There is a huge space of possible software solutions. A very large amount of software that has real value involves using basic methodologies to build interfaces to databases and enable some known business logic. Writing this kind of software is very different compared to making highly general developer tools that need to have finely honed abstractions. LLMs are especially good at whipping up basic apps, but can actually interfere with the process of coming up with and implementing new concepts that operate differently from existing tools. In this respect I find that working with LLMs is very similar to working with graduate students who have energy, drive, and patience, but are also prone to making bad decisions and doing work that ends up being thrown away.