Ui/ux design with Claude has become much better. I have to say we are experts now so most will not able to get consistent output of high quality but for us it takes a few hours to tweak a particular design including animations and all and from there it is a minute per unique page more or less. Much faster, more consistent and of high quality than any pro on Bubble can deliver.
It still amazes us how fast AI is getting better.
I am in the process of moving away.
Bubble you are locked in. WU are super expensive. You dont own the code and you cannt move it.
Using React and Node.js you own the work and can move from host to host.
Iāve had an application on Bubble for a few years and recently switched it to a React site with MySQL db deployed to Firebase - all built with AI. Took just over a month of pretty solid AI coding in my spare time. I had never built anything in React prior.
Costs pennies to run each month, lightning fast, looks better, and itās super quick to deploy new features. And itās fun!
Sounds incredible. Can you fill in ao.e more details how you went about switching the app over? Did you have any prior experience with traditional coding (not low code or basic JS)ā¦
Iāve been using Bubble for eight years to build apps, mostly internal tools for my small business as well as side projects that I hope will scale more. Bubble is invaluable in helping our small team scale without needing extra labor. To underscore how much I love Bubble, I even traveled to NYC last year to attend BubbleCon and have hired a Bubble coach numerous times for complicated troubleshooting.
Recently, Iāve been exploring AI full-stack tools to understand where they fit alongside Bubble. Iāve followed discussions on Reddit and Bubble forums, and even took McKay Wrigleyās AI full-stack dev course called jointakeoff.com, which was great, but these AI tools are advancing so quickly that complex course material isnāt always necessary to get something working.
Hereās how I plan to pick between AI tools and Bubble:
After testing several AI dev tools, hereās my assessment:
bolt.new ā Rapidly evolving with recent native mobile app capabilities (which I havenāt tested yet). When I used it extensively a few weeks ago, I encountered numerous errors that needed constant fixing. Iām honestly shocked it isnāt better given they raised an insane amount of funding, but it still feels unpolished despite the investment. It has potential but requires a lot of annoying intervention.
Lovable ā Very similar to bolt.new. I experienced many bugs requiring constant fixes and found them nearly indistinguishable in terms of user experience. In full disclosure, I probably used it 1/10 of the time I spent on bolt.new.
Replit ā More complex than the others. While powerful, itās overly cumbersome for simple full-stack web development and produces more errors than the alternatives. Not ideal for transitioning from Bubble. Curious if someone has a use case that Replit is better for than the others.
Cursor ā Great if you have technical knowledge. I initially used it with ChatGPT 1o pro, RepoPrompt (very cool) and an XML parser (from Mckay Wrigley) but eventually hit bugs I couldnāt fix. Not an easy transition from Bubble without coding experience.
co.dev ā Currently my favorite. The AI proactively finds and fixes errors before they become issues. Supabase integration is seamless, and environment variable setup is straightforward enough for anyone even if you donāt know what they are. The pricing is comically generous - it seems steep at $50 to start, but I worked on an app for 2 weeks straight and didnāt come close to hitting the cap for the month, while I exhausted my bolt plan super fast. You get many edits, each capable of accomplishing multiple tasks, and you get credits refunded when you roll back changes. Itās a relatively new platform I found on Product Hunt, but itās been the smoothest experience by far. DM me if you want details on what I used it for and Iāll show you my editor. I am one happy user.
In full disclosure, I have not tried the new Bubble AI tools powered by Sonnet 3.7 but am eager to soon. I was not impressed by the first AI rollout though.
Eager to follow along as others keep testing the landscape.
@josh5 good breakdown. We get much better results but have to say that we have 1. Custom prompts for Claude.ai and GPT and 2. No just enough about programming to ask the right questions to Claude in cases of errors or bugs.
We are now working on streamlining it all. We are still stinking about launching a new NoCode builder. We are almost on par with the companies you mentioned while bootstrapping but need 6-12 months to get it to a state that it will also work for people not familiar with code.
Anyway, we have automated almost everything in our business now thanks to AI or tools and software written by AI that we thinkā¦why not stop with our business and go all in on the AI/nocoder space.
Just a note to say when you read āAI can build fasterā or āAI re-built my appā or āWe have wholly moved to AI for developmentā you need to know the context of what was rebuilt. If you have no sense of scale, market, viability, complexity, level of error, level of use, and many more, then itās quite meaningless handwaving.
Itās important to keep an eye on the changing landscape but itās also important to remain critical with AI, as a super skill youāll need keep refining as it improves. Same goes for influencers who may appear to be part of a community but also have their own agenda (sell you things and change your mind).
