You ask an LLM to code something and then just run the code blind without reading it and if it seems to work, publish the application. You can imagine how many performance and security problems it produces.