LLMs have made me enjoy programming even more than I already did. No, I don't use them to program for me, but I love having an on-demand entity to ask any question I desire.
I've been finally working on a plethora of ideas that I have put on the back burner. I think my favorite aspect of using LLMs has been the reduction in friction. Mind you, I predominately use the chatbots, I still love how I can go from 0 to 60 in mere hours with languages and stacks I am complete unfamiliar with. In the past month. Chatbots are great for one demand documentation.
Lately, I have had LLMs create little "drills" for me. I have them suggest poorly written, flawed, or buggy code (like a function at most), and then the goal is for me to try to figure out how to write it better and compare my answer with the LLM's instant feedback. These are dumb, little, fun games I play to pass time, but I feel like it's better than doom-scrolling.
As for letting LLMs do the work, I've tried vibe-coding, and I utterly hate it. If you love vibe-coding, then I won't try to take it away from you. I just never like the results I receive. I read every single line produced, and I am never satisfied with the code. It's always an utter mess or full of shortcuts. An LLM could replace me tomorrow, no question. I have to give the bots credit. I am not a particularly impressive programmer in any regard. I am probably slightly below-average to average after a decade of experience.
Still, I am insecure and controlled by my ego -- if an LLM produced it, you did not make it. Maybe I am vain, but I want all the credit. If I produce something, I want to tell people that I made it.
hirvi74 · · focus · HN ↗
I've been finally working on a plethora of ideas that I have put on the back burner. I think my favorite aspect of using LLMs has been the reduction in friction. Mind you, I predominately use the chatbots, I still love how I can go from 0 to 60 in mere hours with languages and stacks I am complete unfamiliar with. In the past month. Chatbots are great for one demand documentation.
Lately, I have had LLMs create little "drills" for me. I have them suggest poorly written, flawed, or buggy code (like a function at most), and then the goal is for me to try to figure out how to write it better and compare my answer with the LLM's instant feedback. These are dumb, little, fun games I play to pass time, but I feel like it's better than doom-scrolling.
As for letting LLMs do the work, I've tried vibe-coding, and I utterly hate it. If you love vibe-coding, then I won't try to take it away from you. I just never like the results I receive. I read every single line produced, and I am never satisfied with the code. It's always an utter mess or full of shortcuts. An LLM could replace me tomorrow, no question. I have to give the bots credit. I am not a particularly impressive programmer in any regard. I am probably slightly below-average to average after a decade of experience.
Still, I am insecure and controlled by my ego -- if an LLM produced it, you did not make it. Maybe I am vain, but I want all the credit. If I produce something, I want to tell people that I made it.