to start a new app 8 months ago and decided between another Bubble build or taking the risk and trying out AI. I went the AI route as I needed HIPAA compliance and local hosting without spending thousands on a dedicated Bubble instance. For AI codiing, I finally landing on using Claude code, which is a very powerful agentic coding system. Coding with AI has been much faster than I could achieve with Bubble. It was also easy to learn once you get past the initial uncertainties. It was a steep learning curve, but I learnt. I now have a massive fully functioning production app, serving 135 users so far. It is snappy and responsive and down time is minimal as Vercel is stable. Database is on Supabase, which can be backed up (paid for) and AI knows it well, so all security and RLS is all done for you. The app has earnt around $90K so far and is secure (also with the help of the AI agents). The Hosting and AI costs for the app comes in at around $350 a month (including my $200 a month Claude Max account, allowing pretty much unlimited Claude Code use). I am quite technical, but then I told my wife,who is non-technical about it. She started playing around building a business website for herself, but that has now also progressed into a fully fledged app, with AI and Google mail, docs etc integration, client portal, calculators etc. That made me realise you don’t need to be particulary technical to use AI effectively. Building a production app is 100% possible if you can learn. If you can use Bubble, you can learn. So don’t be fooled by misguided beliefs. Claude Code is probably the best system out and uses Sonnet 4.5, it has several tools like agents that can be set for security, API’s, design etc. and it works like a dream,tracking the changes and app design as you go. It can run automatic security and validation checks. My app is 500K lines of code, so not small at all, but it is highly complex. My previous app, which took two years to build in Bubble has nowhere the same speed, design or functionality of what AI allowed me to achieve in 8 months. My wife built more in 3 months with no tehcnical ability, than I did in Bubble in 2 years. So those saying it is not possible yet are in for a rude surprise, when they finally are forced to accept reality. I love Bubble and what it taught me, but I think its days are numbered unfortunately.
Can I see your 90k app ?
I hope you continue to grow your app and come back in 2 years to tell how it’s going and how much is your cost… The main issue I can see in mostly all replies are the lack or full comparison of both.
Here is a brief overview of a video my cofounder made: https://youtu.be/RMHF8rzom9E Signups are limited to Ip country of my choice for the moment.
On Bubble my cost would have immediate been a dedicated instance and no HIPAA compatibility due to lack of BAA (unless that has changed). I actually told my wife to use Bubble and she found it too hard, but she is screaming along with AI coding! Her costs for hosting are free, she only pays for Claude Code Max $100 plan.
Fully control of the codebase and the ability to switch providers fairly easily. With bubble you are completely locked into 1 platform. If you own your code you can move anytime you want with a bit of effort.
Bubble owns your code so you never really own anything other than a login. If bubble decide to ban you or triple prices - you just have to deal with it. If Vercel or supabase do that I can, with a little effort, change to a competitor.
You could of course have a local machine always on to self host with a local database if you really wanted full ownership - but what a headache. That’s how most enterprise companies used to do it but nowadays many have switched to cloud based so they don’t have to manage their own servers inhouse.
Bubble or Vibe Code platform, you are still locked-in one way or the other. For as long as your development team and processes are relying on any monolith (a stack can still be monolithic), you’re pretty much locked in.
You can still stack Bubble as you can (and should) in code. Vibe coded or not.
Vibe Coding
It may sound nice and dandy to say “I can export my code” or “I can switch backend stacks”. The reality is that once you’ve reached a certain scale, you’re stuck. Suddenly what looked like a luxury becomes a huge expense both financially and for business continuity. It’s been a fact of software development for decades.
Code gets complex as you scale, so do the stacks you require, so does the change management. Software engineers get paid the big bucks not to write code, but to draw on whiteboards. I mean this in a positive way.
I whole heartedly disagree with the notion of vibe coding for scale because it disregards the other complexities of software development, maintenance and continuous delivery.
All the “successful” vibe coded app founders I’ve had the displeasure of listening to never admit that after getting seeded, they spend large CapEx to hire developers to redo their vibe coded apps.
Bubble
With Bubble you know exactly what you’re leaning into for the long run: A full stack platform with one dedicated team to maintain it.
When Bubble improves or add new features, you know your apps and workflows will run (Bubble boo-boos aside). Why? Because Bubble has poured the resources to ensure that Bubble apps don’t implode.
A lot of resources and manpower goes into ensuring that a service that is in production, stays live in production.
If you perceive only value in the code, it shows that you do not recognise what software development actually entails.
If you cannot manage WU costs in a Bubble app, expect the same when you run your own code. I mean really run your own code. In your own servers. Cloud or on-prem. Otherwise you’re still just locked into an ecosystem. One that comes with greater complexity and costs.
I feel like this conversation has already hit the point of whose favorite color is red or blue.
Bubble is blue
there are interesting reasons why blue is widely liked across cultures.
Here’s the breakdown:
What science actually says
1. Color preference is not hard-wired at birth
-
Newborns don’t show a universal preference for blue.
-
Infants can’t even see blue well at first; their cones (color-detecting cells) are still developing.
-
Studies on infants show preferences mostly for high-contrast colors (e.g., black/white/red), not specifically blue.
So the idea that “everyone’s favorite color begins as blue because of biology” is not supported.
2. Blue is consistently one of the most liked colors worldwide
This is true, but the reasons are environmental + cultural, not “instinctual wiring.”
Common explanations:
-
Environmental associations: blue = clear sky, clean water → safety, stability
-
Low-arousal color: calming, non-threatening
-
Cross-cultural positive symbolism: trust, peace, reliability (e.g., many corporate logos use blue for this reason)
Because these associations are so universal, blue becomes a very common favorite color.
But this is learned, not inborn.
3. Eyes and visual biology don’t predispose us specifically to blue
-
Humans have the fewest S-cones (blue-sensitive photoreceptors), so our eyes are least sensitive to blue relative to red/green.
-
From a purely visual-processing standpoint, there’s no biological basis for a blue preference.
If anything, biology would not favor blue.
4. Evolutionary psychology?
Some speculative theories propose:
-
Blue skies → good weather → safe
-
Blue water → hydration → survival
These ideas are interesting but still not considered strong scientific evidence for an innate blue preference.
Bottom line
There’s no strong evidence that humans are born preferring blue or that our brains are wired to favor it.
But blue does end up being very widely liked because of environmental cues, cultural meaning, and emotional associations.
Still caught up in “vibe” coding. Guess people forgot the horse and cart drivers protesting in the early 1900 that cars were too noisy, dangerous and unreliable.
If you invested all your time into bubble thats great, but you are not going to stop technology.
Try using a code editor, find a good framework, then add some AI into the mix to help you. Its called coding not vibe coding,
And hosting is not all that complex as some people in this forum make it out to be. You can get setup in a couple hours using digital ocean or some of the new easier hosts that are popping up everywhere. Neither is scaling. Thats why you start with a good secure, reliable framework.
If you business idea turns into a great big success then you should have plenty of cash to recode and scale as required. Enterprises do this all the time as demand and feature requirements and market changes.