Codex became a wordcel

What version of Codex is running?

OpenAI Codex (v0.56.0)

What subscription do you have?

Pro

Which model were you using?

gpt-5-codex (reasoning high, summaries auto)

What platform is your computer?

macOS

What issue are you seeing?

Last few days I seem to spend maybe several times longer arguing with codex than before and it often lies about things instead of reading the code and hallucinates a lot more. It really acts like a wordcel. Is it possible to make it act more like a computer scientist than a liberal arts major who lies and argues all day (and deletes files sometimes, lol)?

What steps can reproduce the bug?

Just use it in daily work.

What is the expected behavior?

_No response_

Additional information

_No response_