ChatGPT for Writing and Recommending Books

TL;DR: Today we’re releasing a new episode of our podcast How Do You Use ChatGPT? I go in depth with author Nat Eliason on how he uses ChatGPT to write books and discover new ones to read. Watch on X, YouTube, or Spotify.

Nat Eliason is a shape-shifter. He’s a writer with a book deal from Random House, a crypto trader, a Roam Research aficionado, a marketer, a book podcaster, a parent, and a seed oil iconoclast. For years, he’s published relentlessly on the internet—accumulating thousands of newsletter subscribers, 70,000 X followers, and 100,000 TikTok followers.

Nat just finished writing his book—about his journey through crypto—and he’s starting a sci-fi novel. But joining the dead-tree club hasn’t prevented him from exploring the most interesting tool for thought since Roam fell: ChatGPT.

ChatGPT is now a home-screen app for him—one that he uses every day—and we sat down for a comprehensive interview on how he incorporates it into his routine. Watch him use it to sharpen his ability to describe writing he likes, find new books to read for inspiration, and even outline the story of the new book he’s writing. We also discuss other small but powerful ways that he integrates ChatGPT into his life—from recipe creation to reading the news and settling bar bets. Here’s a taste:

• Recommending books. Nat is an avid reader and has recently turned to ChatGPT to help him find his next book. We discussed not only describing the kind of book you’re looking for, but also working backward based on books you’ve already read and loved: “So being like, ‘My favorite sci-fi books are The Three-Body Problem, Jurassic Park, and The Martian. Who's an author that would be a really good intersection of those three?’”
• A Google alternative. As a former SEO expert, Nat is looking for alternatives to traditional search engines. “I also just don’t trust Google results,” he says. “I’m sometimes disappointed when [ChatGPT] just uses search results… Most of the results on Google are not the best answers to a question—they’re the ones that are best engineered to be there.” 
• A recipe-generator. Beyond his reading and writing, Nat has found an everyday use for ChatGPT—in the kitchen: “I have a bunch of ground pork in my freezer because my friend gave me a quarter-pig and I was like, ‘Okay, I want to make something with this, like Asian-style with rice. I also have green onions in the fridge, so give me a recipe.’ And it did—a very simple stir fry.” The results were actually good, he says.
• Unearthing notes. Nat is building a custom GPT to go through all of his book-related notes and reviews, so he can easily surface and analyze them. “It goes through all of my 300-notes pages, finds them, and then pulls out the relevant piece from that book and links to that book,” he says.
• Reading the news. “I don’t follow any news stuff,” Nat admits, “so I’ll just ask ChatGPT to summarize the information for me.” Instead of seeking out information on news websites, he’ll see what ChatGPT can give him.
• Better ways of explaining things. Sometimes Nat needs different ways to say the same thing. Instead of writing that he has a “sinking feeling” in his stomach five different times, Nat will ask ChatGPT for alternative ways of expressing the same sentiment.
• Storytelling advice. Nat gives ChatGPT a story prompt—the plot line of Fight Club, as an example—and asks for tips on how to structure an opening scene. “Feed it the hero’s journey. Feed it the seven-part story structure. Feed it the three-act structure. Feed it the five-act structure. You can give it a bunch of them and see what it comes up with.”

You can watch the episode on Twitter/X, Spotify, or YouTube. Links and timestamps are below:

• Watch on X
• Listen on Spotify (make sure to follow to help us rank!)
• Watch on YouTube

Timestamps:

1. Intro 1:08
2. Book recommendations based on genre 3:48
3. Figuring out your own taste in books 15:27
4. Finding recipes for ingredients you already have 19:34
5. How he sifts through his own notes 21:50
6. How to reformat data 24:19
7. How to settle bar bets 25:46
8. Summarizing the news 28:40
9. How to phrase something differently 29:54
10. Writing the opening scene of a novel 31:41

What do you use ChatGPT for? Have you found any interesting or surprising use cases? We want to hear from you—and we might even interview you. Reply here to talk to me!

This episode transcript is for paying subscribers.

Dan Shipper (00:00:00:01)

Have you ever used it for trying to figure out what it is that you're looking for?

