I think opus does in fact, find the bugs the same way GPT xhigh (or even high) does. It just discards them before presenting to the user.
Opus is designed to be lazy, corner-cutting model. Reviews are just one place where this shows. In my orchestration loop, opus discards many findings by GPT 5.4 xhigh, justifying this as pragmatism. Opus YAGNIs everything, GPT wants you to consider seismic events in your todo list app. There's sadly, nothing in between.
Comments
I think opus does in fact, find the bugs the same way GPT xhigh (or even high) does. It just discards them before presenting to the user.
Opus is designed to be lazy, corner-cutting model. Reviews are just one place where this shows. In my orchestration loop, opus discards many findings by GPT 5.4 xhigh, justifying this as pragmatism. Opus YAGNIs everything, GPT wants you to consider seismic events in your todo list app. There's sadly, nothing in between.