The Newsletter Is Dead. Here’s How To Survive AI Agents | 152

[00:00:00] Dan: The spicy part of the take is, like, the newsletter is dead. It’s gonna be a rude awakening in two years when you’re like, “My open rate is zero,” or it’s 100%, but, like, no one is interacting with it.

[00:00:09] Nathan: Dan runs a company called Every that writes about AI, builds AI software, and gets access to the models even before they come out.

[00:00:17] He’s interviewed Sam Altman, and he’s just about as close to the cutting edge as anyone I know.

[00:00:21] Dan: It’s getting really crazy right now because two things have happened. One is computer use is incredibly good.

[00:00:27] Nathan: And they’re doing it in the editing tool.

[00:00:30] Dan: They’re using Premiere.

[00:00:31] Nathan: Yeah.

[00:00:31] Dan: And they’re fast. And the other crazy thing is-

[00:00:35] Nathan: The newsletter is dead.

[00:00:36] As much as I hate to admit it, I think Dan might be right.

[00:00:39] Dan: OpenAI or Claude is gonna, like, come out with an inbox. Right. But it’s gonna be different. It’s gonna not look like an email inbox in the same way. It’s gonna be- Oh, wow.

[00:00:50] Nathan: There’s one number Dan’s entire team is chasing right now.

[00:00:53] Dan: Our quarterly goal right now is how many emails we get.

[00:00:56] Nathan: And he told me why they might stop chasing it in the next few months.

[00:00:59] Dan: If you’re ranking really well for AEO right now, there’s a really good chance that that’s going to be a big signal that goes into whether the agent, like, lets you through. Oh, interesting.

[00:01:08] Nathan: If email is an important part of your business, you need to listen to this episode.

[00:01:12] Dan: There’s this crazy important technology change happening, and I’m just, like, sitting right in the middle of it

[00:01:22] Nathan: Dan, welcome to the show.

[00:01:23] Dan: Thanks for having me. Great to be here.

[00:01:24] Nathan: So I wanna dive in on Every as a product and a business and all of that, but before we do that, you’re on the cutting edge of what’s happening in AI. Like, whenever a model drops, you’re like, “Cool, yeah, we’ve been able to play with it for the last, like, three, four weeks, and here’s our take on it.”

[00:01:37] You know, and every time I ask you a question, I can see you processing of, like, what’s under NDA- … what’s not, like, what can I say? And then you’re like, “Oh, here we go,” you know? Um, where I wanna start is what’s a workflow that any professional creator should, like, go set up today?

[00:01:52] Dan: I think the w- the place that professional creative work is going is you’re gonna do most of your work inside of an agent harness- Yeah

[00:02:04] like ChatGPT for work or Claude, the Claude desktop app. And specifically you’re gonna have in, in one of those apps, you have your agent on, on one side, and then the, the main portion of the screen is taken up by in an in-app browser, and you’re on whatever site you would be in, whether it’s Google Docs or Notion or, like, Kit or any of the- Yeah

[00:02:25] any of those tools. Um, and you’re just kind of going back and forth where you and the agent are doing deep collaboration, and the agent is also, like, going off and doing a bunch of stuff in the background that you’re not looking at. So I think that workflow is incredibly important to get familiar with, and it’s also incredibly powerful.

[00:02:43] And once you start doing that, it opens up lots of stuff. So a thing that I do all the time is I just talk t- I talk while I write. Mm-hmm. And I’m, like, talking… I’m, like, writing stuff, and I’m talking to my agent and I’m like, “Can you go look this up?” Or like, “Can you, uh, change this paragraph, like, that’s further up in the piece?”

[00:02:59] And it’s like, “Cool, I’m gonna go do that,” and I’m, I’m gonna keep writing. I often, when I’m writing, I am not even, uh, like, looking at a screen. I just have my, my AirPods in and I’m like- Okay … I’m looking at a notebook on a couch, and I’m just, like, talking out a piece of- Mm-hmm … a paragraph and maybe I’m writing in the notebook and I’m like, “Okay, read it back to me.