Nat Eliason (00:00:04:14)

Oh, no, that's an interesting idea. So being like, “My favorite sci-fi books are The Three Body Problem, Jurassic Park, and The Martian. Who's an author that would be a really good intersection of those three.”

We were talking about on tour the other day for book recommendations: This Machinehood book. That sounds super interesting. That's exactly the topic that I'm looking for. I might actually go buy that book. That sounds really good.

Dan Shipper (00:00:29:08)

That's partly the magic of ChatGPT. It misses a lot, but then sometimes it just comes up with something where you're like, “That is absolutely perfect and I never would have found that otherwise.”

Nat Eliason (00:00:39:18)

It can't write articles better than me, but it can maybe give ideas for a fiction outline better than I can do on my own.

The other thing you can do is, you can say, like, “What do you think should happen in this scene?”

Dan Shipper (00:01:07:04)

Nat, welcome to the show!

Nat Eliason (00:01:08:18)

I'm excited to be here.

Dan Shipper (00:01:10:04)

Obviously, we've known each other for quite a while. In that time, you've gone through a couple different phases as a thinker, as a creator, as a writer. And I think when I met you,

you were just coming out of the sort of marketing growth kind of phase and you were entering in your Tools for Thought and Roam phase. You've subsequently really gone into crypto. You're writing a book about that, which is super exciting, or you've written a book and it's

coming out soon. And now you're into books. And I am just very excited to have you on this show because you're just one of those people that thinks really deeply about how to use software to help you think for productivity purposes. I think you were one of the first people to discover Roam and really popularize it. You had that amazing Amazing Rome course. I did an interview with you on your Roam system.

Nat Eliason (00:02:03:08)

I actually owe that to our mutual friend Adam, who originally turned me on to it. Another part of this Twitter gang.

Dan Shipper (00:02:11:23)

One of the first Every employees. The first real Every writer. Adam Keesling.

And so, I just have a feeling that you have some interesting ways of thinking about this and interesting ways to use it to extend your mind and your creativity. And so I'm really excited to get to talk to you about it.

Nat Eliason (00:02:27:22)

I'm excited that you're doing this because. I have found some good use cases for it and it has slowly trickled into being a daily driver, phone home screen app. But I am also occasionally impressed by ideas other people have had for how to use it too. So it does feel like we're very in the early days with discovering how can we use this strange thing—which is fun. We haven't had a lot of tools like that in recent history.

Dan Shipper (00:02:58:14)

I sort of think all of the things that we thought in the note-taking days, what the Roam Research or Evernotes of the world, would do for us in terms of the second brain—like that's actually a lot of what is happening. Like the bidirectional links was not it—this is it you know.

Nat Eliason (00:03:15:22)

Yeah, no. These are true Tools for Thought in the way that the note-taking apps can never be.

Dan Shipper (00:03:23:13)

Totally. Totally. So let's dive in. How do you use ChatGPT? Take us through some use cases.

It looks like you're using it for getting book recommendations. I know you've been trying to write a sci-fi novel and you're looking for some inspiration. So you went to ChatGPT and you wrote, “Is there anyone today writing great near-term hard sci fi—science fiction—without involving space like Michael Crichton? Like, why did you write that? Why did you originally go to ChatGPT in the first place to ask that question?

Nat Eliason (00:03:48:21)

It's the kind of sci-fi that I find the most interesting. And I was like, okay, well, I'm looking for more writers like this. What would be a good place—or who would I ask? And the places I would normally ask would be Reddit, Twitter, or Google. And with Google, you kind of have got to open a few links and cross-reference between them. And then with Reddit, you might have to do the same thing. Twitter, you've got to wait for responses. Reddit, if you post questions, you’ve got to wait for responses. It's just much faster because I assumed the information was out there somewhere. And it did a pretty good job. And the nice thing is you can always say, “Okay, cool, can you give me five more?” And it’’ll go out and grab more of them. I mean, the one thing that can be kind of disappointing is it feels like, okay, these first two, it just went to this article, right? I'm like, okay, I probably could have just gone to this article. I bet it's the first article on Google. All right. But it at least filtered it for contemporary ones. 