Youāre right that context is important, but that doesnāt make it meaningless. Just as it wasnāt meaningless when Bubble first released incredible demos showing how they rebuilt Twitter in a day. That was insanely cool and what I showed people when they thought I was wasting my time with Bubble. But I as a novice Bubble user still couldnāt tell the inherent errors in their approach, the difficulty with scaling (this was even before WU were introduced) or other downsides to approaching with Bubble versus typical development. But it still served to inspire and push the boundaries of what was possible.
With every new technology, you have to take a leap of faith before there are numerous proven examples at scale. Weāre in that early adoption phase with AI coding tools - where early adopters are testing boundaries and sharing experiences, but comprehensive case studies at scale are still emerging.
And unlike Bubble apps that have clear fingerprints in their code that identify them as Bubble-built, thereās no easy and consistent way to identify apps built with AI coding tools (at least that I know of). Once deployed, they might look like any other React/Next.js app - thereās no technical signature revealing AI involvement in their creation. This makes it harder to point to examples in the wild, as we simply wonāt know unless the creators explicitly tell us.
In that vein, I just started working on an external tool in the residential energy space and will share as an example if/when I finish and deploy!
Bubble has an advantage over the other AI builders for several reasons -
the biggest and most important isā¦and not to mention several others that I could go into detail aboutā¦
You can micromanage the design. AI only uses the info it collects from sources already available. This means your app is designed based on what is already out there. Bubble lets you make changes to your design the way you want. This is important if youāre building an app for profit. Yes, if you know code you can micromanage the design on other buildersā¦but, most donāt know how because theyāre not code building users.
Right nowā¦and Iāve tried them allā¦Bubble is still the best choice for building an app.
Iāve been testing all manner of AI app building tools. Today for example I gave Replit and Lovable this prompt:
### **Prompt for AI Development of "Riot Events" Web App**
**Goal:**
Create a custom conferencing web app that allows users to host live events, including scheduling, RSVP management, live streaming integrations, and user account management.
### **1. General Structure and Functionality:**
**Public Pages:**
- **Landing Page:**
- Minimalist design with black, white, and gray colors.
- Centered header: "Riot Events."
- Right corner of the header has a "Login" button that links to the login page.
- **Login Page:**
- Simple form to allow users to log in to their accounts.
- **About Page:**
- Brief description of "Riot Events" platform and its features.
**Private Pages (Logged-in Users):**
- **Dashboard Page:**
- Left-hand side menu displaying a list of scheduled events.
- Each event is displayed as a floating tile card, showing:
- Event title in large font.
- Scheduled time, including timezone and RSVP count.
- A plus button if no events are scheduled, allowing users to create a new event.
- Event card includes:
- **Edit (pencil icon)** and **Delete (x icon)** buttons for modifying or removing events.
- **Logged-in Event Owner Page (for event creators):**
- Clicking an event tile redirects to a dashboard showing a list of users who RSVP'd to their event.
- The RSVP list should include user names, email addresses, and timestamps of their RSVP.
**Admin Page:**
- Admin Dashboard should:
- Show all platform users and allow admins to delete users.
- Have a menu toggle between viewing all users and viewing all scheduled events.
- Allow admins to modify or delete any event or userās details.
### **2. Event Management Pages:**
- **Schedule New Event Page:**
- Form to collect:
- Event Title.
- Event Description (max 500 characters).
- Event Time (hour, minute, and timezone dropdown).
- Maximum RSVP count.
- YouTube Livestream URL.
- **Submit Button** to save the event and update the user dashboard with the new event.
- Upon submission:
- A unique 12-character URL slug should be generated.
- The event should be linked to the logged-in user as the event owner.
- **Event Page:**
- Publicly accessible page with:
- Event Title.
- Event Description.
- Event Time.
- RSVP Button that opens a modal.
- **RSVP Modal** collects:
- Userās name.
- Userās email address.
- Submit button to save RSVP information in the backend, associating it with the specific event.
### **3. User Account Settings:**
- **Account Settings Page:**
- Logged-in users should be able to navigate to their **Account Settings** page from their dashboard or a user profile icon.
- **Password Update:**
- Include an option for users to update their password.
- **Form to collect:**
- Current password.
- New password (with strength validation).
- Confirmation of new password.
- After submission, the password should be updated in the database securely.
### **4. Administration:**
- Create an admin user profile:
- **Email:** test@test.com.
- **Password:** 123456.
- This user should have access to an "Admin Dashboard" and be able to manage all events and users.