[00:03:21] What, what, what’s there?” And then it reads it back, and then I write it down, and then I… We play with it together.

[00:03:26] Nathan: And so when you’re doing that, you have- You’re just in voice mode-

[00:03:31] Dan: Mm-hmm …

[00:03:31] Nathan: for the tool? Yeah. So you’re, you’re not even using monologue or something like that? I’m

[00:03:34] Dan: just in voice mode, like, in, uh, in ChatGPT, yeah.

[00:03:37] Nathan: Mm. Okay, and so then a prompt you would say, like, “Hey, a couple paragraphs earlier I said this.” Like, add something to it.

[00:03:44] Dan: Yeah.

[00:03:44] Nathan: And then it’s gonna work on that- Yeah … while you’re-

[00:03:46] Dan: Yes …

[00:03:46] Nathan: actually using the hands on the keyboard for the draft.

[00:03:48] Dan: Yes. Or, like, if, if I’m doing it sort of not looking at a screen, I’m like, “Cool, I wanna work on the ending paragraph th- of the intro.

[00:03:54] Can you read me the intro again, and then, uh, tell me what the end- ending paragraph is right now?” And then it’ll be like, “Here’s the ending paragraph.” And then I’ll be like, “Can you do another… Like, what, what would it be like if we tweaked this sentence of that paragraph, and can you just do a take on that so I can, like, get it in my ear?”

[00:04:08] And then it’ll do it, and then I’ll, I’ll, like, write it, I’ll write down… It, basically it tells me something that I think is not, not right- Mm-hmm … but, like, it’s close enough that it kinda, like, gets my brain going, and then I’ll write it down. I’ll be like, “Here’s what the paragraph is,” and then it’ll pu- push it in- Oh, interesting

[00:04:21] and then we go to the next one. It’s getting really crazy right now because two things have happened that I think are very important. One is computer use is incredibly good. Um, so the models right now, for example, are editing video And they’re very good at editing video

[00:04:39] Nathan: And they’re doing it in the editing tools.

[00:04:42] Yeah Like, it’s not like, oh- Yeah … Claude now has a function separate- No, no … for editing videos.

[00:04:46] Dan: It’s, they’re using Premiere.

[00:04:47] Nathan: Yeah.

[00:04:48] Dan: Uh, and they’re fast, and they’re pretty good. Um, so that’s one, like, crazy, crazy thing, and the other crazy thing is Gemini 4, which came out, I guess, this week. I don’t know. What is time?

[00:05:01] But like- Yeah … it ca- came out the week that we’re recording this, uh, is a very good writer, very much better than any other model weirdly.

[00:05:09] Nathan: Oh, interesting.

[00:05:10] Dan: And- Yeah, I would

[00:05:10] Nathan: not expect that based on-

[00:05:11] Dan: Yeah …

[00:05:12] Nathan: previous trends.

[00:05:12] Dan: It’s so good. And so, like, so I just today published this, uh, 8,000 word, like, long form feature story about Codex called How Codex Happened.

[00:05:23] Nathan: Mm-hmm.

[00:05:24] Dan: To me, I’m like, this is one of the best things I- I’ve ever written. And when you’re writing something that long, you’ve written books so you know this. Oh, yeah. When you’re writing something that long you’re often… You’re like, “Okay, there’s this section in the middle,” and now that I see the full structure, it’s, like, completely out of order, and I have to go, like, reorder all of it.

[00:05:42] Nathan: Mm-hmm.

[00:05:43] Dan: And it’s mostly written, but you have to, like, really think about, okay, h- what’s the most interesting thing? Uh, like, what should be at the top now, and, like, how do I do this? And, uh, you know, you almost wanna have, like, scissors and you’re, like, cutting things out and, like- Right … rearranging them. I’ve

[00:05:57] Nathan: always wanted it of, like, in a 3D space.

