Hello, I hope the person involved in this article will see this.
For example, I entered a file called B.md , which includes a specific A project problem, into the Gemini 3 pro (high) model, “I brought a report b.md that has found problems with our a project at the moment. Please make a report based on this that you wrote implements, tasks.”
Secondly, the Gemini model creates an implement task, task, and reports it to me.
So far, that’s great.
While we were discussing the table of contents related to the task
Problem with the table of contents 2 was found and I press the review button to review the report, for example. I think paragraph 2 needs to be improved, I think it needs to be corrected
That’s where the problem comes from.
I gave my opinion by pressing the review button on the report, and I haven’t approved it by pressing the process button yet, but gemini doesn’t modify the sentence in the report and just goes straight to code modification, this is a serious problem.
In addition, Gemini only tries to modify the code associated with the task when performing the task in the report, but does not analyze the risk when performing this task by analyzing the association points.
In the case of Claude, a boundary task is created to analyze the association points when the relevant code is read and modified to implement the task and implementation task again, but Gemini does not.
When I wrote Gemini, in conclusion, the task does not show up in the checklist, but the code is already modified, which creates an abnormal situation.
Google also thinks about why many people use “claude” when using anti-gravity.
The Gemini model, which ignores procedures, instructions, is no longer reliable.
I’m using a pro-rated plan, and recently the limit has been increased to 7 days for the 5-hour limit on the Claude model.
Turns out it’s confirmed that the weekly limit has been applied, but the usage is ridiculous to be honest, I was only thinking about the 5-hour limit. What do you mean the 7-day limit?
–
For me, the Gemini model maintains a 5-hour limit policy, so far it is not subject to a weekly limit.
It makes sense that Google is encouraging the use of the gemini model for anti-gravity.
However, it ignores the above procedures and cannot even solve the relevant problems
Gemini model is no longer available…
Please. Please mitigate the risk for Google 7 day limit or make a correction for Gemini.