The AI agent started executing actions even though the “Planning” mode was enabled. Instead of only creating a plan, it began implementing the steps automatically.
This happened after a plan was already created and I was trying to further refine it. The agent then started execution on its own.
This has now happened to me twice. I’ve used “Gemini Pro” the 2nd time (not sure about which one I used the 1st time).
“Planning” is a misnomer.: it really means it thinks harder about its responses. The agent is designed to take action, so you’ll have to get in the habit of telling it not to (“reply only, make no changes”), or setup a rule to change its behavior.
Antigravity with Plan mode using Google Models, most of the it didn’t work for me. I read this is pretty common with Google Models, that is the reason because I am only using Claude Models.