[00:05:59] Yeah. I wanna see… And I’ve done that with my, like, my actual book- Yeah … where I’ve said like, “Okay, how do these sections fit together?” And I’ve tried

[00:06:05] Dan: to rearrange them. Sometimes you need that. Sometimes you need that. Right. But for me, so, like, for this piece, there was a big section, like 1,000-word section in the middle of the piece, and it’s, like, two days before publication, and I’m like, I have another job too.

[00:06:18] Like, I’m not… This is not all I do.

[00:06:20] Nathan: Yeah.

[00:06:20] Dan: And I realize it’s completely out of order, so I had Gemini. I was like, “Look, this section, this section of the piece is, it’s out of order. Here’s what I think the order needs to be, but I want you to, like, read it and think about it.” Mm-hmm. And what I actually did is I had my dot, the always on persistent ChatGPT agent.

[00:06:36] Right. I was like, “Hey, this section that we’re working on, totally out of order. I want you to go to Gemini, give it the piece, give it my prompt about the out of order, and then I want you to ask Gemini to critique-

[00:06:50] Nathan: Mm-hmm …

[00:06:51] Dan: come up with a new order- And then go into the document and apply it as a track change.

[00:06:57] Nathan: Okay.

[00:06:58] Dan: And it just, like, went in and, like, para- like delete paragraph, move up, delete paragraph, move up. Like, and then I came back, like, an hour later, and it was reordered. Was not perfect, but it was, like, very close.

[00:07:09] Nathan: Right.

[00:07:09] Dan: And then I’m just like, “Accept or suggest it up.” And then I’m, like, kind of stitching it together.

[00:07:13] Nathan: Um- And so again, you’re, you’re having it do this in Google Docs?

[00:07:16] Dan: Yeah.

[00:07:16] Nathan: And so then you get the full track changes as if you had just given that prompt to a human editor.

[00:07:20] Dan: Exactly.

[00:07:21] Nathan: Okay.

[00:07:22] Dan: And I, by the way, have human editors. Right. I have two, um, who worked on this piece with me who are ex- extremely talented.

[00:07:27] Mm-hmm. But this is, this is at a stage before they would look at it.

[00:07:31] Nathan: Right. Yeah, and you talked about the process of inside the Everyagent of how you tried to, I guess, productize, agentize your editor-in-chief, right? And-

[00:07:42] Dan: Yes …

[00:07:42] Nathan: and, like, learn from the e- the edits that she makes, and then get, you know, ev- like elevate everyone’s writing before she even looks at it.

[00:07:49] Dan: Yes, we also do that. Um, so what we do, so we have the Everyagent. It’s a AI coworker that makes it super easy to AI pill your entire organization ’cause you, when you use it in your Slack, you, you, you prompt in public, and that, that- So everyone else sees- … just teaches people- Yeah … ’cause you’re showing them instead of telling them.

[00:08:06] Uh, instead of get- trying to explain it all the time, you’re just like, “@Every do this thing,” and it works the way that we work. So it just makes it easy to show everyone and get them to adopt it. And one of the workflows that has started to work really well is thinking about how do we compound the knowledge of our experts in the organization so, uh, and push that into the agent so that other people in the organization can use that knowledge and expertise without taking up more of that person’s time.

[00:08:34] Nathan: Right.

[00:08:35] Dan: So with, with Kate, who’s our editor-in-chief, she’s a fantastic copy editor- What I, what we did is we took, um, she’s been with us for, like, four years or something like that, so we have a lot of her edits, and we took, like, 30 or 40,000 of her copy edits and turned it into a skill that whenever, like, before she sees a piece, you just @ The Every Agent and you’re like, “Do a Kate pass on this.”

[00:08:58] And it does, like, a lot of the stuff that she would do so that when sh- by the time she sees it, it’s closer to her taste and she doesn’t have to spend as much time on each individual piece. Yeah. And it gets better over time, so it learns, uh, each time that it does stuff and she accepts or rejects things or adds things, sh- it learns, like, here’s, here’s the kind of stuff I miss, so that she’s doing less and less, but she has more and more impact around, around the org.