Dan Shipper (00:05:01:00)

That's a thing that if you Google for it, there might be an answer if someone has already created a list. But for a lot of those kinds of queries like, for you, that does not involve space. Like there's probably no list that's like specifically Michael Crichton-like writers who've written about things that don't involve space. Like I don't think that's a preexisting thing. And yeah, sometimes if you're a creative person, you're like, you want to find examples that fit this one of one type category that cuts across lots of different categories—cuts across lots of things that people may have composed to list, but there isn't one list because that's so specific to you and what you like. And I think ChatGPT is really incredible for surfacing things like that because it can go through all the lists that have ever been composed and filter it automatically for this one-of-one category that you just came up with. 

Nat Eliason (00:05:58:00)

Yeah. And I also just don't trust Google results. And so I'm sometimes disappointed when it just uses Google results for the answers. Because, like you were saying, I used to work in marketing. I specifically worked in search engine optimization and most of the results on Google are not the best answers to a question. They’re the ones that are best engineered to be there. And so it's hard to find actually good information. You used to be able to go to Reddit and find it, but even Reddit is being engineered now for SEO, so there is a little bit of like, How do we surface good information, not just well-engineered information. I think it is at least a little bit better than just going off of search results. And it certainly helps to filter them, right, because that other list had, what, ten people? But this one has filtered it down based on the criteria of being alive and not talking about space too much. 

Dan Shipper (00:06:53:00)

And I think that's a good note for a tool like this. It's sometimes nice that it uses web browsing, but when it does web browsing, the results it gets are going to have a really big impact on the quality of the response it returns. And if you explicitly ask it not to use web browsing, you might get different results. But the tradeoff there is like when it's not using web browsing, it's much more likely to hallucinate. So you may get it recommending authors or articles or books that don't exist, but it has the potential to be more creative because it's not being limited by the search terms it happens to use to find the answer to your question.

Nat Eliason (00:07:38:18)

Yeah, and it looks like it didn't use search for the second set. So I’d be curious to look at some of these authors.

Dan Shipper (00:07:44:23)

Have you heard of any of these? Do any of them sort of ring a bell for you?

Nat Eliason (00:07:48:06)

No, and so, what I might do is instead of going and searching for them, I would say, “What are the best known works from each of those five authors that fits?”—and then just kind of see what it comes back with.

Dan Shipper (00:08:08:00)

Now it's browsing. So it's going to be pretty likely to probably find something real or say, “Hey, by the way, I made that person up.”

Now I love this. I love this use case. Once you know what you're looking for, it can surface things that you would never have found otherwise.

Have you ever used it for the previous step, which is like trying to figure out what it is that you're looking for?

Nat Eliason (00:08:37:03)

Oh, no, that's an interesting idea. So being like, “My favorite sci-fi books are The Three Body Problem, Jurassic Park, and The Martian. Who's an author that would be a really good intersection of those three.” That's a good idea. I haven't tried that.

Dan Shipper (00:08:56:23)

That is one of my favorite things to do because I have all these lists of authors I like or quotes

I like or whatever. And I can just take the list, paste it in, and be like, “What are the patterns here?” And it just uncovers patterns of my tastes that I never would have put words to before. But once I have it, it's a really powerful tool. That is what I like. I really like emotionally resonant, philosophical, analytical writing. And I know I know that because of ChatGPT.

Nat Eliason (00:09:32:06)

All right. This result is coming back pretty interesting, actually. I already know that the second one wasn't actually fitting what I was looking for, right? Because human interaction with alien ecosystems—I'm not looking for space. SB Divya, this Machinehood book. That sounds super interesting, right? That's exactly the topic of writing for. I might actually go buy that book. That sounds really good.

Have you read it? Have you even heard of that book?

Dan Shipper (00:10:02:06)

I've never heard of it. I think that’s partly the magic of ChatGPT is it misses a lot. But then sometimes it just comes up with something where you're like, “That is absolutely perfect and I never would have found that otherwise.”

Nat Eliason (00:10:19:23)

Okay, you can't see this because it's on a different tab.But I'm looking at it on Amazon. It’s god a 4.2 rating. It’s not like a 4.7, 4.8, mass-market, everybody just loves it. Only 600 ratings. So it's not super widespread, but it's an editor’s pick on Amazon for best science-fiction and fantasy. And the top endorsement is from Ken Liu, who says it's Zero Dark Thirty meets The Social Network and it got nominated for a Hugo Award. This is like a perfect intersection of—Oh, wow. It was endorsed by Ray Kurzweil, too. This is a fantastic recommendation, right?

