> Another version of this issue is when you push back but you were NOT "right to push back". In other words, the LLM original solution was better than the pushback.

Indeed, it's easy to surface this by sending one model a "Review" of their proposal to another, then bounce them back and forward, ask which one is best and both models will almost always say something like "The other proposal/review is better", I'm guessing because somehow they think it comes from the human, and "human is always right" or something.

I've had those as well. Sometimes I'm asking clarifying questions because I'm not sure about the solution, and the LLM "interprets" that as pushback (as opposed to curiosity / enquiry), and sycophancy takes over. Sometimes it will simply change the code without ever answering the questions, or it will answer the questions along with it, but incorrectly - or with bad assumptions.

> Answer grounded in truth, with evidence and concrete proof, no guessing or assumptions allowed, no changes to files on disk.

I've used this a bunch as a suffix to try to prevent that, works OK in most cases, but not always obviously, works better in the system/developer prompt if you have access to those. Seems I've used that about ~1000 times since 2025/08 when I started using codex (- transcription duplications, so maybe 1/2 of that?).

    $ rg -a -o "Answer grounded in truth" ~/.codex/sessions | wc -l
    1046

This is why I love plan mode and switch to it as soon as I see the agent starting to deviate from the plan because it hit a wall. Gotta get back to aligning on design decisions, approach, reverting any debug code, etc. before progressing. (Damn, I have more instructions to add to AGENTS.md lol)

Although this was more of an issue when using opusplan as Sonnet loves to do this. I only use Opus now if using Claude.