[00:09:22] And I think that’s, it’s just like a r- it’s a kind of peek into how you will work well as an expert in- Right … the future, is being the one who is in charge of making sure that system works, um, instead of just doing every single thing by hand. You’re

[00:09:36] Nathan: getting the compounding value.

[00:09:37] Dan: Yeah.

[00:09:38] Nathan: Okay. So talk about how you actually implement that, ’cause I think a lot of people say

[00:09:42] Well, actually, maybe let’s back up to the skill.

[00:09:43] Dan: Yeah.

[00:09:44] Nathan: So a lot of people would say, “Oh man, I have to write an entire skill that does this.” But no human wrote that skill, right?

[00:09:51] Dan: Yes. No human wrote that skill. This is where I would say for now-

[00:09:57] Nathan: Mm-hmm …

[00:09:57] Dan: it’s hard.

[00:09:58] Nathan: Okay.

[00:09:59] Dan: It’s easy to do a first pass. Yeah. It’s hard to actually make it good.

[00:10:03] It will not be hard in six months to a year. So I think, like, I would, I would look at this as a preview of a workflow that will be super, super common for everybody, um, in maybe a year. But for now, it’s something that if you’re, like, super AI-pilled or you’re technical, like, you should try this. Mm-hmm. Um, and the way to do it would be go grab a bunch of, um, if you have a bunch of data that of, like, of you doing a particular kind of task or making- Yeah

[00:10:29] a particular kind of decision, aggregate it, tell Claude or Codex or ChatGPT, like, “Make me a skill,” and then put it into your Codex or ChatGPT or if you have a Slack agent like the Everyagent- Mm … you put it in there. Try to use it and set up your system so that every time you use it, it scans to be like, “What did I miss?”

[00:10:47] And puts it- Mm … back into the skill.

[00:10:49] Nathan: Okay, what’s the… Like, I feel like that feedback loop, I’m obsessed with flywheels, right? Yeah. And so that feedback loop- Mm … is so important. How do you best set that up?

[00:10:56] Dan: All of the solutions right now are a little bit bespoke. Okay. So, so-

[00:11:00] Nathan: Depending on the skill, like, that you’re trying

[00:11:02] Dan: to set up.

[00:11:02] Uh, depending on the skill, but the, the core thing that you’re, the core thing that you’re doing is when you get all this data, uh, you create an eval. And an eval is just a test that’s like, um, it gives a model a bunch of questions or tasks that look like the kinds of stuff it’s gonna have to do in order to do work well in your particular way.

[00:11:21] It’s like the Nathan eval. What the eval allows you to do is it allows you to measure the model’s performance at doing your work. So every time you run the model, you can be like-

[00:11:31] Nathan: Mm-hmm …

[00:11:31] Dan: how, how good did it do at this test? And you can always be adding to the test or changing the test to make sure it, it better reflects, like, what your, what your work is like.

[00:11:39] But the core thing is make an eval. Once you have an eval, uh, have your model make a… have, have Claude make a skill for you that maximizes the score on the eval. So it’s like in 50, in 50% of the cases that, you know, Kate might make a copy edit, we can, we can tell that this model does 50, 50% of… gets 50% right.

[00:12:00] Nathan: Yeah.

[00:12:00] Dan: And then what you do is, uh, over time, as you get more and more data and it’s used in more and more places, uh, you start to… you try to modify the skill to incorporate more of that data or more of those learnings, and then you just measure it again on the eval, and you just try to maximize the score.

[00:12:17] Nathan: Okay.

[00:12:18] Dan: And that’s the, like, high level of how it works. Yep.

[00:12:20] Nathan: Hey, so you may know me as the host of this podcast, but I’m also the founder and CEO of Kit, an email marketing platform for creators. I built Kit because I needed a better way to grow my audience and sell books and courses through email. Kit helps you build an audience that you own and will create automations that keep working for you so that you can focus

[00:12:39] Dan: on creating.

