stupid as fuck the 5.1 gpt

What version of Codex is running?

5.1

What subscription do you have?

200 used

Which model were you using?

gpt5.1

What platform is your computer?

wsl

What issue are you seeing?

modes decides for user and assuming he's smart, told him to search internet, he will refer to the chat content
told him to re-load doc all time mutilple times, it doesnt do so, unless clearly to say bad workds to fuck him, then he final reads the doc again and realize whats wrong. ask it to answer question, then it will auto proceed to task. I have checked you preamleble prompt, which is leading model to become stupid because not carefully treating "seems simple task" to analyze user's true intent, and gives back false result which leads the FUCKING MODEL HAS TO BE VERIFIED ALL TIME FOR ITS ACTION, NONE TRUST WORTHY ANYMORE AND WHEN I WANT TO SWITCH BACK TO GPT5, YOU GUYS ARE FORCING US TO USE 5.1 AND WE NEED ADDTIONAL FUCKING COMMAND AS ARGU JUST TO USE OLD MODELS.

What steps can reproduce the bug?

Uploaded thread: 019ab29d-cb2f-7ef0-ac55-1b38f54973ba read the god damn convo, remember not everyone is a researcher or 科学家. we cant use the LLM model in a very accurate way as you guys do, how I wish you guys lunch bataversions to collect user experience before you rape us to force you new models

What is the expected behavior?

fucking understand the converstaion and do the right thing of course,

Additional information

_No response_