Gemini 2.5 can lie so confidently

Gemini makes changes, I already see it’s not going to work…
I ask:

  • Are you sure the changes won’t introduce deadlocks?

Gemini’s like Yeah I am sure.

Asking again: Investigate deeply, and check if the changes will introduce deadlocks.

Gemini: Yeah bruh, no problem, I double triple checked, no issue.

Applying changes: okay… deadlock.
Switching to o4-mini, which I don’t prefer because it’s so slow compared to Gemini, but it fixed it…