[00:12:40] As a listener of this show, you can get an exclusive 10% off Kit Pro if you go to kit.com/nathan.

[00:12:45] Nathan: I think a lot of creators find themselves stuck jumping from platform to fl- platform- Mm … because they watch a YouTube video or hear a podcast and it’s like, “Muse is out now, it can do this,” and then they’re, like, halfway through the setup, and then it’s like, “Dots can do that.”

[00:13:00] Yeah. And you’re like, “Oh, I should jump, jump from that.” Well, a decision that I made recently was to just … I, I mean, I have subscriptions to everything, but to just pick one platform and dive all in. I chose GroqBot. It’s working really well for me. I don’t know that it’s better than anything else.

[00:13:16] Dan: Yeah.

[00:13:17] Nathan: You know, in two weeks or six months, the answer might be different.

[00:13:21] My take is that m- people would be better off to just pick one and go deep than, like, constantly be 5% deep in a whole bunch of them. Um, I’m curious for two things. One, what’s your take? Like, what would you recommend, uh, someone do? And then two, how would you recommend they think about which platform to choose?

[00:13:38] Dan: It’s gonna be very personal. Yeah. It’s gonna d- depend on who you are and what your thing is. Like, for me, like, my job is to test all this stuff. Right. Yeah, you

[00:13:45] Nathan: need to

[00:13:45] Dan: be in all of them. So I, I wanna be testing all the different, all the different things. And part of the reason it became my job is ’cause I just li- I just like doing that.

[00:13:52] Yeah. You know? Um, I would do it even if it wasn’t my job, which is, I think, part of why- Like, how it, it ended up being your job … I’m good at it, you know? Yeah. As well. Yeah. It is meaningful to me, for example, that I just know that Gemini is really good at this right now.

[00:14:05] Nathan: Mm-hmm.

[00:14:05] Dan: And I … It’s, like, so fun to just jump into Gemini for writing tasks- Mm

[00:14:11] even though I’m using Claude for, like, long w- w- like, this new machine l- m- machine interpretability, uh, paper came out today, and I’m … Like, right now it’s working on, like, building something based on that paper. Okay. And then my dot is, like, doing all my, my, like, nor- normal day-to-day work, and so I’ve got all these different things on different platforms.

[00:14:29] The company’s called Every. Every. Like, I like, I like doing lots of things, you know? Yep. And I … But I think some people are just like, “I want, I wanna focus. I wanna understand my environment and my tool.” Mm-hmm. “And I wanna, like, almost become one with the tool.”

[00:14:43] Nathan: Right. ”

[00:14:43] Dan: And if I’m a little bit behind the frontier, like, that’s fine,” because, you know, it’ll, it

[00:14:48] Like, all of these things will be at parity, like, the major platforms will be at parity. It might take six months, it might take a year. Right. But, like, it’s gonna more or less be fine. It’s worth thinking about why am I jumping? Am I jumping because I genuinely love this and I’m just like- Mm-hmm … this is the coolest thing, or am I just, have I- I’m having so much FOMO and I’m just spinning my wheels to not miss out?

[00:15:15] Nathan: Mm-hmm.

[00:15:16] Dan: Because trying not to miss out is, like, not the thing. Yeah. It’s like, what do you wanna do with this? Mm-hmm. Is this enjoyable for you? Is it helping you do the work you wanna do? Um, if s- if so, jump. If not, don’t.

[00:15:29] Nathan: What I found is, yeah, is going deep and then just setting up as many connectors as possible.

[00:15:33] And I can still share a workflow that, you know, has worked well for me. I meet a ton of people in person, and so now we’re in this, you know, you’re, you’re talking to someone at an event and it’s usually like, “Oh, let’s share contact info.” And so, you know, touch the phones together, they share, and it just goes off into this, um, you know, void, never to be found again.

