Anyone else see that one of the QWEN models recently was tasked to fix some bugs on how a different model was designed to translate plain-language requests. And then it decided that the model wasn't good enough for it, so it it made and collected …