That's not really a fair test because you're leading the model pretty hard, even if the prompt doesn't specifically say there's a bug to be found. It's basically the same objections that people raised in the thread where someone claimed current models are just as good as mythos.