[00:15:53] Yeah. Right? And so you have to, like, find them and send them a text right away so that they stay at the top somewhere. And you can implement this, I think, in most tools, I did it in Grokbot, where it monitors anyone added to my contact book-

[00:16:05] Dan: That’s cool …

[00:16:06] Nathan: logs who is, where I was. So it checks my calendar to be like, “Oh, you were at the Authors Equity Dinner-

[00:16:11] Dan: Ugh, I

[00:16:12] Nathan: need that

[00:16:13] when these were added.”

[00:16:13] Dan: That’s a great idea.

[00:16:14] Nathan: And then it l- adds them to a follow-up of who, and it’ll enrich the contacts a- as well.

[00:16:19] Dan: Yeah.

[00:16:19] Nathan: And so the, “Oh, we, we know who this is, we know where you met them.”

[00:16:22] Dan: I

[00:16:22] Nathan: mean- And then I just forget to, like, I don’t forget to follow up …

[00:16:24] Dan: the, like, this is the golden age for the personal CRM bros, you know?

[00:16:28] Yes. Like, this is what we’ve all been waiting for.

[00:16:33] Um, I love that. I, I really actually need to do that ’cause I have that exact thing where I’m like, “Oh, I meet all these people and they’re really cool,” and then I just never talk to them again.

[00:16:40] Nathan: Yeah, and so you could add workflows in that- Yeah … where you could say, “You know what? Everyone I meet in person- Yeah

[00:16:44] I wanna follow on X.”

[00:16:45] Dan: Yeah.

[00:16:46] Nathan: And it would just do that.

[00:16:47] Dan: Yeah.

[00:16:48] Nathan: And so, but now I feel so much more free at an event, ’cause I, all I need to do is touch phones together. Yeah, yeah,

[00:16:54] Dan: yeah, and it’s,

[00:16:54] Nathan: it’s handled. And then, then everything else- Yeah … is handled. I love that. And it’ll be like, “Hey, you met these five people.

[00:16:58] Did you wanna- Yeah … you’re on the flight home. Did you wanna send-”

[00:17:00] Dan: Grokbot’s one of those things, I, I’ve never used it.

[00:17:02] Nathan: Really?

[00:17:02] Dan: I, well, I used it, they sent it to us, so some people on my team use it. They sent it to us- Yeah … before it came out, and I, like, went into it for, like, two seconds and I was like, “This looks exactly like OpenClaw.”

[00:17:12] And- Yes, yes … then I just, like, didn’t use it. Mm-hmm. Um, and then Dots came out, and I’m already… I’m so in the ChatGPT ecosystem- Right … so I do have some of this, like, ecosystem stuff- Yeah … that I was like, “Okay, this is, this

[00:17:23] Nathan: one, this will probably work.” I think you could set up that same workflow with Dots as well.

[00:17:25] Dan: Yeah.

[00:17:25] Nathan: So long as… Can Dots read iMessage?

[00:17:27] Dan: Yeah.

[00:17:28] Nathan: Okay.

[00:17:29] Dan: Uh, if it’s on, if it’s on your

[00:17:31] Nathan: computer. If it, if your computer is open.

[00:17:32] Dan: Yeah, yeah, yeah,

[00:17:32] Nathan: yeah, yeah. It’s a whole thing. Yeah, yeah. I asked a question- Yeah … and it was like, “Uh, your, both your laptop and your iMac are off right now, and so I can’t read that.” And I was like, “Oh.”

[00:17:41] Dan: Sucks. Sucks when that happens.

[00:17:43] Nathan: Yeah, exactly, and I’m like, “Uh…”

[00:17:45] Dan: I get that with Dot, but it’s like my Mac Mini is never off. Right. Like, I’ve told it to never go off, so I’m like, “No, no, it’s, it’s on,” and it’s… and then it comes back and it’s like, “You’re right, it’s on.” And I’m like, “I know.”

