I’ve recently taken to considering Large Language Models like essay assistants. Sure, people will try and use it to replace the essay entirely, but in its useful and practical form, it’s good at correcting typos, organizing scattered thoughts, etc. Just like an English teacher reviewing an essay. They don’t necessarily know about the topic you’re writing about, but they can make sure it’s coherent.

I’m far more excited for a future with things like Large Code or Math or Database models that are geared towards very particular tasks and the different models can rely on each other for the processes they need to take.

I’m not sure what this will look like, but I expect a tremendous amount of carefully coordinated (not vibe-coded) frameworks would need to be made to support this kind of communication efficiently.