Is gemini 3.5 flash High bad or is it my code?

Basically since the update, I am getting stuck on bugs that is not fixed after several hours. Basic and main feature that I had in my app is not functioning anymore and it is adding more and more bugs basically. With debugs command, try/catch and what not. This was not an issue prior when I use 3.1 flash or pro. Don’t know if my code have gone bad or what it is but I can not fix anything or go forward right now. Pulled back 30 git hub versions. Couldn’t fix stuff there.

Like I did a huge surgical change in my code prior to 3.5. Literally changed a whole map library and it one shotted the change with 3.1 gemini flash. We talking about like 1000 lines of code and changing map library where it is main feature. And now it can not even solve simple bugs! Something must be going on.

Vibe coding and AI is the most frustrating thing ever created. It is like cooperating with a person who switches between all illnesses - part time dementia, part time schizophrenia and it just goes on. Then you give that person all power to fix your stuff. That will drive any person mad :smiley:

Please, explain your workflow, that’s not normal.

Correct. Explain workflow in greater detail. Provide more specifics about the project as a whole and I will use Flash 3.5 High inside Antigravity IDE to vibe code you an app that teaches you how to avoid all your problems before the end of the day Sunday as long as you get me those details within the next 117 minutes :+1:

As a non dev I basically just use Anti-gravity IDE right now and very basica, that worked before. Tell what I want to change or fix bugs. Use a very good debug prompt that worked fine before that analysis hypothesis. It says it understands it, gives me implementations plan and we proceed. He does not fix issue and I use dev tools in chrome to tell him any error messages or logs. Repeat. This worked fine and built a rather advanced app so I feel my workflow is the same or even better in some regards. Thats why I look at the model.

Basically need more understanding on how to read you code and also understand what you are doing to also ask or prompt the ai to fix such issues, I am not a full stack dev but I was in your position once where I use to try to fix issues and bugs but end up working on it for hours and sometimes days, but now I have clear understanding of my code and what I have been doing. Hope this help, Vibe coding is not frustrating its the lack of understanding framework you use and language that makes it frustrating.

My app is quiet complicated with a matching algorithm and there is no way I have time to understand how everything works. Some may have the skills, experience or patience but most casual vibe coders don’t.

We need to separate between casual vibe coder and semi-pro. For casual, vibe coding can and is frustrating when you get stuck on issues and the app degrades as you try and you basically don’t know why. I don’t even know something similar in life that acts like this.