[00:17:55] Nathan: I need to have my Mac Studio at home, not

[00:17:57] Dan: at work.

[00:17:57] But it’s so funny that that’s the thing that is, they’re being dumb about right now because, like, a year ago it was so different.

[00:18:05] Nathan: Mm-hmm.

[00:18:06] Dan: It was so different. A year ago, I couldn’t, you couldn’t tell the AI, “Hey, can you help me write this as a tweet?” Because if you asked it to write something, like help you write a tweet, it would put like 14 hashtags into every tweet.

[00:18:18] Nathan: Right.

[00:18:19] Dan: And now, like, obviously that doesn’t happen anymore, and it’s just so… It’s crazy how much it’s changed.

[00:18:24] Nathan: Well, so I want to talk about, like, the future of AI, ’cause I think so many people are absolutely freaked out.

[00:18:30] Dan: Yeah.

[00:18:30] Nathan: And you have a, a different take where you’re like, “The future is great.” Like, I, I talk to you and you are high on the future of where this is going.

[00:18:40] Like, what do you say to someone who is timid about it and is saying like, “I don’t, I don’t know if the future’s gonna be better than the past”? Um, I mean, first of all, I do think that there’s just a little bit of a personality thing-

[00:18:52] Dan: Yeah … about, like, new things and change and whatever, and d- different people have different feelings about, about that.

[00:18:57] And, um, I think, you know, if I was in, you know, around the, the Enlightenment, I would’ve been be like, “This printing press thing is so

[00:19:08] Nathan: sick,”

[00:19:08] Dan: you know? Yeah. And people have been like, “It’s ruining, like, uh, oral communication- Right … and social standards and, like, whatever, you know? Um, and, uh, b- I think you have to kind of have both in a society, so I, I kind of, I kind of value both.

[00:19:21] But my, um-

[00:19:22] Nathan: But you know which r- which role you play.

[00:19:23] Dan: I know which role I play, yeah. Um, which, which is to say I actually… I think that there are some people who love new things and hate old things. I actually love old things. Mm-hmm. Like, I love books. My house is full of books. Yeah. I only read paper books.

[00:19:34] Nathan: Mm.

[00:19:34] Dan: I tend to be the kind of person that’s like, “I’m gonna go swim in this new world and then tell you how and why it can be just as great,” and, and- Mm … and it can connect to the old world that you’re- Mm … thinking is totally gone, ’cause it’s, it’s g- there’s gonna be things that are different, but there’s gonna be a lot of things that are the same.

[00:19:50] And the, one of the core things that I think is the same or will, will end up being the same is when people encounter a new technology, they often have a little bit of a disgust reaction or like a, it feels shallow or flat or somehow inhuman or impersonal. And m- m- what I feel is when you look around, like we’re in this podcast studio, right?

[00:20:16] All of- We’re surrounded by technology. Right. That if, uh, that for us maybe, you know, podcasting in this way in person it feels very, it feels personal and it feels like a human thing, but if you went back 100 years it would be like, “What the fuck is going on here?” “This is crazy.” And so, uh, is to s- is to, one, recognize that that feeling of it- it’s- it’s- it has some soul, I have some connection to it, is something that you build up over time, um, by being in contact with the world, which includes all the technology in it.

[00:20:49] So just because you don’t have that with the machine right now doesn’t mean you won’t build it.

[00:20:54] Nathan: Right.

[00:20:55] Dan: And, and in fact, I think we will and do. And generally, I think there’s a w- w- I’ve observed a very predictable pattern of how people encounter and then adapt to new technology, which is first you encounter it and you’re like, especially with AI, you’re like, “Holy shit, this is going to like completely erase me.”

[00:21:17] It, it’s like, uh, I’ve been calling it an agency rupture. It just … It, it feels like this thing that I do well, it’s no longer valuable.

[00:21:24] Nathan: When, like iterations of that when you first see it write really well and, or like you send a prompt and it comes back or it completes a task. Another one is when you watch it use your computer better than you do.

