My God, the AI Studio filters are really getting more ridiculous than before. I’m working on a big project with thousands of lines of code, a biological and cellular simulator in which I’m trying to use exomes containing tumor telomeres and control the tumor and prevent it from mutating to prevent aging and use tumor exomes to increase the length of telomeres so that a kind of immortality or not aging occurs at a certain age that is chosen. But this is just a C++ simulation and I have to block all my biological words in an impossible and difficult way, or rather, out of 10 requests, 1 request is rejected and my entire quota is used in those 10 block responses and I have to move forward with my biological simulation project in an impossible and difficult way, and that’s with a lot of prayer and when they say they’re blocking it now, you’re stressed and nervous. And with this, the project will be completely behind schedule and destroyed. Even the level of concentration you need for a large project like a biological simulation project will be destroyed. You can say that you have to endure the stress of the quota that will be reduced from your 10 blocked requests. You have to endure the stress of whether your prompt will work now or not. You have to endure the stress of feeling like you are saying to yourself, “Oh my God, it will work now,” “No, it will be blocked now,” “How can I get past this?” and a thousand troubles that you could have just moved on with your project. Before and now, I will tell you that even with tokens, maybe 200 thousand or more Gemini passes, especially 3.1 Pro, are answered completely quickly and with poor quality without thinking or reasoning. You also have to regenerate so that the model can decide whether it should reason or if I want to deal with the next catastrophic problem. Well, my project is about 3500 lines of C++ code and structured files.And it’s complicated. Now, when I want to integrate several project files in an implementation phase, the implemented version of the first phase is up to 1200 lines of code, 1500 lines of code is good, and it’s in C++ format, clean and correct. But in the last code, especially if the code is a few hundred lines of code, when everyone is writing correctly and finally it’s all in C++ format, suddenly all the code jumps out of C++ format and all the code and text in the pass becomes txt and gets mixed up. I even changed my prompt or emphasized it for the model many times and tried, but it repeats every time and forces me to convert 1 implementation phase into 2 implementation phases, which wastes my time. Code that should be implemented in 1 phase is in 2 phases, one in the performance phase, one in the user interface phase and other functions of the user interface or control room. I even have to stress that I don’t want all the code to suddenly jump out of the template and become txt and I want to rewrite the prompt, rewrite the phases so that 2 phases become 1 phase and make my work more complicated and the project will progress slower. I tried to fix all the problems I had in AI Studio with the environment and model, especially the blocking filters titled: OtherBlock, which is very annoying and blocks the simulation content of the simulation prompt with an unreasonable title OtherBlock, which you don’t even understand where the prompt came from, or something else. I will try to fix these problems in the next process and updates. As someone who works on a large project and biological simulation, trying to move the project forward with these problems makes it impossible and extremely difficult, makes me nervous and stressed, and causes the project to be destroyed.And I know these problems like the back of my hand and I have experienced that I am using thousands and hundreds of thousands of tokens every day and for this reason the quota of 10 messages that could be very important in the progress of the project is reduced only because of 10 identical requests and trying to get through them with stress and nerves, and that with a quota of one out of 10 requests that is blocked. I am a very loyal and loyal fan of Gemini and AI Studio, but these problems are bothering me, especially in my big projects. Now I have tried with this report to try to make these problems better and correct in the next updates for people like me and other users. I wanted to say that loyal users like us should not lose out on a few ridiculous bugs. I hope my report is helpful and will be effective in the next updates. Thanks to the Google representative. -From Zero-
thanks for the detailed feedback, this looks like the safety filter (otherblock) reacting to biology related words even tho its just a coding project, plus a formatting issue with long c++ outputs. two tips: adjust safety settings where you can, and try breaking long code responses into smaller chunks so it doesnt drift into txt format. safety settings info here: https://ai.google.dev/gemini-api/docs/safety-settings
Thank you very much for the advice and yes, unfortunately, the block filter, even though it is a C++ and simulation project and is safe, is sensitive to words and has become difficult to move the project forward.