### **5. Backend Requirements:**
- **User Profiles:**
- Each user should have a profile that includes:
- Email.
- Password.
- Date created.
- Status (active or deleted).
- **Database:**
- Use Supabase for managing user profiles, events, and RSVPs.
- **Event Data:**
- Store all events, including their details (title, description, time, RSVP count, YouTube URL).
- Each event should be associated with the user who created it (Event Owner).
- Event data should be accessible through unique URL slugs for public access.
- **Live Streaming:**
- Use the YouTube API to embed live event streams on the event page via an iFrame.
Iāve had success with these tools for āmicroappsā that have single purposes.
Has anyone being able to build something more complex? Would love to play with some examples (if you are ok sharing of course)
I have different opinion on it - design is actually one of the easiest parts of building with code. In fact, itās even faster and simpler than using Bubble. There are tons of high-quality component libraries available, and in my experience, no one really needs more than what these libraries offer. Theyāre highly customizable, tooājust look at ShadCN-UI or headless UI libraries like Tailwind UI.
Where Bubble truly shines, though, is in how effortlessly it handles databases. Everything is deeply connected, making it easy to execute actions and events. With code, you either have to deal with complex SQL queries or rely on ORMs for a more manageable approachābut even then, Bubble still wins in terms of speed and simplicity.
Most people choose Bubble because it allows them to ship an MVP at lightning speed, and itās cost-effective to build with. As much as I love Bubble and want to argue otherwise, I canāt ignore the reality: in the AI-driven future, Bubble will inevitably lose some of its competitive edge.
Yeah, they are cumbersome if you donāt know some of coding. Iāll do it next after my 8-apps sprint.
We recently built an app with code, using bubble for authentication and database queries.
Quite powerful and also makes it easy to control privacy rules and database changes without breaking the app.
We are caching data on the front end to prevent excessive usage.
You should try breaking this prompt up into at least 10-20 different prompts. Some of these steps are quite involved so youāll never get the output you want in a single shot at this point.
Thatās cool! Do you cache on a regular interval to get new data but limit WUs?
5 minute interval, and 30 seconds for data that needs to be refreshed more often.
We also allow people to refresh manually once every 30 seconds.
Some context, we build a full fledge multi tenant system handling transactional mail, marketing mail, authentication with magic links, extensive loggin with various configurable depth levels, specialized report handling, website, landing pages, security config where you can add additional security measures such as device fingerprinting, full or part of IP and more. Also out of box every piece of code has extensive JSDoc commenting both on script level and function level such that the code documentation and manual gets auto updated
What happens if you feed this prompt to bubble AIā¦.?
Important to understand that you need to split it. People think it is easy but after 6 months of only doing production ready apps with only AI we have learnt a thing or two. You have to put in the work and you have to understand at a basic level how code works with all of these tools and AI in general. We even learn every single day and adjust very carefully our prompts.
Should we have 6-12 months of time, I am pretty sure that we can rebuild Bubble. It are amazing times!
For anyone who stumbles across this post, please consider that it would be wise not to let a few persons opinion deter you from adding a new tool to your tool belt.
With Bubble you have a screwdriver with AI (particularly bolt.new), you have a high-powered drill. We all still own and use screwdrivers, but most of us prefer a drill for all but the most basic of tasks (and by the way, both tools would be fairly useless if you were working with nails).
Anyone who isnāt using bolt.new should be and I am a huge fan of Bubble, long time user and plugin builder. As much as I love Bubble , the pricing model is wrong and the proprietary platform is problematic (Not being able to add collaborators is an unnecessary and likely harmful āpremiumā feature - #5 by bubble.trouble).
bolt.new on the other hand is open source and uses open source platforms like supabase and expo. Once we build the code on the platform we can take it anywhere we want immediately including downloading it and running it locally if needed.
It already has support for mobile app (through Expo), integrated database support (through supabase) and one-click deployments like Bubble (through netlify).
Itās not that Bubble isnāt good itās just that bolt.new (and Iām sure there will be other AI tools in the future) are 1000x better. As has already been said, you have to learn the tool to use it well and itās frustrating in the beginning, just as even though we may not remember well, Bubble also had itās own frustrations in the beginning.
Donāt take my word for it though, go see for yourself ā https://bolt.new Note: You must use Chrome for your browser with bolt or the web container will not run properly!
(P.S. You will run into errors, donāt give up. Do some research on how to fix them just like you would if you ran into an issue on Bubble. AI isnāt perfect it still needs intelligent human interactions)
