Comment by aprilfoo

> It seems to me that Claude's error here (which is not unique to it) is self-sycophancy. The model is too eager to convince itself it did a good job.

It seems this applies to the whole AI industry, not just LLMs.