Treating LLMs as programming books

LLMs can often perform localized miracles, quickly building something that works, or at least gives the appearance of working. When an LLM does something that you’re not an expert on, it’s like magic. It can write Terraform HCL, knows every single bash tool invocation, tackle your Tailwind issues (after a few false starts), and debug your Kubernetes cluster, all without breaking a sweat. Of course, the more you know about a topic the less impressive it is. It often makes things more complicated than they need to be, it’s rarely elegant. Still, if you care more about building something and getting it in front of users than you do meticulously crafting each line of code, you can produce something really fast. Working with LLMs over a longer period of time, your standards for acceptable code change. It’s not necessarily that they get worse, but the pattern I’ve seen is caring less and less about each individual line of code, and the focus during review shifting more towards overall…