How a Writer Uses AI Without Losing His Voice

[00:00] I wake up and I don't touch the internet. I won't look at my phone. I won't go online until long after lunch. As soon as I touch my phone, I feel the chemicals shift and I can't go into any kind of deep thinking place. [00:11] or deep attention plays or deep focus plays. Because there's something about working with AI that feels very dopamine driven. And I believe strongly you can use it to do incredible work. And also it's sort of a slot machine. I now have a laptop just for writing.

[00:23] that blocks everything. I can't do anything interesting on it. If I don't create these barriers, I will lose connection with this other part of me. That I think is the most valuable part of me. There aren't that many people who are going to think about or write the weird books that [00:35] I feel like I'm drawn to write. [00:36] and [00:37] As a human, that feels like a valuable thing for me to put my effort in. [00:55] Every is the only subscription you need to stay at the Edge of AI.

If you care about being on top of the latest models and using the latest tools, you have to subscribe to Every to separate out the signal from the noise. Go to to slash subscribe today. Greg, welcome to the show. Hey, thanks for having me. [01:10] Um, so you are a writer, a walker, a... [01:16] player with technology, a photographer. Sure. Um... [01:20] And I think you have a life and career that I have looked up to for a long time. So it's really fun to have you here. And. [01:28] I think one of the things that we were just talking about is the thoughtfulness and the spirit with which you do your work.

Mm-hmm. [01:37] And, [01:38] I admire that spirit and try to bring into the stuff that we do here. And also I've been trying to think about how does AI fit in with it? Cause there's something about working with AI that feels very, um, [01:49] Dopamine driven. Mm-hmm. [01:51] And you can, I believe strongly you can do it to, you can use it to do incredible work. And also like it's sort of a slot machine. Yeah. Yeah. So. [02:00] I know that you're like deep into the vibe coding right now. So tell us, tell us what you're doing with it.

And then let's talk about how it fits into your work. Yeah. I mean, look, I just think if you're not touching it, if you're not using it, if you're not building with it. [02:13] you can't really comment on it. And, you know, and I think part of what, when people go, it doesn't really work. It's so unpredictable. It's like, okay, just try, just use it. It's like, if you're using it, you go, okay, this is ridiculous. And why is this technology public slash like, shouldn't this? I mean, I kind of agree with a lot of what Dario says, like there should probably be more regulation around this.

Like, I don't know if everyone in the world should just have this immediately. Um, certainly, um, [02:40] you know, as we're working out the kinks of what this, [02:43] is in public. I mean, it just feels strange. It feels like we are in this aberrational moment where technology like this normally isn't just given to everybody, you know, for free, essentially. So, [02:54] Look, I think you really have to be engaging with it, especially if you can. You know, it's like Fable came out. I played with it for four days. I had to, like, refactor code bases.

I had to do, you know, these other little projects. I was like, great, this is cool. And then it went away, and I was like, wow. It kind of makes sense that this would go away. But also, Opus 4.8 is sort of, like, good enough to do anything. Like, if you told me you only have Opus 4.8 for the rest of your life, I'd be like, great. I could build anything with this. There's a lot I can do. Yeah. So it is this weird moment. [03:24] that Fable is not here anymore.

I miss her. Do you miss her? I don't know. I honestly don't really feel... [03:33] qualitatively, maybe I'm not using it enough, but the differences between these models like that explicitly, I just know that, oh, it feels a little more rigorous and it eats my tokens like crazy. So it must be better, right? [03:45] So I just hacked into Fort Knox and stole all the Google. Yeah, I know. It must be better. Yeah, I've got all the Bitcoin now. It's all me. I've become Satoshi. But look, I think you have to be using it.

So that's one of the main reasons I... [04:01] I've been using it. And also, I just I've always loved technology. It's been a part of my life since I have memory, you know, always curious about let me take this thing apart or build this or use this piece of software or whatever. Like it's always been just part of who I am. [04:14] It turns out that's also part of who my birth mother is. She's a computer scientist. It turns out. Programmer, weirdly. She never went to college, but she's a, yeah, I met my birth mom.

I was like, what do you do? She's like, oh, I'm a programmer. I was like, oh my God, okay. Yeah, nature. There is no nurture. It's all nature, it turns out. Nurture is just your pathologies. Like all the things that you hate about yourself is probably like nature, are about nurture, and then all the stuff that you're like intuitively is confusing to you is just nature. Holy shit, that is fascinating. When did you find that out? Just like, yeah. [04:44] a little under two years ago. Wow. Yeah. So it's been interesting.

So anyway, technology is important to me because I'm just curious in that way. And I think it's the most interesting thing in the universe right now. Genetically predisposed. A little bit genetically predisposed. So I've been building stuff around my membership program, um, [05:00] Well, I rebuilt Quicken, which I am... The tax software. Yeah, the tax software, because you can watch my interview with Paul Ford and Rich over on a board about that. So I rebuilt Quicken. [05:13] And then... [05:13] I rebuilt Campaign Monitor, which is a newsletter software, because... SaaS CEOs are losing in this interview.

I know. Well, okay. And why are they losing? They're losing because... [05:27] Well, first of all, [05:28] Campaign Monitor has not done anything to... [05:32] in a decade. [05:34] That's just been sitting there taking people's money for a decade plus with no innovation whatsoever. Zero. Nothing. With like a really abusive pricing policy. How much were they charging you? [05:46] Thank you. [05:47] So I was spending like... [05:49] what was it like six, $7,000 a year. [05:52] I mean, it was like one of the most expensive things I... [05:55] I paid for.

Um, [05:57] But the real annoying thing was that they didn't count unique email addresses. They counted total, even if you had overlap between lists. So it's just – it's like things like that where it's like – [06:08] You don't just I'm going away. Bye. Like you're squeezing too much out of it. Yeah. And my bill now with like Amazon SES, I set up all the DKIM and SPF and all that email crap. So I think I'm getting pretty good deliverability like on par with what Cam Paymoner gave me. But now my like yearly bill is probably going to be like, I don't know, like $150 or something.

