If I can get 100 lines generated instantly while explaining it in 25, scan the answer just to validate it and then, no wait, add other 50 lines as I forgot something before. All that in minutes then I'm happy.
Plus I can detach the "tell the AI" part from the actual running of the code. That's pretty powerful to me.
For instance, I could be on the train thinking of something, chat it over with an LLM, get it where I want and then pause before actually copying it into the project.
Plus I can detach the "tell the AI" part from the actual running of the code. That's pretty powerful to me.
For instance, I could be on the train thinking of something, chat it over with an LLM, get it where I want and then pause before actually copying it into the project.