Asking how to read stderr and stdout at the same time destroys GPT-4o

The solution is arguably not super straightforward, but `2&>1` can be emulated by creating a pipe2 and then passing it to both stderr and stdout for the command.

LLMs are useful to me in a lot of ways, I just found it hilarious that it runs into the same pattern over and over and then just gives up, but _still recognized that it's failing_ - most of the time it just confidently outputs broken rust.

Update: https://chatgpt.com/share/684ed2b1-3cf4-8011-a5b6-77c28969ec...

Seems like it's really an infinite loop. It keeps getting more frustrated and snarky.