Comment by julienchastang

As usual, when we see a thread on this topic on HN, the reactions tend to be bimodal: either "Yes, AI has transformed my workflow" (which is where I mostly fall), or "No, it's over-hyped." The latter often comes with an anecdote about how an LLM failed at a relatively simple task. I speculate that this diversity in opinion might be related to whether or not the user is employing a pro-tier LLM. Personally, I've been very impressed by ChatGPT-4 across a wide range of tasks, from debugging K8s logs to coding and ideation. I also wonder if some of the negative reactions stem from bad luck with an initial "first contact" with an LLM, where the results fell flat for any number of reasons (e.g., poor prompting), leading the user to conclude that it's not worth their time.