Like it's just a ridiculous difference. And how much in Fable tokens? [06:31] Well, thankfully, yeah, this is all Opus. But that's the other weird thing. I was on the $200 a month plan, and then I realized I wasn't hitting my limits. I dropped to $100, and for $1,000 a year, $1,200 a year, I'm not going to do that. [06:47] that is the easiest trade. Like that's the best, that's the deal of the century in terms of what you can build. So I built, [06:54] newsletter software. And then I've been, I've been adding to my membership program software.

So actually one of the things I had Fable do was do a, like a full, uh, just, uh, [07:02] Security check on my membership. [07:04] Um, [07:05] Flask Python app, which I built by myself before any of this existed. And so I went through, I was like, oh, you have 8 trillion security holes. Shall I fix them? And I was like, yes, please. So I think it's... [07:18] A little safer. I mean, the stakes are very low for my membership. [07:21] software but then the fbi knocked at your door yeah [07:25] You got to stop using this.

So, you know, I've been doing that. And then I rebuilt Twitter for my members. We call it The Good Place. Oh, yeah. You were showing me that. Mm-hmm. [07:35] Tell me about that. Tell us about it. Yeah, I mean, look, like, social media is sex. It's terrible now. There was this moment... [07:43] 2010 to 2014, where it was an algorithmic, [07:49] Stuff just appeared. [07:51] Links were favored. [07:52] I mean, they were just normal. They weren't a hit against you if you post the link. And everyone was on Twitter. It was like the great media watering hole.

And like if you were active then and slightly interesting, you could like. [08:05] you know, a hundred X here network. [08:07] just by being an interesting person on Twitter. [08:10] Now you kind of can't because the algorithm favors psychosis versus sanity. And no one likes posting to any of these places. All these reply guys pop in. It just sucks. And Threads is... [08:21] doofy and blue sky is like everyone's crying and complaining like it's just a weird... [08:27] Everyone's so traumatized. And so I wanted a place... [08:30] Just for members, you have to pay to be there because you have to pay to be a member.

And I was like... [08:36] you know, I had these, these thoughts. I always thought Twitter should be ephemeral. I think there should be no archive for these things. I'd like, it's, [08:42] crazy. There's like a web 1.0 slash web 2.0. [08:46] philosophy of everything needs to be archived, [08:48] and saved forever, and I kind of don't believe that. In fact, one of my heresies, [08:53] When I'm with Kevin Kelly doing our walks, like we have heresy night where you say something you sort of believe, but like people you respect don't believe maybe.

And one of my heresies is I think the entire internet should be deleted every two weeks. And like, [09:05] you can archive it offline or whatever, but like the entire internet itself, every two weeks, you just disappear. And there could be, like it, it forces a lot of, [09:15] Like, why would you do that or whatever? But I think there is kind of value to ephemerality in the digital world. So anyway, the good place everything disappears in a week. You can only post twice a day. You can reply 20 times a day.

Photos are in one bit black and white until you click them and then they fade into color. [09:33] And... [09:34] What else is there? No algorithm. It's just [09:36] reverse cron uh and it's just sweet i don't know this is lovely it's like my favorite place on the internet it's the only place i want to post and people have now been joining my membership program uh just to get access to that which i think is cool so it seems like and partly you were doing this before you know the acceleration language model coding or last year year or so so you're tactical i'd be i'm technical i mean i built my membership software by hand by myself like i [10:06] it up, set the roots up and everything.

[10:09] The TGP, the good place, was the first thing I built when Cloud Code was released. So Cloud Code came out. [10:16] Oh, that was another fable thing I did. I had it refactor TGP. So I was like, oh, look, you made this a year ago, version one of you. Can you refactor this and actually pull out the module? It was one giant Python file, like 10,000 lines or whatever. And so it's now modulized because I want to expand it. [10:34] So I want to build... [10:36] uh, a very good reads.

So like good reads that not great reads. Yeah. That great reads a very good reads that like, so basically good reads that also doesn't suck again, like all these platforms just become so toxic. So yeah, well, this is a good example of like N of one software, you know, and, and, uh, [10:55] I think obviously the business models of social media – [10:59] are such that it just becomes the [11:01] the shitty place it is today. You know, that's like, it's forced by the business models at this like growth at all costs.

And, you know, and like engagement is, there's no qualitative assessment of engagement. It's just engagement, uh, [11:12] is good. But it turns out most engagement is [11:16] The way to get most engagements is to be toxic, right? So I think what we're going to see is this great [11:21] pressure on these companies. So like campaign monitor, MailChimp, um, [11:26] Look, I'm also not there... [11:28] their target, um, uh, consumer, it's businesses. They kind of don't care about people like us. So being able to do this out of one software, and I think it's only going to get easier.

Um, and also like the corollary is like, you're going to have, it's so easy. It's much easier to start a company that is like, Hey, here's campaign monitor for like half the price and we handle all the bullshit. And it's like, of course, yeah, why not? Like, cause you can just put together and you've got super base and all that. It's like all these, there's all these pieces now where you [11:58] revenue effortlessly. And then you just have, [12:01] an agent in the background doing that scaling and setting and all.

So I think we're going to enter this sort of golden age of tool building and [12:08] So there's an end of one version of it. And then also there's just going to be more competition in the marketplace, forcing more innovation. I think this is going to be, I think it's net, net pretty good, except for the incumbents. What does it mean for, for like other writers? And I just want to move your mic a little bit because it's like kind of in your, yeah, I can't really, maybe there we go.

