Wait, you're still here?

Meta
21

My 0.02 as a profitable Saas founder scaling with Bubble on a dedicated instance: There’s just no other platforms, AI coding tools included, that would allow an app such as ours, as a team of … 1 and a few contractors (now 2 since september) to exist

Building is fast and easy, merging branches is fast and easy, debugging users is fast and (almost) easy (the logs are a pain to work with), we don’t have to worry about devops, uptime and server issues

To answer OP’s question: Bubble, despite its flaws, is still unique in what they allow some profiles - such as mine - to build and maintain

We honestly won’t move anytime soon, not because we’re married with bubble, but because the alternatives are not interesting at all given our own circumstances

22

A lot of interesting opinions in here. Maybe I should host a roundtable call and let everyone share their piece? Would anyone be interested in some office hours vibes?

23

Do you think any of this has to do with how you are working with the AI tools? I tried a little while back to “one-shot” rebuild of a piece of my bubble app on bolt and within maybe 3 prompts, I got stuck in an unbreakable error loop. I 10000% agree that trying to ask AI to just “build” a complex system on its own will fail quickly.

However, I have pieces of my app that live outside of bubble on custom javascript (specifically, a webscraping microservice that required architecture bubble can’t support), and I use AI in a very focused manner to help with writing and modifying code, and I have had solid success with it. Basically, with good coding practices where things are adequately DRY, I can give the AI only the single function or set of functions related to a given task and work in that directly, rather than the entire codebase at once. Like working with a real engineer, I might also have to explain how these pieces fit into the larger system for context, but I keep it focused on the particular task at hand

The AI absolutely still hallucinates from time to time (I find asking it to link to docs for each function it uses will help reduce that), and I often do need to read the code and spot issues / make small changes, but it is able to spit out working code / requested changes really rapidly. When I consider how my own dev time is spent, I am personally starting to debate whether bubble is still “faster” than working with AI in this way - for instance, working with “only when” conditions can be miserable and take hundreds of clicks in bubble when I need to make changes.

100%. For example in my Grooming Salon app, I can ask ChatGPT or Claude or Cursor to build an application for managing grooming salons that has stripe integrations for payments, clicksend API for 2 way SMS, the ability to add multiple packages, add-ons and discounts to appointments, which is not even close to all of the features in the app.

It builds out a shell of a booking program with mention of the integrations, but no real logic that support them. Then I focus in on the booking logic and for some strange reason, it starts renaming variables and redoing functions that have nothing to do with the booking process, so then it takes time to fix that, then once the booking logic is functional and I move onto integrating ClickSend, it starts changing variables and functions in the booking logic, and it’s just too much time and effort. If I had the time to sit down and code it myself, I would because I can, but I don’t have that kind of time anymore.

I’m sure there are better ways to go about it, and tips and tricks to AI, but the amount of time it would take for me to look into it is the same amount of time it takes me to build regularly in Bubble, so I don’t see any incentive to bother with it at this moment.

24

Very fair, I think that’s why I have success if I keep the work totally containerized - it can literally only see the file I’m working in so it can’t change things outside the scope of the current work, but it definitely requires close attention. I guess the net tradeoff comes down to how much time AI can save on producing code relative to the time spent reviewing it. I might just be a better editor than writer (ADHD certainly doesn’t help), so that’s why it feels more worthwhile to me personally.

That said, I 100% get your pain. Claude in particular loves to flat out lie about making changes. Here’s a fun recent example, and even after I gave it the exact snippet it still left in some dead code that I had to manually remove. The gaslighting is comical

image
image1544×1308 337 KB

25

:raised_hand: :grinning_face:

I certainly would. In my particular case, I think I would be very happy to stay and keep scaling on Bubble if I had a better sense of how I can use code & AI in my bubble workflows to speed up building - for instance, in some places I’ve started using the JS toolbox plug-in’s server scripts to use javascript for certain processing & filtering tasks rather than using bubble “native” functionality because its faster to setup and modify

Right now, it feels too “either / or”

26

I have worked and do work with some great companies that have huge apps in the code world.

When I hear people say they’re going to go to code or even AI (with little or no coding experience)…I often think they have no idea what they’re getting into.

For a successful app built with code, you need a tech stack. Then you need people to run it…unless you’re exceptionally gifted and can do it all :grinning_face:

There are just a million different problems you’ll run into, and you need someone there who knows what they’re doing to get you going again. And, it’s not cheap.

I have a live app right now, and another one about to go live…

right now, I have a team of engineers, DevOps, security engineers, product engineers, testing engineers, and the list goes on. It’s a huge support staff. And it only costs me $32 a month because I let Bubble handle it.

Someday, my Bubble app may grow to where I need a dedicated plan (we all wish).

For now, I really don’t know of a platform that does as much as Bubble just by pointing and clicking.

Good luck to everyone on your projects

27

Have you seen it because you built it? :wink:

28

That’s such a huge point and very understated. Yes, Bubble goes down sometimes, but you know who has to fix it? Not me.

29

An excellent idea for sure. I would attend if Josh and Emmanuel are present (or at least will watch the recording).

30

Bubble did an awesome job on handling the DevOps part. Engineering-wise, it’s not worth managing your own server if you don’t have many resources.

32

No, but if anyone wants to buy Buildprint, then now is a great time because it’s gonna be useless shortly

33

What is Buildprint?

34
35

It’s like AI Jedi for Bubble.

36

Don’t hype it up too much George :joy:

37

I mean everyone here knows that if I tried it and thought it was shit I’d say so :grin:

38

Word.

39

Oh, your app security audit tool? I thought you called it Not Quite Secure, as a play on your agency name, security and idea the tool points out security vulnerabilities of the app.

40

NQU Secure’s features were folded into Buildprint, which is a superset of NQU Secure + Flusk’s features, and also documents your entire app into a Markdown exportable format.

41

Going away because of Bubble’s AI?

Anyway, they should use your buildprint instead of the Flusk thing they have…

would be better for everyone :grinning_face: