today in LLM replacing engineers: Ran a code review agent with GPT 5.5 across all my 12 requested code reviews, the only one that the AI had no comment on (i.e. code should be merged as-is) is the one that had to be abandoned because we were working on the wrong thing. Context still matters. AIs need to have more advanced context and need capabilities to reason about the motivations of changes. Building the right thing is just as important as building right.