Sounds great. Will the code that is generated be accessible? Meaning WCAG 2.2. Or would we be able to have the full ability to make changes to the generated code? Think synching with Github.
My honest opinion. As fast as some AI startups are shipping and iterating, I doubt Bubble’s ability to keep up with the pace of the AI era. Its infrastructure doesn’t seem AI-ready, and new AI-driven platforms will likely make app development far more intuitive through conversational interfaces and MCP.
1 Like
It’s been clear for years that AI was going to revolutionize software development, as you linked to in your article from 2 years ago.
Which begs the question, why did Bubble go all-in on mobile 6 months after that article and nearly a year after ChatGPT was released to the public when converting a web application to a native mobile app is something AI does almost perfectly TODAY* ?!?
Bubble’s focus on mobile has led to neglect of so many critical parts of its ecosystem (including some critical prerequisites to use AI) and we still have to deal with so many unrequested changes, outages and super slow editors.
Moreover, besides mobile, one of Bubble’s biggest focuses the last 18 months has been on WU calculations (or miscalculations), which will almost certainly be extinct within a few years even if Bubble survives.
*Converting Bubble Apps to Native Mobile Apps would be extremely simple if we could export workflows and logic, which I understand we can not (although some innovative individuals are getting awfully close to being able to extract/scrape this info from a Bubble app), but Bubble can export this information, so Bubble can build those native apps for us with the click of a few buttons and without delaying critical features and development for years…
1 Like
can you walk me through this? In my DB I can set things to private with rules and stuff? What exactly are those settings doing if you’re saying that things are still publicly accessible?
quick reminder for everyone here that we have an AMA with the founders next week:
2 Likes
- This time could have spent working on accessibility. Especially for software operating in the EU.
- Rather than designing with AI, AI could be used to build workflows that have already been designed. The design output it spits out right now is not anywhere close to being intuitive. Very bad UX & UI.
I’ve worked in Product Development for 10 years. I’ve used Lovable, Tempolabs, v0, and Replit.
I’m still using Bubble to build and test my MVP because of its ease of use, design, and development control. All the other AI tools can not provide the design control and flexibility I need to build a great experience. I’m speaking very kindly when I say your AI feature will never be able to compete unless you build it around making our lives easier with creating workflows.
Be true to yourself bubble and optimize what you do best. Don’t chase the carrot.
2 Likes
AI REQUEST: Create and Manage option sets
example: create all the values for a set based on my description + relevant attributes
3 Likes
For the Bubble team - my most critical feedback:
Bubble AI needs to significantly improve on design/UI/UX. It lags far behind other AI code builders to the point where it’s not an apples to apples comparison. The competition use Tailwind UI or other modern design systems, and have been trained on modern SAAS applications. It seems Bubble uses its own design system and/or is trained on existing Bubble apps, which are very ugly!
For apps, I’m finding the output is similar to a clunky government website that’s had a nice rebrand. For landing pages, it’s pretty embarrassing - I wouldn’t launch a website with any of the Bubble AI pages. Compared to the outputs on vO here - which are beautiful
Competition has raised the quality bar and the devil is in the detail. At a minimum Bubble AI needs to produce build modern, clean, responsive websites on par with v0/Loveable/Bolt. Right now Bubble AI gets me to <50% on design, whereas the AI builders are comfortably >80%.
It’s irrelevant if Bubble is miles ahead with editing features if the output is an ugly Bubble template that I have to spend hours and hours editing. Frontend design is easily where most of my hours are lost on complex apps - and right now AI isn’t saving me much time on this.
3 Likes
Completely agree.
So many of the core Bubble elements are either buggy (Table) or outdated (Repeating Group, Group Focus, Input, Popup, Date/time picker). And basic app functionality from Bubble’s plugins like Drag/drop and Keyboards shortcuts don’t work properly.
Which makes me wonder… is the AI going to use the current native Bubble elements? Or can it whip up its own custom elements on the fly? Or is Bubble going to bring their current elements into the 2020s? If using the current elements – which I fear is the case – wouldn’t the AI then just be helping me build apps with an outdated look and feel?
I agree with Josh that the real power of AI is in the “day 2” use cases of iteration, adding new features, finding bugs, etc. but the last year has left me with a feeling that Bubble has not yet fixed the foundation on which they’re building these features upon.
I was curious about the AI website building so I gave it a try with a prompt about creating & managing groups. Here is what I encountered. (1) None of the core design elements were what I wanted (i.e., sizes, colors, fonts, capitalization). There needs to be some styling questions to feed into the AI. (2) Nothing was connected properly - almost all of the buttons had empty or incorrect workflows, even though it seemed to guess the correct name for the buttons, like “Add Member”. (3) None of the classes were quite correct, and it didn’t use any option sets → even though it looked like it would have given the visual content. It also used the wrong classes for certain sections of the website. (I.e., a “member” should not be a “user” in this case, since a “user” requires an email and password - which aren’t necessarily required for a “member”).
In summary, the AI did a decent job of giving me an example visual representation of a possible website - but followed no “best practices” and would basically require me to rebuild the entire website from scratch. So, I found it marginally more useful that just finding a similar website on the internet as a reference.
There are many design and flow decisions that the AI would really have to ask the user about to build properly - just like any website development team would ask you for preferences and context. Guided AI is the way to go here, with the AI solving smaller problems.
Bubble should focus on stability instead of AI. The best apps can’t run on servers that are down every week.
1 Like
Will bubble.io provide a way to create MCP servers similar to n8n and xano?
Unclear but I wouldn’t expect anything in the MCP front in the near term
Any updates on the AI Iteration function?
1 Like