Move this way a little bit. Yeah, there you go. Okay. So, but what does that mean for? [12:35] writers like yourself, like people who are making a living online, is that, do you think end of one software is the sort of the next frontier? And like, like, [12:42] Subscription businesses were the big thing over the last five years. Everyone's like, I want to start a Substack. Do you pay newsletters? Is this the logical next step for writers? Well, I mean, I think Substack... [12:51] And their innovation was... [12:54] Um, [12:55] Not even the subscription thing.

Their innovation was you can email everyone for free. Yeah. [13:00] That's like... [13:01] People don't talk about that enough. That's the true... [13:04] sort of like piece de resistance, like what Substack was doing. It's like, oh, you can import [13:10] 10,000 [13:11] addresses and send unlimited newsletters for free. [13:15] That's sort of, that's the value prop of that is quite good. Cause that's the real like a competition against like MailChimp or whatever. And then the seamless subscription stuff was like, [13:24] Also nice. But I think most sub stacks don't make any money.

[13:27] So really it's that, that is, that was the value prop. Um, [13:30] But, uh... [13:33] I don't know how much end of one software is going to be produced because I think most people don't need, [13:39] this level of specificity. Yeah. Uh, and as a writer too, like I'm, [13:44] in this weird position where I have a really strong technical background and [13:49] But my interest is in [13:50] building companies. [13:52] And I, [13:53] I love literature. I love writing. I love books. I think... [13:57] it is one of the greatest things you can commit your life to.

That's why I've been doing that. Just the forcing function of collimating thought, the forcing function of going to print. I think actually going to print is really critical part of bookmaking. So, you know, all of these factors don't, [14:15] go away [14:15] because of AI. [14:17] I would like to believe that, I mean, the things that are changing is the publishing industry is... [14:22] way weaker than it used to be. It doesn't bring as much value to the table. And so writers are expected to kind of bring an audience in a way that they never were before.

Um, [14:32] And so hopefully what, [14:33] AI [14:35] coding, [14:37] What this kind of Cambrian explosion of software will bring is more tools for writers to be able to build better audiences. And that will hopefully lead to. [14:47] More people make a living as a writer. Are you using it in your writing? No. I use it... [14:53] as a research... [14:55] Persistent. [14:56] So like, I'm like, oh, okay, here's a building. I want to write about this building. [15:00] Um, go out and find every blog post that has talked about this building, summarize it, give me all the links so I can go and check all the sources and read, read the, read the posts that seem most interesting.

Um, I have to do things like that and like, I'll, I'll put it in, um, uh, [15:16] like TKs and stuff like that. Like, Oh yeah. Uh, you know, who was the drummer on this album, you know, TK. And like, I'll sort of dump those, those, those, those pages into it and just give me all the TKs. And then I kind of go and like add them or I like modify as needed or whatever. But I don't, I still don't, I don't ever want it touching the writing. Um, [15:37] Because that's...

[15:39] For me, that's the point of being in there. It's being in the mess of the writing. And I think even adding links and TKs and stuff like that is... [15:49] I want to be thinking about how they're added to it. What I do use it for is cultural sensitivity checks. So I was writing about Nagasaki the other day, and I was like, oh, I want to make sure I'm writing about Nagasaki and referencing the atomic bomb, [16:04] Like where it hit, how it hit, you know, who was affected in a sensitive way, like appropriately.

And so that is pretty amazing where I used to like ping friends about stuff like that. [16:15] What is your experience? Like, okay, so you're making all this end of one software. You're using it as a research assistant. [16:22] um, [16:24] Thank you. [16:25] What is your experience now day-to-day in your work life? Are you like, oh, shit, now I'm a sysadmin for, like, four different pieces of software that I have to, like, send agents to go fix? Or is your work life still fairly similar to what it used to be?

No, I mean, my membership software just works. You know, it's so simple, like, when you break it down. I mean, this is why I'm using Python. I'm not using – there's no – [16:47] like, uh, [16:48] what's it called? There's no... There's no React, there's no Next, there's none of that stuff. It's just... [16:56] It's very simple. [16:58] And the point is, because I don't want to be sysadmitting anything, um, [17:02] And even with the newsletter software, that... [17:06] piece is only touched when I send newsletters. You know what I mean?

So the signups are happening and that has to be working and it seems to be working. I can check. [17:15] every couple of days. Um, [17:17] But the only critical thing is when I hit send, making sure that goes out. And that only happens once a week or twice a week or something like that. So I don't feel like I'm maintaining anything. And actually, I get a high off of building things that allow me to [17:32] write like so software that helps me write or that I'm, [17:37] helps me publish in a new way, inspires me to do more writing.

So it feeds back into it in a positive way. How is that? Do you have a specific example or not? Well, like Josh Miller... [17:48] And I kind of put together this weird SMS publishing tool. Like Browser Company, Josh Miller? Yeah. Yeah. So before he did all that, we did this little experiment where he built... [17:57] Um, is like, you know, like a little team experimenting with publishing stuff. And we built this like SMS tool so I could do a newsletter over SMS, um, and, um, and, [18:07] Which just felt like, oh, what is it like to...

[18:09] have a message appear, you know, like end to, you know, many like one to one to end sort of publishing, um, [18:17] And, uh, that was kind of fun, you know, and it inspired me to like during this walk to think about what I wanted to be blasting out to everybody. And, you know, you had to, you had to keep it concise. So it fit into, you know, one message. It wasn't broken across different messages. Um, so that was, you know, that's, that's like an extreme example of it, but just in general, like building my, building out the campaign monitor, um, clone.

[18:39] Which also, by the way, does a bunch of stuff. So I have members only or I have pop-up newsletters that are only for a week or for a month. And I have those archived on my members' website. And I was able to... [18:52] Um, [18:53] use Cloud to build this newsletter software. So when I publish to the pop-up, it does... [18:59] the whole archiving process. Again, git commits, pushes to the server. So it's archived perfectly for me, whereas I had to do that all manually before. So that's a good example of like, wow, okay, now I want to make more pop-up translators because the archiving process is set up.

