I do put substantial amount of work into reviewing code: that's essentially what my job is.

And even if everyone is using the same LLM model, quality varies on lot on who's driving it.