Critical Bug Report: 100% Build Failure Rate and Poor UX Since New UI Launch in AI Studio
I’m reporting what I believe is a very severe issue.
Poor Code Iteration: To start, the Google Code Assistant’s behavior is frustrating. It rewrites an entire large file for even a one-line code change.
(With all due respect, this development experience is significantly worse than Cursor, Claude Code, and even newer tools like Qoder and Github Copilot.)
Constant Build Failures: The most critical problem with the new UI is the build success rate. Previously, it might have succeeded about 20% of the time. Now, it fails nearly 100% of the time. The builds are interrupted with various red and white exclamation mark icons, but with no explanation as to why. I’ve ruled out local issues (DNS, VPN, network instability) as all my other development tools are working perfectly. This points directly to a new problem within AI Studio.
Unusable Mobile Preview: The new preview layout is a disaster for mobile development. While it offers a large preview space for desktops, it lacks the customizable windows and UI that platforms like Firebase provide. This makes the UI/UX completely unsuitable for debugging mobile projects. When the window is narrowed to a mobile viewport, the code editor’s layout becomes distorted and awkward. I honestly have to ask: who was the genius that decided to remove friendly support for mobile previews?
such As [Forgetting to update the code]、[CheckPoint can not work] … …
I use ‘Build’ for over 8 hours a day. Believe me, I have cultivated more than enough patience to tolerate a massive number of bugs. But you have sent the model’s success rate into a nosedive, and it is driving me to the point of breakdown and depression.
Concluding Thoughts: In closing, I want to say that I still have high hopes for this product and want to see it succeed. If the build success rate were not at zero, the core cloud development experience would still be quite good. (The old Build had its own share of problems, like the persistent and extremely laggy scrolling, but it didn’t break the fundamental convenience of a cloud IDE. Many of those issues aren’t much better on Firebase anyway. In fact, if we disregard the build failures, the current cloud IDE experience here is superior to Firebase’s.)
app builder was working but lost in ide restrictions and scaffold. ai is not self-aware of environment and restrictions. iterates on code to fight itself. constant Uncaught error preact, etc.
Wanted to hop in to support your report, mate. It has been really noticeable that the quality of the model used in Google AI Studio Build is becoming worse, with half-baked implementations and the removal of full functions/flows for one piece that didn’t work. It was really enjoyable to build through that platform (despite the expected, reasonable bugs of the process). I was really delighted at how it would not just do ‘the thing’ I asked, but consider the whole implications of including it into the project and even go in ways I hadn’t initially envisioned. For example, I started the conversation asking it to implement lazy loading on a lightweight app, and after three message exchanges, without having it implemented it properly, while pointing out the faults to be corrected, it just said ‘I am completely removing all lazy loading from the application’s structure.’.
Within the past year, I’ve built dozens of apps through it and deployed a few, but with the current model, I really can’t trust it with my codebase as much. It takes lots more iterations to get one thing done, and while it’s doing that, it’s breaking other stuff.
For me, the ‘issue’ of detrimental performance was noticeable upon the update that broke the UI from three blocks into two (horrible, BTW). Now, every new input on the code mode triggers a reload of the app, whereas previously you were able to ‘run’ it on demand. I had been building an app for some days and everything was cool (old model). I had kept the AI Studio window open during those days. For certain reasons, I had to rebuild the app from scratch, so I opened a new window and landed in the new UI, and right there I started noticing something was off. It was putting a nav bar with tools, but where the previous model had integrated each one consistently and synergistically, this one just stashed them in the UI without any logic, regardless of going beyond the UI limits and without a sense of cohesion. It’s those many things that, as a weary user of the platform, let us understand ‘this has gotten worse’.
The Build features was working just fine prior to the recent updates.
Now I experience issues with the code not updating, but I am provided with responses discussing updated files. Experiencing errors from simple commands, and I have spent more time reloading old checkpoints to get back a portion of the work I’ve done than actually making progress.
The best version of build, was when the code displayed above the chat box. The Build feature began to head downhill when the “preview” and “code” tab where separate, and then with the most recent update it just shit the bed as if it was thrown off a roof and left in the rain.