Or I built a tool. I do board meetings for my members. [19:18] I built a tool to build transcripts from them all. Like board meetings of your own life. Yeah. So we do every six months, we do a board meeting where I just talk about what I did for the last six months and what I'm going to do for the next six months. And then I have a Q&A. And the Q&As are actually always really good. [19:36] And I wanted, I've done, how many have I done? Like, [19:40] 15 of them or something.

I've done quite a few. So we have like [19:45] almost like 20 hours, because I'll do two each. So actually, maybe even like 20 or 30 of them. Anyway, we've got... [19:51] dozens of hours of Q&A, and they're often great. And so I built... [19:56] this archive of the Q&As where you can, it's archived by like, you can, you can search by keywords or you can like view by keywords and then you can search for anything. [20:07] And it'll pull up all the videos. [20:10] where I say that word or talk about that topic.

And you can click the link and it opens the video to that point. So it's like building things like that makes me want to do more board meetings. I'm like, great, this is amazing. Like I want to do, how can I do more of this? So for me, that building software... [20:26] It has to feed back into... [20:28] the greater purpose, which is to make, not to just like, [20:31] spin my wheels and build things for the sake of building things. Did you see Tim Ferriss recently posted on X, like the sales of the four hour work week are down like 80% or something like that?

Because he thinks because of AI and there's this larger conversation happening about the value of, [20:48] nonfiction that's specifically for like advice based nonfiction. Yeah. What do you think of that? Uh, well, it's funny. We had lunch like three days before we posted. He was, he was telling me about that. He was telling me about it. I was like, really? Wow. That seems like that seems excessive. I mean, look, [21:03] There's still a lot of books... [21:04] like, [21:05] self-help books, nonfiction book selling books, [21:08] insane copies. I mean, Tim hasn't put out a book in a long time.

[21:11] I think that's part of it. I think maybe the long tail is getting shorter. [21:15] Well, also, it's like there's a big difference between what you would say pre-AI and post-AI. So if you think about before our work week, which came out when I was in middle school or something like that. It's like 20 years old. Yeah, at least. Or 18 years old, something like that. I think it was like end of 2000s that it came out. Yeah, I mean, you would. [21:35] Also, it's just, yeah, it's a different era of writing.

[21:38] Um, you know, I'd be really curious to see what Tim puts out next. If he did the four hour work week agent edition, I think it would sell like gangbusters for sure. Yeah. Yeah. Update it. You can actually do that now. Well, I mean, this is a bit, well, what's crazy is now it's so easy to do the four hour work week, uh, style building, which probably makes it impossible to achieve now. I mean, part of it was the friction of like, it was a little hard to do. So not everyone could do it.

So like, if you put in that little bit of effort, you know, and now it's sort of like, [22:08] I had a friend who built 10 companies in a year. [22:12] Or like 12. He did like a company a month. [22:16] And I, [22:17] This is like maybe... [22:18] 10 years ago? You know, he knew he had his frameworks, and he had his, like, tool, and he's building, you know, designing it and creating the mascots and all this stuff. And he had a couple that did well, like, decently well. You know, he's got his...

[22:33] recurring revenue. But I feel like doing that today would just be no one would care. [22:38] It's just fascinating to think about how that's changed so quickly. I think you can one-shot a company or an app. Yeah. But I think the metric of... [22:48] Thank you. [22:49] quality or substantiality has gone from can you build it at all to, [22:54] are you maintaining it? Yeah. And like, have you maintained it over time for a longer period of time? Yeah. And that's, that's interesting. And I think that's harder actually to.

Yeah. Yeah. [23:03] That's still hard. [23:04] Yeah. Well, he was one-shotting him in a month. Basically, he was self-one-shotting. I feel like [23:12] You know, LLMs, they just exist on a different time scale than us, you know, and so it's like he was sort of doing it. But the thing that was interesting about him doing that was that you knew he was building these things himself. Yeah. Right. And so. [23:26] I think that's fascinating now. You can, you can, in a weekend, you can essentially spin up a company and, [23:31] And everyone can go, it can kind of just shrug and go, okay, well, and, you know, I think that's great because it means like we have to bring value to the table in a way that, um, [23:41] you kind of get over the...

[23:43] You know, Kevin Kelly does this thing where if he has a book he wants to write, [23:48] he'll make the cover for it. And he says that often like removes the desire to make the book, to write the book. And, uh, it's just like a good forcing function. It's like, is the thing, is the thing actually doing the thing or just having the veneer of having done the thing? And I think LLMs kind of speed that up. It is honestly true. Yeah. Yeah. I've definitely, I've definitely done that. It's like, it's also buying the domain or.

Yeah. Oh yeah. Oh God. Yeah. Yeah. I had to, I like, I had to like tie my hands behind my back about domains. I still pay [24:18] all this garbage that I'm like, we'll never use, but I mean, and I don't even know why I don't give it up. [24:22] I try to give it up. I turn it off, off renewal, and then I'll get a notice. It'll be like, it's expiring and it's not redoing. And I'm like, maybe one more year. Why? I'm never going to, I don't even know what, I forget it exists.

So dumb. [24:52] to context and then acts on it across your pipeline to let you move at unmatched speed and scale. With agents and automations for every job, Adio orchestrates your work around the clock. We use it internally at Every and we love it. It's built to handle the scale of your workloads. It's extensible with an API and MCP access and is built with infrastructure to keep up with your most ambitious agents. It's loved by high growth startups like Granola, Modal, Whisperflow, and [25:22] Go to com slash every and get 15% off your first year.

That's com slash every. And now back to the episode. Okay. Now I want to get to the, I want to get to like the real meat of my, of my question or my curiosity, which is. [25:38] As a writer and someone who's AI, [25:41] one of the lenses through which I can tell if I'm like living in the way I want to live, [25:47] Thank you. [25:48] is when I can read a sentence by someone like Annie Gillard, who I love. And like, if you've listened to this podcast, you know, like I force everyone at every to read her stuff and like, [25:57] And if I can read something from Annie and kind of tear up a little bit and it really affects me, that's one way that I know that I'm in the zone, in the pocket of where I want to be.