Dan Shipper (00:11:11:01)

I love that. You saw it here live, folks.

I'm super curious, actually: Your evaluation criteria for whether or not a book is good. You just went through this whole list of things—like where does that come from?

Nat Eliason (00:11:34:09)

It's a good question. I think that there are multiple reasons to read a book, right? One is pure like popcorn—I just want to be entertained. I just want to have fun and I want to go on a little journey here and there's nothing wrong with that. That’s a great reason to read a book. And if that's what you are looking for then you want the 100,000 reviews, 4.7 stars, Oprah's Book Club—whatever. You want those books because those are pretty much guaranteed to deliver that experience if you enjoy that type of story. But if you want something that is not going to be—for lack of a better term—a cookie-cutter plot, like something kind of new and interesting or there might be rough parts to it, but it's going to maybe expand a little bit of what you think of

as what you can do with a novel. Or it's going to introduce some interesting ideas—it's going to do something a little bit different, so you're not going to feel like, oh yeah, I've done this before. Then you kind of have to look for something that has, maybe the not-great reviews, especially if it's a lot of like five-stars and one-stars, right? That's a really good sign. Or it's got incredible endorsements, but then mediocre ratings. Some smart literary people really like this, but then a lot of like mass-market readers didn't like it. That might actually be a really good book. Those things to me are often really good signals. Because it's the books that don't really say anything but are like wonderful junk food that get the best ratings. Usually, you see that with nonfiction too. If I'm looking for that kind of experience, that's the kind of like a rating-spread or whatnot that that gets me. 

Dan Shipper (00:13:37:00)

So you're kind of like, if there's a love-hate dynamic with the crowd-sourced reviews on Amazon, that's a book you want to look into because it's probably taking a real position.

Nat Eliason (00:13:51:05)

Yeah. And if I think about some of the books that I have enjoyed most, like Antifragile or like Taleb in general. Very divisive person, divisive writing style—a lot of people hate him, a lot of people love him. But he’s very opinionated. It's interesting. I mean, another one of my favorite fiction books I read in the last year or two is this super manic kind of dark novel called Stephen Florida. It's got 200 reviews on Amazon with a 4.0 rating, but it has a very strong endorsement on the cover from the author of A Little Life, which is incredibly well-known, literary fiction. So it's like there's a disconnect there. And it's a great book. So to your point, this would be interesting way to use ChatGPT.

We could even try that. Let's just give it a go, “I really enjoy manic dark fiction, like Infinite Jest show and Stephen Florida.

Dan Shipper (00:15:15:16)

You might want to ask, how would you describe that?

Nat Eliason (00:15:27:00)

“How would you describe the overlap— How would you describe the taste in books for someone who really likes Infinite Jest, Choke, Fight Club, Stephen Florida?” Do I have any other examples? My fiction bookshelf is at home. Oh, “Blood Meridian.”

I mean, that's a really good description.

Dan Shipper (00:15:56:03)

Isn’t it? Like, that bottom one. Yeah, “intellectually challenging author, unconventional ideas and don't shy away from darker or more complex themes. One of the things I get from this actually, which is so interesting, is it's a huge part of how you write. I'm not surprised that those are the kind of books you like because I think a big part of your writing is sort of challenging conventional narratives and offering an unconventional one. I think your writing is highly intellectual. It's a really interesting thing.

Nat Eliason (00:16:35:14)

Yeah. This is super powerful. It's a great way to expand your readership. Another interesting question for me to ask would be, “Who are some women who fit this theme? Who are some Asian writers or Indian writers or Russian writers who fit this theme?” Because I mostly read American white guys—and it’s not a deliberate choice, it’s just what ends up coming in front of my view—but there are definitely people with different backgrounds who touch on this style too. And so it'll be an interesting way to broaden that lens without doing it in a forced, contrived, checking-the-diversity-box way.

