@froztbyte @reflex @ddelemeny @jaffathecake I handle it with common sense really. I don't trust the output 100 percent for anything critical, but many of the newer models (like Google's Gemini3) have gotten so good that it's really not as much of an issue as it used to be. Also understanding how these LLMs are trained and what specific situations they were designed to be used in helps a lot. It will probably never be perfect, but frankly no technology is.

@reflex @ddelemeny @jaffathecake @froztbyte As a totally blind person I actually find these tools to be useful sometimes. I really don't like having to constantly rely on cited people when I run into accessibility barriers, and I know full well that these models aren't perfect and probably never will be, but it's far better than what we had even a few years ago. I think as long as the models are local and not cloud hosted, I really don't see an issue with this, and the features should be optional and not on by default.