I have been able to work through a full project without encountering the critical issues reported earlier. The restore functionality is operating correctly, and the workflow is stable once again.
It is worth noting that the response time between requests has increased slightly compared to before, but it remains within an acceptable and functional range. Nothing that should block your progress.
For those who followed the advice to wait — it seems the right moment to cautiously return to the platform. As always, testing in a non-critical environment first remains a good practice before resuming work on important projects.
Further update on Google AI Studio Build: new functionalities have been added in this latest version.
One particularly welcome improvement is the handling of secrets — the platform now requests them manually, which is a solid step forward in terms of control and security management.
That said, there are still a couple of notable absences. The conversation reset button has not been restored yet, which was a practical and frequently used feature. Additionally, I was unable to confirm whether the option to download the project as a ZIP file is available, so that remains an open question for now.
Overall, the platform is moving in the right direction. The core functionality works correctly, and the new additions are meaningful improvements. We will continue monitoring for further updates.
I have not had the chance to test that model yet. I will look into it and share my thoughts once I do.
So far, my experience has been with the Flash models — specifically Flash 3 and Flash 2.5. Both have performed reliably and consistently. What stands out the most is the response time: even for heavier requests, the wait time remains remarkably short, which makes a real difference when working on demanding projects.
That said, it is worth keeping in mind that this is a newly released model. As with any launch, early issues and instabilities are not uncommon. The sensible approach is to give it a few days to settle before integrating it into any serious workflow. Patience at this stage can save a lot of headaches down the line.
How are you able to work on a project without the chat reset button? What happens when your context window overloads? “back and working“ is an exaggeration when we don’t have such core functionality yet.
It looks like all models except 3.1 are working again now. I’ve been building a simple puzzle game for the last few hours using all models. 3.1 works sometimes but is very slow. Can’t do much else since the chat refresh still hasn’t been fixed yet. I hope they get it completely fixed soon.
My issue is that when I use it to build a app, no matter what model I use, it generates the code files properly. But when I try to preview the app, I get this page in the preview panel. All the code files which were generated now just say “Permission denied”. This is the same issue I’ve been facing for more than 48 hours with no exceptions. It never previewed the app properly. Any edits made to previous apps would run into the same issue, rendering all of them useless.
That is a fair point, and I understand the concern. However, speaking strictly from my own experience, throughout the work I did on my project I never once needed the reset button. The context window did not become an issue for the type of tasks I was handling — and I use AI Studio Build to create entire SaaS applications, so it is not a light or simple workload.
I acknowledge that depending on the nature and complexity of your workflow, it can absolutely become a limiting factor. Every use case is different. My statement that the platform is back and working was based on my personal experience, and for what I needed, it held up well. The missing reset button is still a pending improvement that Google should address, but for the way I work, it was not a blocker.
If you are seeing a 403 error in the preview, it is likely that the application requires your Firebase credentials to initialize. Make sure to provide the necessary keys when prompted — without them, the app will not start and the preview will have nothing to load.
I thought that too. However even if I enter my firebase credentials it doesn’t work. I even tried to make another app with the prompt “Make a app to take notes” and the same issue. As you can see below, all code files just say “Permission denied”
hi @aroshidze my trick is to work feature by feature. Every time I finish one, I just “remix” the app. This completely resets the context to zero, gives me a decent versioning system, and prevents unwanted side effects.