Dan Shipper (00:17:30:22)

No, I like that. Frst, you get the words, you get the dark, complex themes, like grittiness, all that kind of stuff. And then you get the “I want you to expand this definition in this one dimension.” Like, “I want to read writers who are not just like white guys, who else have done this.” But it would be almost impossible to Google for that. What you need ordinarily is to ask someone else who has the same taste as you, but is slightly different in that dimension. It’d be hard to find but ChatGPT is that person.

Nat Eliason (00:18:15:06)

Totally. I should show you—

Dan Shipper (00:18:16:21)

Have you read, above it said Don DeLillo, Thomas Pynchon, now it’s saying Margaret Atwood, Toni Morrison.

Nat Eliason (00:18:28:16)

See, this is interesting. Toni Morrison and Octavia Butler have both been on my list to get to at some point. So it's like it's kind of confirmation. Margaret Atwood, obviously, I know of her but she hasn't been on my list as much. And then Joyce Carol Oates. I have her short stories on my shelf at home, but—

Dan Shipper (00:18:58:12)

Lydia Davis is great. I really, really love her short stories—they’re really good.

Nat Eliason (00:19:02:18)

Dude, I don't like reading short stories. I don’t know if that’s weird.

Dan Shipper (00:19:08:16) 

I was the same way. And then I read some Lydia Davis and Ted Chiang and I was like, “Oh I actually like short stories.”

But yeah, let’s move on. So, you’re using it to help you find new things that are sort of on the same vibe of things you like. What else are you using it for?

Nat Eliason (00:19:34:12)

Okay, I mentioned this one. I just want to bring it up in case somebody else brings it up in these interviews. It's so good for recipes. So basically I was like, I have a bunch of ground pork in my freezer because my friend gave me a quarter pig and I was like, okay, I want to make a stir fry—or I just won't make something with this like Asian style with rice. I also had a bunch of green onions in the fridge, so, give me a recipe.  And it did. Like a very simple stir fry and you know, I changed some things out here. I just didn't use vegetable oil, I used ghee instead. But this isn't a complex recipe. It gives you all the instructions you don't have to get through, you know, 2000 words of someone's life story to get to the recipe. And it's good— I've cooked this a couple of times now. It's a great recipe. But here's the really fun thing, okay? Because, like, this is all I did, right? Asian, ground pork, good with rice, green onions, and it comes back with this. The problem was I didn't have any hoisin sauce, so I said, I don't have hoisin sauce. Can you modify the recipe? And it's like, Yeah, no problem. And it replaced it with brown sugar, which is apparently a good substitute, which I didn't realize. And then, everything else is the same. But it's like, all right, cool, you've got this recipe, it needs to be modified and it can just do it on the fly. And the thing I haven't tried that I bet would work is if you take a picture of a recipe with your phone and upload it and say, hey, can you extract this recipe and then modify it to replace this ingredient because I don't have it—I bet it would work. I haven't tried that. But if you're working with a cookbook, I'm sure—

Dan Shipper (00:21:17:16)

It definitely will. The vision stuff is so good. It's so useful for stuff like that. 

Nat Eliason (00:21:20:00)

So yeah, it's like if you've got stuff in your freezer that you haven't put to work yet, or you want to try new recipes but don't want to flip your cookbook, I mean, this is really, really powerful. It's awesome. I love cooking with it. 

Dan Shipper (00:21:39:18)

That's amazing. If you scroll up on your sidebar, I think I saw something like recipe-related in your custom GPTs. Did I remember that right?

Nat Eliason (00:21:50:16)

Oh no. Okay. So this is just something else I've been playing with is the book notes. Because I have, you know, like 300 of my book notes published on my site. So I've been trying to work on a custom GPT to surface results from them.

Dan Shipper (00:22:06:18)

Is it working?

Nat Eliason (00:22:07:23)

It’s close. It sometimes has this issue—

The way I have it set up is it looks through—

Right now, it's just going to the web version, but eventually I'll upload all of them so it'll just be able to search. But what it does is it goes through all of my 300 notes pages, finds them,

and then pulls out the relevant piece from that book and it links to that book. So you can be like, Oh, this book seems to be talking about this idea. If I want to explore further, here's a great place to start.

Dan Shipper (00:22:49:16)

Do you find yourself using this?

