Post

I've spent all day struggling to write a single page of a popular science article. I bang away at a word processor; give up; start diagramming on paper. Take some notes; draft a few sentences in pen; return to the computer...and very slowly I figure out what I was trying to say in the first place.

It's a struggle, but it's gratifying once things start coming together. What I have on paper tonight is a clean expression of idea I didn't have well-formed when I began this morning. Some people think we should hand off this painstaking form of labor to LLMs and get back to doing science.

In my case, this painstaking labor in the process through which the science gets done. Science isn't measuring stuff in a lab. It's thinking deeply, extracting the heart of idea from the soup of thoughts running through my mind, molding it, and finding a way to communicate that idea to others.

I could have finished this piece in five minutes this morning by prompting an LLM. And I think my editor would take it. The rest of the article is solid. This is just the concluding frame. But had I done so, the thought would be flabby and woefully incomplete. I'd be wasting my readers' time.

6 reposts223 likes

There's an attitude that if you use an LLM to do a bit of grunt writing, it's OK as long as you honestly review it. "Yup! That's exactly the itinerary I would've written" or whatever. But people are missing (and I think maybe don't even know) that sometimes writing IS thinking. You have to do it.

The process of figuring out how to express one’s thoughts improves one’s ability to express future ideas in addition to forcing evaluation that may lead to further insights. George Saunders said something like: in the act of writing you discover things missing from the original idea.

I don’t recall where I saw/heard what I paraphrased (I have gone through jags reading/listening to him talk about writing) but he discusses a related theme in Swimming in a Pond In The Rain…where he talks about the need to not fixate on one’s original plan because you discover things ….

Sometime this year i started to do this with my blogging. Instead of just writing out my post long form and asking the LLM to copyedit & also critique and suggest rewrites based on that, I started writing in points & asking it to expand. Only this year it started to mostly work (1)

Often it would be short x/bluesky ranks expanded into full blog post. I would use llms to iterate, highlighting parts I want changed or i would add <url> and prompt something like use argument from there in this section & llms these days I find pretty much get what I mean (2)

Q: is this still writing to think? I also wonder if people can tell which of my blog posts were developed this way vs those I wrote in long prose and LLM merely copyeditted and critiqued to change.

Absolutely not a waste! I really appreciate your post & pulling back the curtain to share the struggle is real but also how to push through without cheating yourself or your audience with using LLM. It's tempting. Hands down. But also, only you can unlock the message of what you are trying to share.

Keep fighting the good fight for human thought and work; nothing feels better than a hard won piece of writing. Question for you: any reliable program that can review and recognize patterns of writing generated by an LLM? And, why aren’t you posting pictures of your dogs?!??!! Monster.

I think the most interesting part of writing is figuring out what you’re trying to write. You sit down with an idea of what you want to write…and it’s broken and clunky and flawed and every attempt to make it flow is awful and slowly you pare it back and re analyse what you’re ACTUALLY trying to say

And eventually if you’re lucky and took lots of walks trying to explain to yourself what it is that you’re trying to express (I always walk alone and pretend I’m talking to my wife about what I’m trying to say so she doesn’t have to get bored hearing me!) , bam! There it is! Clean and clear.