Who do i complain to about this. I pay monthly for access to AntiGravity and now - since the new implementation - i can’t get any decent output from it. This is truly unacceptable, a waste of money and thoroughly disheartening. Just 3 months ago you had a good program that was responsive, hugely capable and reliably able to create outputs from very complex instructions. I restarted Antigravity today only to discover you’ve gutted it and turned it into a dysfunctional ‘lite’ version of its former self - where it cannot even create a quarter of what it was capable of 3 months ago without issue after issue and poor amateur-looking outputs. I cannot now rely on AntiGravity to help me develop complex projects. Gemini 3.1 Pro seems … than ever and Claude is limited, so it’s not possible - without paying a lot of money - to build anything of worth. The number of errors has increased 10 fold, even simple requests cannot be carried out without breaking. The graphical output is also truly poorer than before. You’ve taken the best automated ai creation platform on the market and ruined it completely! I’m beyond anger and disappointment here. Bring back the program that actually worked and was worth investing time in. I know revenue in ai is an issue but if you can’t serve basic customer needs, wants and requirements, you’ll soon have no revenue at all.
Hey! Lots of general comments here to unpack, would love to hear specific things we can fix. Happy to push on those things!
I also had some issues with the recent antigravity. Go to antigravity releases and reinstall a more stable version before 2, and then everything is sorted.
Hi Logan, thanks for reaching out.
- Gemini 3.1 Pro Low and High, seem incapable of coding in an efficient and logical manner. They spend an inordinate amount of time thinking and going over and over decisions, which ultimately seems to end in confusion, with the output riddled with mistakes and coding errors. I’ve witnessed both get caught in loops for 10s of minutes on end. I then have to switch to Claude, which fixes the errors and patches the mistakes fairly quickly.
- Claude usage (for me) only lasts around 1 - 2 hours. 3 months ago i’d get 2 - 3 days from Claude before it told me to await a refresh.
- Separating the code windows away from the prompt chat has blinded me to many of the mistakes that are happening. In the earlier version of Antigravity I could go in and either fix code errors, or find and prompt what might be the issue - which would then be fixed. Now i’m not sure where I find the code? Is it now in a separate app? and if so why?
- I asked AntiGravity to help me code an app that allowed me to create agents that acted as company team members. I gave it a brief with as much detail as I have given in the past when coding other apps in AntiGravity. The app’s UI was a shoddy mess. Headlines were partially hidden on the vertical, links were broken, buttons didn’t work and were visually broken. The interface was basic beyond even what we’ve become used to as a basic layout. I asked it to create a toggle button that once pressed, would activate my mic and allow me to talk with my created agents. It couldn’t create a toggle that worked - when pressed it would immediately turn off. It took 5 iterations to get the toggle to work. I had to enter my private key to connect to my paid version, so that voice activation and verbal communication would work (standard) it struggled to create the file necessary to hold the key. I requested that it create a prompt box in the app settings where I might instruct it regarding the avatar settings. It created a basic prompt box that collapsed so small, not one character could be seen. I got that fixed and typed in a prompt for it to create the image of a young female assistant type - it created a man. I fixed this, and tested the mic - it didn’t work. I couldn’t get it to recognise input from the mic (this never has been a problem and isn’t an issue from my end, as other ai agents have no problem). I typed into the prompt, to have a conversation with my new avatar - having instructed AG to create a natural conversational output - and received a male robotic voice reply that could only repeat one phrase over and over.
None of this was ever an issue in the past. It feels like the whole experience has been completely dumbed down. Even when I got super specific with prompts it would correct some issues and break other things. I became so frustrated with the output, I eventually asked it to remove all files it had created that day, and then sat down to write my complaint.
I know much of this might sound subjective to you and maybe it is, but the whole experience of using AntiGravity to build apps has - for me - become untenable, which is a crying shame, as the power it had just three months previously was astonishingly good. It helped me prototype some very complex and useful app ideas. The user experience then was astonishingly simple, efficient and reliably stable. I knew that if something wasn’t correct, I could easily prompt a fix within one or two tries. Some of my prototypes are still in development and now i’m afraid to touch them because I can no longer trust that they won’t be broken beyond repair.
Not sure if this gives you enough detail, but AntiGravity is missing its robust, comprehensive build quality. It actually feels like we’ve gone back a whole year or more, in terms of its output ability.