Nat Eliason (00:22:53:17)

No. One, because it's not totally working yet. But, two, because I want to make it part of my book influencer thing. So basically, I want to have this as something that people who follow me on TikTok and Instagram can like go use if they're ever looking for book recommendations. Because people ask me all the time in my comments to say, “Oh, can you recommend a book on X, Y, Z?” And it's like, I can't respond to all of them, but they can come here and they can plug in a question and it'll tell them every book I've read that touches on that topic. And it links them to like my notes on that book. They can go see all my notes, they can see my review of the book.

And look, this is a pretty good result: The first thing that came to my mind was Ten Percent Happier, but  oh, Deep Work talks about anxiety too and Work Clean was not one that I would have thought of, but it totally does now that it's mentioned.

Dan Shipper (00:23:47:18)

But one thing you could do, which could be kind of fun, is you could load into a GPT your list of books that you've read like one sentence about what you liked or didn't like for each book, which would fit into one prompt. And then you could have ChatGPT recommend, as you, based on—it doesn't even have to be like a book that you've read. It could just be like a book that fits the taste it knows you have.

Nat Eliason (00:24:12:09)

Yeah, that’s a good point.

Dan Shipper (00:24:16:12)

I'd be really curious to know if that works.

Nat Eliason (00:24:19:21)

So that's been a fun little project. There's a couple other little ones I wanted to quickly show you.

Okay. Yeah, if you've got something—

So if you have data that you just want formatted a slightly different way—like these are timestamps for a YouTube video—pasting it in and saying, can you put brackets around the times and remove the spaces? It's way faster than doing it yourself, which is like a silly thing, but it's helpful.

Dan Shipper (00:24:50:22)

Yeah, I do think ChatGPT is really good—there are all these things that we do in life that we're, we're doing subtle translations for. We're subtly transforming a piece of information that's in one format into another format. And computer language was developed or machine learning evolved in large part to do actual translation between English and French or like French and German or whatever. But I think it turns out now that we have language models that work really well, that there's a lot of translation that needs to happen between English and English. And sometimes it's just like one format and another. Sometimes it's like there's a subtle way

that people in tech talk, for example, like if you're trying to write a product spec for a piece of software you're building, and if you're not a tech person, you can't write that. But now you can. And I think that's so powerful.

Nat Eliason (00:25:46:05)

I agree.

Okay, another quick one: Bar bet type of things—it’s really good for. So I was having this conversation with my friend Neil. There will be—probably within our lifetime companies that build businesses around converting atmospheric CO2 into, like, stuff because if you just pull off the carbon atom, then you can use all those carbon atoms for something else, right? And that would be like one way to try to reduce carbon dioxide in the atmosphere. Or maybe it would just be profitable to do it. I don’t know. And so I was like, well, how much air would you need to make a diamond? Like, could you have a business that just captures CO2 and then poops out diamonds and oxygen? On the physics level, it should be possible, right? But is it energy efficient? Whatever. And so it's, well, how much air is it? It just does all the math. It walks you through every step. And dude, the volume of air is much smaller than I expected. It's a little less than a cubic meter. And so then I said, okay, well, how much CO2 does a person exhale in a day? And then it gives you how much air you give out in a day—about ten cubic meters of CO2 per day. So you could make about ten diamonds a day.

Dan Shipper (00:27:18:01)

Just by breathing.

Nat Eliason (00:27:20:05)

Just five or ten one-carat diamonds per day by breathing.

Dan Shipper (00:27:24:05)

Like, man, I mean, like I'm working so hard now. I don’t need to be working this hard.

Nat Eliason (00:27:32:14)

I know. We’re wasting all of this! But I thought that was super interesting, right? I don't know what I was reading recently—maybe it was Anthology of Biology or maybe it was Novacene. It was Novacene. He was talking about how we might eventually have diamond-based processors because it might just be a superior material for it. And so if we just have machines that pull carbon out of CO2 in the air and then poop out diamond processors like self-replicating machines—actually it’s not that crazy of an idea physics-wise. It's kind of fun to think about it. It's like a very Where Is My Flying Car-type idea.

Dan Shipper (00:28:24:06)

Novacene is the book by James Lovelock, who's the guy that got the Gaia hypothesis that sort of looks at the earth as one living organism as opposed to many organisms.

Nat Eliason (00:28:40:17)

Let me see if I have any other fun ones for you.

