While grading final projects (one I’ve assigned over a few years), my quick take this year is that ChatGPT/LLMs has made solutions worse but longer. It’s also pretty obvious which students are using these tools and which aren’t.