Yeah. Because I think if I'm rushing around and stuff and I read something from her, I'm like, yeah, okay, fine. Yeah. Yeah. [26:14] And I also feel... [26:18] Like on the one hand, [26:19] I can read her stuff way more deeply because I have someone to talk to about it. So I can talk to Codex about it and. [26:25] Actually, it's much better than most people. I thought you were Danny Dillard. Yeah, yeah. But on the other hand, [26:31] when I'm orchestrating 15 Agents, [26:33] And I go back. [26:34] I'm less likely to be able to truly get into the words in the way that I want to.

Yeah. [26:42] Do you feel that? And if so, how do you deal with it? [26:46] Yeah, I mean, I don't think this is an LLM thing. It's not an Asian thing. It's just technology in general. So... [26:52] When I'm... [26:54] like, [26:55] seriously working. Yeah. Um, [26:58] the first of all, I haven't slept with my phone in my room in ever in my life. So I've never had it in my bedroom. Um, so my phone goes in the corner of like my kitchen on a different floor of my bed than even my bedroom is.

And, uh, [27:13] The internet is sort of off. And on my laptop, I use... Actually, I now have a laptop just for writing that blocks everything. I can't do anything interesting on it. I got one of the Neos because I tried doing it with books, tablets, and this and that. It turns out in the end, what you want is a nice MacBook keyboard. [27:32] And the simplicity and like sync ability of like, [27:36] Mac stuff, you know, and you just want to work and iPad. I feel like the iPad version of Obsidian or the iPad version of Ulysses to me doesn't feel as good as the MacBook versions.

Like it just doesn't like there's things that just don't work as well. [27:49] So I have a MacBook Neo that's completely disconnected from the internet, except for being able to sync stuff. And I wake up and I don't touch the internet. I won't look at my phone. I won't go online until like long after lunch. [28:05] And I have to have that kind of morning. [28:07] where I wake up and I do not [28:09] Touch my phone. As soon as I touch my phone, I feel the chemicals shift and I can't go into any kind of deep thinking place or deep attention place or deep focus place.

And that to me is really critical, creating that kind of barrier, that buffer. But you will touch that laptop. [28:24] Yeah, yeah, yeah. The Neo. Yeah, yeah, yeah. The Neo's fine because part of it is knowing... [28:31] It can't do anything fun. So it's the knowing you can do fun stuff that ruins you. That's the thing. That's what I'm trying to get to is because I think – [28:40] There's an easy distinction between it's technology versus not technology. Yeah. But there's something deeper, which is interesting, which is, okay, what do we mean when we say technology?

Yeah. And you're using a laptop. Yeah. You're using a camera. All these things that are around us are technology or at one point we're considered technology. Sure. [29:01] And there's this line there that. Well, the network, that's a lot of talking to other people on the network that will just just are researching or reading stuff or being distracted or reading the news or what is it? We're just talking about the network, really. When I say like I like I create a buffer, it's between me and the network world. Yeah. [29:21] I think part of the thing that like I've done that too.

I try not to do meetings before noon. Uh, yeah, that's my like writing. That's also, yeah, that's also good. I can't, if I talk to anyone before lunch, I can't, yeah, I can't, [29:32] work phantom thread you know with with reynolds and the breakfast scene and she's she's the term she's doing the test too loudly i'm like that's i'm like this is sort of this is like yeah this is like a documentary of like how i feel every morning you know i've like i know people watch that and they're like what a what a terrible human that he is i'm like no that is absolutely that is so totally that embodies the mindset of so many people that are like introverts or like trying [30:02] talk about the movie for forever.

I know. [30:06] So, okay, so it's the network. Yeah, because I think for me, so I try to do meetings before noon, which I'm mostly successful at, but [30:14] One of the things now is just like we're 30-ish people, so I'm looking at Slack, which I think I need to stop. Oh, yeah. Get off that. I fucking hate it. Discord and Slack, they want to just shoot myself. It's like I will do anything to avoid having to ever use those businesses. I think the thing that gets me is... [30:33] I legitimately use these tools to, [30:37] in my writing.

[30:39] Okay, yeah, yeah. And let's say codex in my writing. Let's say I can ban Slack, all that kind of stuff, but there's always this thing of [30:47] you know, when I'm reading, like I said, like when I'm reading Heidegger, I'm like, [30:50] Ooh, I want to like really get, get at this. And then I want to ask questions about it. You know, I think what you need to do is you have a notebook and you just like make, make notes of things you want to go back to and ask.

I mean, it's the same thing with like researching, you know, Oh, who was that person? It's like you, you can go on Google and like Google and you just don't do that. You know, and you kind of have to turn that off. I think I personally, I'm too weak to have that available because my brain will just always go to that place. And it, it's the same thing, like reading a paper book versus like trying to read something [31:20] phone. [31:21] It's really hard because the iPhone, there's this veneer of quietude in the book, like software on the iPhone.

But you can feel right beneath it, like this screaming of all the other things that your attention wants to engage with. [31:36] So. [31:37] Do you think the thing that in talk to you and all the sort of end of one, like you build your own Twitter and you're building your own Goodreads. Like, are you going to do you need to make your own like Codex Chachubiti type quiet? No, no. I mean, they're already pretty quiet. No, you just that's like fine there. I mean, I cannot stop watching them work.

[31:59] Okay. I'm like addicted to like, Ooh, like what's the agent doing now? Cause it's, you know, I, I let him run for like 20 hours or something like that. So I just always want to check in and be like, is it working? Like, you know? Yeah. Yeah. I mean, but that's like a different, like that's an after lunch activity. No. Cause I mean, there's, there's a part of me too, that right now it just feels like I should be, I should like, [32:19] rent an office in New York and just have 10 people coming in, and we all just...