I've used it for news stuff sometimes. I literally just, I don't follow any new stuff. Like I just don't let it into my life. But sometimes I'll hear something mentioned by a couple of people, but, I don't want to go to New York Times or whatever. I don't want to read a piece on it, so I'll just ask ChatGPT to summarize the information for me.

Dan Shipper (00:29:11:14)

I love that. I honestly did that recently when I had to write a summary article of what happened at OpenAI Dev Day. And I started by just being like, what happened? Even though I was there. It was really helpful because it just puts it all into a format that makes it easy to be like, oh yeah, that happened. You need to put that in there.

Nat Eliason (00:29:29:13)

Yeah. I think was really smart. I can keep going—

Dan Shipper (00:29:34:04)

Yeah, scroll up if you don’t mind.  I would love to just look through it and see if I can pick through anything that looks kind of interesting. And if you don't want to show us anything, that's totally fine.

Nat Eliason (00:29:45:12)

Totally. It's great for writing…

Dan Shipper (00:29:50:04)

Yeah. How do you use it for writing? Do you have any examples of that?

Nat Eliason (00:29:54:18)

Yeah. So I haven't found it very useful for nonfiction writing to be honest and—saying this with the utmost humility—I think it's just because I'm much better at it than it is. And so, if I need something explained, it can be helpful. How would I describe this thing? Or, so I can fact

check myself. But actually writing, I haven't found it very useful.

But with fiction writing, I've found it more helpful just because I’m much weaker there. And this is one good example. This emotion comes up kind of often—I feel a sinking feeling in my stomach. But you don't you don't want to write that 10 times in a book or like 20 times in a book. And so it's really, really helpful for alternative ways to describe things.

And, you know, some of them are a little contrived—”a heavy sensation weighed down my insides.” And even some of these can be modified to be better—”I was stricken with nausea” or “a pit formed in my stomach” or “stomach dropped.” These are all good alternatives. So it's very helpful for that. I've been using it for—I don’t know if I want to show this one on the chat, but I can describe it. Basically like pasting in details on a scene in a novel and then asking what questions it has about that scene is really helpful.

Dan Shipper (00:31:40:21)

What's helpful?

Nat Eliason (00:31:41:11)

I will just do a live one quickly. Yeah, let me just show you.

I'm working on an opening scene for a sci-fi novel. I'm working on a sci-fi novel where—what happens in the sci-fi novel?

Dan Shipper (00:32:05:12)

You could even take a chapter of a book that you have that's already a released book and just paste it in there. If you have one available.

Nat Eliason (00:32:12:05)

Oh, yeah. Let's just take Fight Club. “I’m working on a novel where a man is disenfranchised with society and slightly psychotic starts a secret underground fighting with other men as a dark form of therapy. His best friend turns out to be a hallucination—spoiler alert. In the opening scene, he is flying home from a business trip and discovers—I can’t remember if this is the opening scene or not—that his apartment has exploded.

And then, normally I would paste in five or six bullet points about things I think

are going to happen in the scene. But the other thing you can do is you can say like, what do you think should happen in this scene?

Dan Shipper (00:33:28:06)

And we should probably preface this by saying ChatGPT knows about Fight Club. So it may—

Nat Eliason (00:33:32:23)

Yeah. It's probably going to just pull in things from Fight Club. But some of these questions are really good, right? Like, “What's his emotional response to seeing his apartment blown up? Is he secretly happy about it? Is he freaked out? What's the interiority?

Police and firefighters should be in the scene. What does that interaction look like? All these little things can be quite helpful if you hadn't thought about them. And what I usually find is when I do this, I'll ask it to give me like ten responses and usually does. It seems to be going towards 10 right now, and one or two of them will be very helpful. Most of them won't, but they'll occasionally have some very helpful details in them.

It's also helpful for—if you had an idea for a story you wanted to flesh it out, you can say something along the lines of, “I'm still working on fleshing out the whole story. Can you give me a Joseph Campbell hero's journey outline of what this whole story could look like? And I've found a couple of interesting ways to do this. So one is to feed a bunch of these story

frameworks—so feed it the hero's journey, feed it the seven-part story structure, feed it to five- or the nine-part story structure, feed it to three-act structure, feed it five-act structure, right? You can give it a bunch of them and see what it comes up with and you can regenerate each response a few times. And then like from that you can make an outline or you can like apply whatever you like in it to your existing outline and then you can feed that outline back in and say “What questions do you have about this outline or what holes do you see or like anything along

those lines?” 

