I love large language models. I love building tools with them. I love using them for research. I love meandering chats about nothing much at all.
Occasionally, I’ll ask one (or more) to give me some ideas for blog posts based on previous writing of mine (including all of my morning pages). Every time, it makes me just a little bit sad.
I’m not saying that large language models can’t come up with interesting ideas to write about (or interesting writing of their own for that matter), but it always feels like they lack a bit of… soul?
You could say it’s a skill issue on my part. It probably is.
I’ve been trying to make sense of why it makes me feel this way, and the thing I’ve landed on so far is roughly: when you spend the time to look through your past writing; to sit there and slowly, painfully think of ideas, you’re not just thinking—you’re feeling.
You don’t only feel something when you finally land on what you believe is a good idea, either. You feel something as you get close to an idea. You feel it as an idea slips through your fingers whilst you try to grasp it, before you realize there wasn’t much there at all.
It’s a feeling (or a combination of feelings) that’s just completely missing when you craft the perfect prompt. When you sit there waiting for a tool to think on your behalf. When it trots out a long list of decidedly mid ideas that apparently reflect all of your own thoughts.
It took me a while to pin this down, but when I did, I found a new sense of appreciation for those feelings. For the struggle. For the almost-idea. For the break-throughs and for the many ideas that never would be.