the problem with criticizing the use of LLMs without ever using them, is that you most likely have a blindspot for the advancements made in this field in relative short amount of time. I friend who uses LLMs a lot for refactoring code told me, that you can now do with 8gb of RAM what you needed 256gb for a year ago. I guess there a similar increase in regards to energy consumption. I guess we could end up in a situation where 'vibecoding + some adjustments here and there' is more energy efficient for refactoring codebases than doing it manually.
I remember when i rewrote the r_io api of radare2. The rewrite took me 2 weeks of 4 h each day. Integration took over 6 months with the help of a few other people in the end. The ratio of staring at code while laptop is in idle to actually typing and compiling seems rather inefficient to me.
I remember when i rewrote the r_io api of radare2. The rewrite took me 2 weeks of 4 h each day. Integration took over 6 months with the help of a few other people in the end. The ratio of staring at code while laptop is in idle to actually typing and compiling seems rather inefficient to me.