Dan Shipper (00:35:44:16)

Interesting. So when you say “feed it”  are you putting all of those structures in at once? Are you starting a new chat? You're saying, “I want to do Joseph Campbell's hero's journey. Here's the premise of my book. Like, can you help me flush it out into a hero's journey and then going on to the next one? Are you doing like, “Here are a bunch of different structures. I want you to flesh out this idea and all those different structures” and doing it all at once.

Nat Eliason (00:36:10:17)

I'll probably do it all in this chat, so I would be like, “Okay, great. Hold on to that thought. Can you try again using this structure? But like, don't carry over anything that you just thought of start from scratch or something like that.

But like, this is a pretty good outline. And like, again, it's probably pulling from Fight Club, but it's not doing like it's not it's not doing a bad job. We can pull a pretty great story out of this. But this is fascinating.

Dan Shipper (00:36:45:18)

And this is great, too, because, you know, I'm sort of I'm sort of in the beginning stages of fiction

writing again and like I've been wanting to outline because I do think that outlining is helpful for doing great plot-driven stuff. But I'm just not like an outline person and this is cool for that.

Nat Eliason (00:37:01:19)

And the thing I've found is kind of like with nonfiction-writing, it's not very good on its own. You have to give it a lot of your own creativity and guidance, but it can give you that extra 10-20% 

or it can help speed up your process. So it's been very, very helpful.

Dan Shipper (00:37:18:14)

I'm sort of curious. Even moving away from some of the practical hands-on stuff like for you, like I think a lot of people try Chat GPT and they're like, oh, like it doesn't give me the right answer or whatever, and they sort of get frustrated and move on. But it seems like you've managed to be okay with the fact that it's dumb sometimes, or maybe it's dumb a lot. And you've also managed to find lots of different places in your life where it can be useful. Tell me about that.  Tell me about the mindset that you have to learn to use this well. 

Nat Eliason (00:37:54:00)

Hmm. Well, I mean, before we go on, I should mention this use case in particular, the outlining I actually got from our mutual friend Nathan Baugh, who talks a lot about storytelling and fiction writing on Twitter. So people who are curious about this type of stuff should go follow him because he has a lot more resources on it and something we've talked about quite a bit.

But yeah, that's a good question. I mean, I think it's like any person, right? When you work with other people,you realize that they're good at some things and not good at other things. And if you expect them to be amazing at everything,  then you're just going to be disappointed by it.

So I kind of just think about it that way in the sense that, right now it's like a very eager teenager with an internet connection—is probably where I would put it. And I'm better than a teenager at like most intellectual things probably, but not everything. And certainly not at the things that I have never done before. So, yeah, it can't write articles better than me, but it can maybe give ideas for a fiction outline better than I can do on my own. And it can certainly do these little things that I could do but don't want to do. Right? The way you might outsource something to a VA. And so I don't know. I think just like having that idea that this is not a god, right? Like we're still very early in its evolution and kind of like having that mentality is very helpful and being willing, I think also like, I don't know, I mean, we all grew up in the era of, you know, having to install drivers on our Windows machine and like pop the cartridge in the N64 and blow on it and pop it back in—it takes a few tries to get stuff to work, not everything is an iPhone. At least it used to be that way. 

I feel like people are getting spoiled by technology now. Back in our day… you had to struggle with it!  You didn't know if you're getting a virus or a song from LimeWire. It was exciting!

Dan Shipper (00:40:12:20)

It gave you the gumption to explore new technologies. Not like kids today!

Nat Eliason (00:40:17:08)

Exactly! Not just ask Siri for everything.

Dan Shipper (00:40:22:00)

No, this is wonderful. Thank you so much for for taking the time

to show this to us. Excited to get to chat with you once again about sort of Tools for Thought stuff. And I'm psyched for your book to come out.

Nat Eliason (00:40:34:05)

Thanks, man. Yeah. July. It’s coming up soon.

Dan Shipper (00:40:41:18)

All right. Thanks, dude.