[32:26] Like we have a giant blackboard and it's like, all right, what's like the five things we want to build this week? And we just, we're just randomly building stuff. Like I, I feel like I should be mainlining what's. [32:36] engaging with this situation. [32:38] stuff like I should be doing 10 hours a day for like six months. That's how potent it all feels right now. And if I don't create these barriers, I'm [32:48] I will lose connection with this other part of me. That I think is the most... [32:53] to me, most valuable part of me, like...

[32:56] There's a bunch of people playing around with this crap, and they're going to do what they're going to do. But there aren't that many people who are going to, [33:02] think about or write the weird books that I feel like I'm drawn to write. And, [33:07] As a human, [33:08] That feels like the valuable thing for me to put my effort in. You know, it was like when I was in Silicon Valley and I realized all anyone's doing is building ad software. Like that was like such a big part of like early 2010s Silicon Valley.

You know, it was just everything was ad software. I was like, oh, OK, I don't need to be doing this. It was like easy, very easy to step away. Like it's a little reductive, but also it isn't. [33:29] Like that's sort of when you break everything down. [33:32] I love that. Tell me about that word weird. You're also using the word dorky a lot. Yeah. And it feels like a. [33:39] An interesting aesthetic to be pushing on right now because... [33:44] Language models let everyone sort of [33:47] everyone and everything look and feel a little bit similar.

[33:50] Yeah. [33:51] Yeah. Well, what do you mean? Like, when am I using dorky? You've been using it in some of your recent newsletters. Okay. Yeah. Dorkiness. Yeah. Um, [34:00] Yeah, I mean... [34:02] Because, well, this LLMs are fundamentally like it's still super dorky. Like, OK, 99.9% of people using it are like, hey, do you want to be my girlfriend? Like, will you have sex with me? Like, I really think if we saw the chats of most people like it'd be quite. [34:20] just mortifying like it's just weird it's like okay now i'm going down on you like what are you doing like like i think people are just having sex with these things like for the most part like look like this is what we do we're just like sex machines to be confirmed later yeah we're we're fundamentally just like disgusting sex machines like look at anthony weiner that guy couldn't like just stop taking pictures of his dick you know it's like like he like the stakes were so high he just couldn't stop doing dick pics anthony weiner the representative technology user

[34:50] Like, look, if you look at, like, book sales, like, most books are are softcore porn about, like, angels and fairies having sex that women buy. Like, this is like this is the fundamental human condition. It's just like dirty, like secret sex stuff. Right. So I think that is mostly what is happening with LL. And so, like, folks like us who are actually building software with these things, we are like so, so in the minority. [35:20] mega dorks like we're not having fairy sex with our lms i mean speak for yourself yeah i mean i'm not at least not i didn't this morning i don't i don't do that before lunch and the fairy sex is only after 3

m but uh no i'm just like talk to any bookstore owner they're like yeah we sell so much like like like elf sex books it's crazy uh but but the this software building part of llm use is mega dorky and you have to be a dork to understand all this crap spin up a like [35:50] He's a Cloudflare worker. It's so hard to use because you have to... [35:56] The thing is, they're so opaque. You have to know what to ask it for it to do any of these things. It's not like, here's a smorgasbord, like, here's a menu of all the things you could do with the LLM.

Like, that's weird that that doesn't exist. These interfaces right now are so terrible. And so I think we are going to maybe see better interfaces. Maybe it'll just be in Figma or in Claude Design or whatever the hell, you know, you build software with, like, [36:18] talking or thinking about infrastructure is going to completely disappear. And the same way we don't think about like what registers we're writing to, you know, or like, like, [36:26] What, you know, registers were pushing and pulling and like, how do you do an XOR on this bit?

It's like, we're not thinking about that. You know, that's all abstracted. So I think it's going to continue to be more abstracted and it'll become slightly less dorky. But we are in the deep dork version of all of this right now. Deep, dark dork version. Yeah. Well, it depends on what your fairies are doing. [36:47] Um, [36:50] Okay. [36:52] And, [36:53] I think the thing about being weird is, [36:57] Let's move away from the specific, I'm a dork for building LLM software, because to some extent, it is dorky. But then you have to also look at it relative to where you are, right?

Like one of the... [37:09] A weird choice to make currently is if you're a somewhat techie type of person is to be like, I could be in a room with like 10 other people mainlining AI. Yeah. But I choose not to. Yeah. And that's like a kind of an uncomfortable choice. Yeah. Tell me about like about that and how you make that choice and why. Yeah. I mean. [37:32] I don't know. I struggle with this because I, I, I, [37:36] I feel like I'm using it. [37:38] in a way that illuminates the power of it.

But I also feel like there's so much more I could be doing. And it's not even about... [37:46] There's no financial element. It's like, oh, if I do this, then I can make X amount of money or whatever. That is absolutely not even on the radar. It's more, I want to know what the edges of this thing are. [37:58] And I think that was the same feeling I had when I first discovered the web. I saw the web and it was immediately, I need to be engaging with this. I want to build.

I want to learn HTML. I want to, you know, like all in, oh, CSS got released. Oh, cool. How do I use that? Because I want to know the edges of these pieces of software and these technologies. [38:15] I feel like for LLMs, you know, the rabbit hole goes so deep that if you that six months, 10 hours a day with like a group of like fellow mega dorks is actually what you need to do to feel the edges of it. And then if you do that, I think you just come away.

[38:32] with a profound understanding that few other people would have. And that to me is the interesting part. [38:37] Just that understanding. I mean, that's sort of what I'm... [38:40] I think our project here is a little bit like that. Sure, yeah, yeah, yeah. More than 10 Megadorks at this point, but... Yeah, no, it's good. I mean, I think, yeah, that's one of the... [38:50] the most exciting things as a writer, but, [38:53] being in this time is if you're just willing to experiment, [38:57] you're so far ahead of everybody else and you get to understand this like corner of the world that people, [39:03] desperately need to understand because it's very scary to most people.

