yes exactly! it's able to "guess" the output of the code without running it in the traditional sense. I've even able to get it to do loops correctly. This is ki

yes exactly! it's able to "guess" the output of the code without running it in the traditional sense. I've even able to get it to do loops correctly. This is kinda the reverse of it producing code based on comments. But it goes further: now you can make up API function calls that don't exist and compose them with code that does. It feels like the entry to an entirely different space of prompt engineering.