I’m not sure what caused this but maybe this is why my credits keep draining so fast. But using Gemini 3.1 Pro (Low) in Antigravity and during any interaction with my code/tools/anything this message appears every time and repeatedly. “I’m focusing intently on tool specificity. I’ve internalized the instruction to avoid cat for file creation/appending at all costs. Instead, I’m now actively seeking alternatives to running grep within bash commands; grep_search is my new default for that task. This shift is crucial for efficiency.”
So for instance my last request took a total of about 5 minutes, half of that was just that message spamming in “Thought for 2-3 seconds at a time”. Maybe it’s doing more behind the scenes and this is the only message appearing however when it actually does something it say more or takes even longer with: “Refining Tool Choice
I’m now deeply immersed in refining tool choices, adhering closely to the directive against ls, cat, grep, and sed. I recalled both critical instructions, and I now remember listing all relevant tools before executing any commands. While seeking”
But when it’s actually editing/changing files the amount is usually 7-8 seconds+ so that feels like it’s actually doing something. However, why is it constantly thinking about tool specificity?
I do not have any sort of tool definitions or rules in my GEMINI.MD file and I noticed it started happening a few weeks ago and just hasn’t stopped.
Yes, I noticed this from the latest update as well. The last two updates have been a huge step backward in productivity. Enough to grab a Claude Pro subscription to start testing alternatives.
Honestly at this point due to maxing usage after just a few file changes on the Pro plan I need to find an alternative. I mostly code with a bit help from Gemini/AntiG so I’m not a heavy user like I’m sure most are.
I would love to go to Claude but upon signing up they immediately banned my account. When asking why they said I violated TOS which I’m not sure how I could have done due to not doing anything and I’m from Japan so not like its a banned country. Maybe something I did in AntiG triggered that but it’s a simple JavaScript app I’m coding and only used Claude a few times, 90% of the time is Gemini, no idea.
I’m now prioritizing the most useful tools available to complete the next steps. I am assessing which tools will provide the most efficient path forward. I’m focusing on their respective strengths to solve the particular requirements.
This is happening to me as well. At every step this is appended:
Prioritizing Tool Usage
I'm now prioritizing the most useful tools available to complete the next steps. I am assessing which tools will provide the most efficient path forward. I'm focusing on their respective strengths to solve the particular requirements.
It is absolutely burning through tokens. I used to never get close to the limits of the pro tokens. Now it won’t last 10 prompts.
I am also fedup. Excellent models at Google are going to waste.They are not fixing silly bugs. I get this everytime when using Gemini models (doesnt happen with Claude models) and dont know the actual agent thought chain: I’m focusing intently on tool specificity. I’ve been refining my approach to file manipulation, particularly emphasizing grep. It’s become clear: I must prioritize dedicated tools and avoid less efficient methods like cat inside bash when file creation or appending is required. The aim is to make my process more efficient and elegant.
Came here to say me too. Use to use CLI then it become beyond slow, like 10 seconds if your lucky just to read 1 file and things that use to be quick are extreamly slow.
Switched to anti gravity and its got the speed back thank god but yes it keeps spamming stupid instructions at itself.
Is google cooking so hard that are nurfing their own products to charge more, hide the good stuff? whats happening?
I’m not entirely certain, but I suspect the reason Google hides Gemini’s thinking process might be to protect their intellectual property. It seems some developers have been using these “thinking traces” to train open-source models, essentially teaching them to mimic how the more advanced closed models reason and perform.
Yeah these last couple updates have made me start losing faith in the developers. Have they gone on vacation and passed everything to their Agent Smith internal tools? If this is the result of that – I am not impressed.
Also one time Gemini 3.1 was stuck inside a terminal and was communicating to me from inside the terminal !!! This was 3 updated ago.. now the rate limits are being hit like crazy!! I’m.now sitting on 7 day cool down for Gemini 3.1 models and 3 days CD for claude models.. 6-8 atomic small coding tasks = 7 day CD ?
These same tasks previously consumed less than 3-4k tokens earlier!
Thank you for bringing this to our attention. We have escalated the issue to our internal teams for a thorough investigation.
To ensure our engineering team can investigate and resolve these issues effectively, we highly recommend filing bug reports directly through the Antigravity in-app feedback tool. You can do this by navigating to the top-right corner of the interface, clicking the Feedback icon, and selecting Report Issue.
as you can see the Submit button never becomes available. If I click the what looks to be unchecked checkbox for server logs it then complains I am not attaching server logs lol. Which still can’t hit Submit on.
I can never actually check any checkbox visually and messing with them doesn’t ever allow me to hit the Submit button. This is why I’m making a post on this discussion board because your own built-in report feature is broken as well. There is no way this is properly QA tested.
I’m just mindblown that this hasn’t been fixed yet. It has ben over a month since this problems first appeared. It affects Google’s most advanced and capable model Gemini 3.1 Pro so it is a big deal. How did they still not adress this or issued a fix for this I can not comprehend. Here is the oldest mention of this problem that I could find: https://www.reddit.com/r/GeminiAI/comments/1r9unsg/prioritizing_tool_usage_gemini_31_pro/
This reddit post is now more than 2 months old which is unbelievable.