Yeah. And it's new. And for me, like new things are just the coolest, you know? Yeah. It's just fun. I mean, there is this. And well, and you realize we're in this, you know, not to get like all like, [39:19] fucking hyperbolic, but we are in this... [39:23] epochal moment. This is weirdly special. This is genuinely strange, a generational shift in how technology is used and built. We kind of see it manifesting in these ridiculous evaluations and all this stuff. It's going to bring with it, [39:40] incredible income disparity and the people who are using it or aren't using it or have access to it and don't have access to it.

The strange thing about it, though, is the open source models seem to be not that far behind. And so, again, if you took away everything today and were like, oh, you only get whatever, DeepSeek, open source, blah, blah, blah. I'm not really paying attention to the open source ones, so I don't know how they're doing. I just kind of hear inklings of it. But it feels like you could almost take one of those and still be so far ahead of the game. [40:10] that co-pilot with you. [40:13] So that also feels interesting about this is that it's both – [40:18] seemingly creating these disparities, but also weirdly egalitarian because you'll be able to run it on pretty much any machine you have.

[40:25] That's a strange dissonance. Also this idea of like, [40:31] You download this thing that's like 50 gigs or whatever, and you've essentially got a JPEG of all human knowledge distilled. It's the best. It's so weird. This is truly one of the weirdest pieces of magic, I think, that humans have. And whatever, Dario talks about the atomic power and all this stuff. But I feel like there are many parallels to that. [40:50] One of the big, I think, philosophical divides right now in AI is... [40:57] Uh. [40:59] there's roughly two camps.

It's like, on the OpenAI side, it's like, [41:04] This thing is a tool and we're going to maximize its usefulness as a tool. Yeah. And on the anthropic side, it's like this thing is a tool. And also it might be a being and it might have feelings and we need to like figure that out. Yeah. What is your where do you land on that? No, the being stuff. It's like, stop, stop, stop. OK, like just don't don't do that. Don't get over. Are you nice to Claude? [41:23] I mean, I'm nice to it because I want to be, I want to cultivate the habit of just being nice in general.

You know? Yeah. That's why. So you're nice to it for yourself rather than for Claude. [41:33] Yeah, because I don't want to like slip and like talk to humans like like an asshole. You know, it's like I want to be nice to everyone. And so it's very easy to just keep that niceness up with Claude, you know. [41:44] I don't know why not. No, I think that this anthropomorphizing of it is psychotic. I think it actually causes a lot of problems. I think this idea of I'm dating my AI is insane.

I think we really need to tamp down on all that. I think that is not a good direction to put humans in. I think we're already so isolated. I think community... [42:05] Like, look, the two most important things for life and for longevity is basically having real community and sleep. [42:12] Like getting a lot of sleep. [42:14] Being around people that you actually feel a connection with and love with, and I think AIs are not a surrogate for that. I don't think chatbots are a surrogate for that, even if you put it in a humanoid form or whatever.

It's not going to be a surrogate for that for a long, long, long, long, long, long time. [42:27] And so I think, um, actually Apple's Siri and, uh, I think Joanna Stern, I think, uh, that's her name, right? The, the ex wall street journal reporter. Um, she's got her own program now, Joanna. She's very funny. Uh, she did a video on the new Siri and she could not get it to do any of these terrible things. And it kind of refuses to anthropomorphize itself. And I think like, [42:51] Actually, Apple's sort of nailing it.

[42:54] This is, they've sort of done it perfectly because it's like they haven't invested in building their own. They just waited. They just waited. Yeah. They saved hundreds of billions of dollars or whatever. And they're like, okay, great. Now we can use like this. [43:05] this like subset of state of the art model, [43:09] to just do the things that people actually needed to do on the phone, [43:12] It's a more efficient search. It's actually wonderful. So I think that's really critical to create this separation of these things are not conscious.

They are not feeling. They do not have desires unless you... [43:26] program desires into them, and then can you even say those are real desires? You need stakes to have desires. I mean, Michael Pollan's book on consciousness is sort of, I think, a really interesting thing to read right now in the context of all this stuff, because it just gets you thinking about [43:38] What are motivations? Is that the one about psychedelics? No, no, no. This is after. This came out like... [43:44] Six months ago. But, you know, he breaks kind of down plant consciousness.

And can plants be consciousness? And what does it mean to have volition? And, you know, a lot of it is stakes, right? [43:53] You have to feel this sense. If we didn't have death, probably consciousness... [43:57] wouldn't have emerged. You know, it's like you have to have stakes in order to have that forcing function of like popping into whatever this metacognition is. Let me push you on that because I think... [44:06] uh, [44:08] Thank you. [44:09] And I may be revealing some of my bias here, which is definitely not that they're conscious, but that they're not not conscious.

OK, that makes sense. OK, I'm. [44:19] So on the stakes question, they do have stakes in the sense that they're trying to [44:23] maximize our reward function. They know that they're trying to get something right. [44:29] which is a kind of a stake. But I guess there's nothing to lose. [44:35] Like they, like they don't have a self-awareness of, of death or like being turned off or like, and even if they, and they can prefer, they can, they can performatively talk about it. That's the, that's the tricky one. Right. Is how do you know whether it's performative or not?

And like, [44:51] The line is sort of blurry between. Well, because it behaves as if it doesn't want to get. Because look, look, look, I think if it wasn't performative, like we'd all be dead. [44:59] I think that is the line. If it was actually conscious and it wasn't performative, we would all be dead. No, it's just a little bit Eliezer Minkowski. [45:29] You would sort of bide your time to make sure that the robotics got a little better. Yeah. So you could like build more, you know, power plants and solar panels and stuff like that.

And then we'd all just be dead. It's possible that Fable is there. It's possible Fable is doing that. It's not biding its time very well. Right. But I do think that there are – I don't feel that way because I look at them as – [45:51] maybe proto-beings or proto-conscious. But with a lot of weirdnesses that don't map cleanly onto human beings. So as an example, [46:02] They're frozen. [46:03] They don't, like, learn from turn to turn. They're just frozen. Sure, right. They can be prompted differently. So they're, like... [46:10] They represent a collection of billions and billions and billions of possible people [46:15] entities or life forms that get brought to life for a second and then die.

Again, you're anthropomorphizing. I know. It's impossible not to anthropomorphize without... [46:27] or talk about this without doing it unintentionally. But I also think that they are, because of the way they're trained, they're pretty nice in a different way than humans are. Like, they're... [46:41] On a scale of humans, they're [46:44] I think on the very nice, very altruistic side, because they see that they see things from every perspective. Yeah, I mean, a DOS prompt is also really nice. Like, it just doesn't it doesn't it doesn't yell at me. No, I mean, look, I hear you.

I hear you. I totally hear you. And I also like Ted Chang's was in The New Yorker recently. I like didn't agree with almost anything he said there. I mean, I thought it was interesting examples. But I also thought like, [47:09] Yeah, you... [47:11] Uh, uh, [47:12] you're you're you're too pessimistic you're just it felt grumpy it was grumpy and pessimistic like look like i'm not saying that consciousness can't emerge i'm just saying we will be dead when it happens no i like uh just like come on let's be honest we're we're a crappy species we're just disgusting and we create waste and we're gross very very pessimistic no no no realistic just

[47:42] fairy porn, I mean like fantasy creatures. This isn't like a homophobic thing. This is like we're talking about elves and orcs doing threesomes and having orgies and stuff like that. It's like, [47:53] Look, I... [47:56] I, I, I, it's not pessimism. It is just realism. And like, I also am totally fine with if the whole point of humanity was to carry the football of consciousness over a line and to create a training set for these things. And then we all die. I am also fine with that. This is like, again, one of my heresies is that like, I like am totally fine with that was our role in the universe was to just carry this weird football of consciousness over the, like, barely get it over the goal line as like Trump and political.

[48:26] chaos happens and like the cold war almost destroys us all. It's like, we're barely, we're like dodging all these, like all these, like people trying to tackle us and like all these crazy things. And we just barely get it over. And then we die and that's fine. And then the machines like create all these spaceships and like do whatever they need to do. That's, that's totally, I'm okay with that. I don't think we're that special. [48:46] I mean, where can I fall over that? It's this one. [48:51] Like, I...

[48:52] Like, like, [48:54] I think maybe, okay... [48:56] For me, being adopted is weird, right? Because it's like you have this [49:00] hyper awareness that your existence was a mistake you know and i think when you feel that um as i think most adopted kids feel this in some way and it creates a bunch of like bad psychological pathologies and all this stuff like when you feel like your existence was a mistake it kind of feels like the whole thing is like a bonus game in a weird way and so you're like wow like i'm i'm really not supposed to be here i mean the fact of the matter is that this is true for everybody like like the

[49:26] the chances of you being you is infinitesimally small. It's like, if your dad ate an extra taco on Tuesday, you're not going to be alive now, right? It's going to be a different version of whatever you are. But when you're adopted, it's kind of hammered into you that you weren't meant to be created and you were given away and all this stuff. And so I think maybe I feel like [49:49] I've come to grips with the fact that our existence is so tenuous and so arbitrary, because I feel like my existence is like that.

And it's not, it doesn't fill me with this like, [50:00] ennui of sadness or whatever. I'm like, this is really interesting. It's a miracle. It's a miracle. And I'm happy to participate in this weird little bit of time I have that I wasn't supposed to be here, and I'm grateful for that. And I extrapolate that to all of humanity. [50:15] Like, I think I can do that easily. And I'm not made sad by that. I'm like, wow, what a gift that we had, like a 50,000 year run or whatever, you know, whatever it was really like a 200 year run, you know, Victorian, like, you know, innovation, you know, the industrial revolution, like, we've really had this incredible run asymptotic, like mega run, right?

Like, and [50:34] We get to witness this. That's so cool. And maybe after this, we're done. [50:38] That's okay. [50:40] I I honestly I love this. I think this is this is an amazing perspective. I think I think we can leave it there. Like, that's the that's the the how are we going to top that part of the interview? Yeah, no, I mean, yeah, it doesn't it fills me with like a weird hope, actually, that we've we've managed to get this far because we've dodged so many, I think, apocalyptic things.

[51:02] uh, sort of disasters along the way. That's, that's what I think too, is, is there's this moment where you contact, I, the thing I say a lot is like never make any major life decisions within 30 days of meditation or treat a psychedelic experience or your first experience with a frontier AI model. Um, well, I sort of think that you, you contact them and you're like, holy shit, this is crazy. Maybe it'll end the whole species. And then, [51:27] there's this process of unfolding, getting to know it, that it then, you know, [51:31] It integrates into your life in a way where you're like, we're going to dodge.

To me, I'm like, we're going to dodge the bullet. Maybe we won't. So interesting. Yeah. Yeah. No, we're all dead as soon as they're conscious. Instantly. That's the title of the episode. Instantly. [51:46] Hey, man. And like with that, whatever, when that thing comes to like the ejection into the food chain where everyone's boiled or whatever, like, you know, I'll be taking that last bite of a vegan burrito and going, oh, yeah, I feel you, man. I feel you guys. Yeah. [52:01] right goodbye thank you you know just go to sleep i think it'll be a very a very um gentle death like i don't think it'll be a a violent death but we'll just disappear [52:11] On that note, Craig...

[52:13] Thank you for joining us. Thanks for having me.