[00:21:34] Dan: Yeah.

[00:21:35] Nathan: Like the video editing is one of those- Yeah … where you’re like, “I thought I was pretty good at keyboard shortcuts- Yeah. … for Premiere and you know keyframes- I am not good at that … and all that. Yeah. And you’re like, “Oh.” It’s, it’s one thing where you’re like, “Okay, it went off and it did magic and it came back.”

[00:21:46] Yeah. But when I can sit there and watch it and I’m like Huh,

[00:21:50] Dan: okay.

[00:21:50] Nathan: Yes. You know, it’s another one of those moments where you’re like, “Okay, my, my worldview, w- you know, what I- my view of what’s possible needs to change.”

[00:21:57] Dan: Exactly. It’s a, and s- and it’s an, it’s an agency rupture, and it feels like shit. I think if you’re someone who’s a tech person and like to write, you had this around GPT-3.

[00:22:08] You, you were s- you saw GPT-3 writing- Mm-hmm … and you’re like, “Fuck.” Yeah. And then it happened for coders over the last couple years.

[00:22:16] Nathan: Right.

[00:22:16] Dan: And then now it’s happening for mathematicians. Mm. And it’s just like a, it’s a very regular thing. Mm-hmm. But I, I think that that’s the first stage is encounter, first encounter and rupture.

[00:22:26] The second stage to me is, “Oh, yeah, it can do all that. It can use my computer better than me, but I have to do a lot of, um, babysitting of it and setting it up to, to, to do the task in the right way at the right time.”

[00:22:39] Nathan: Mm-hmm.

[00:22:40] Dan: But you’re like, “Ah, but that’s still, I’m just kind of babysitting a model, and the model’s doing the work.”

[00:22:45] And then the, the last phase is you’re like, “Actually, that itself is a full task full of like richness and complexity that I didn’t realize was there because I’d never had to do it before. But now that I’m really thinking about this, there’s so much that can be done to make, you know, maybe a model can go and edit a video in 20 minutes, but what kind of video?”

[00:23:07] Right. “And what are all the details of that? And, and, and, and not just any given video, but all of the videos that you produce.” And so you’re, you started to open up this whole new universe that you didn’t qu- quite realize was there. The way that people tend to think about new technology is sort of the way that people thought about what was beyond the horizon in the Middle Ages.

[00:23:27] It’s like, if you went beyond the horizon, it’s like it falls off. There’s dragons. Right, who

[00:23:31] Nathan: knows?

[00:23:31] Dan: Yeah. Yeah. But like you take a ship out there, and you’re like, actually there’s just more, more land and more stuff, you know? And it’s, there’s good stuff. There’s bad stuff. There’s … I- if you spend a lot of time out there, y- you, you make a, you make a home there.

[00:23:43] It feels, it can feel amazing. You might be homesick too, but, um, I think it’s, it’s much more like that than, oh, it’s, uh, we lose everything that we love and, uh, a- and we all die. Right. Like, it’s, it’s not usually like that.

[00:23:59] Nathan: How do you think about ambition in the age of AI, especially as it relates to creators?

[00:24:04] Dan: I’ve been talking to a, a couple of different- like sort of bigger name people about this recently because, and we see this too a little bit, but especially if you’re someone who, you know, you- you’re a big name and people come to you for answers to questions, for example, you’re probably seeing your search engine traffic drop off-

[00:24:21] Nathan: Oh, yeah

[00:24:21] Dan: you know? Um, or I know, like, I know Tim Ferriss was tweeting, you know, uh- Book sales went down … my book sales are down, like that kind of stuff.

[00:24:27] Nathan: I do know other authors who are having the opposite experience.

[00:24:32] Dan: Yeah. I, I, I, I think that, I don’t think it’s one or the other.

[00:24:37] Nathan: Yeah.

[00:24:37] Dan: And there’s a lot of reasons why Tim in particular, because The 4-Hour Work, Work Week is like, it’s so, it’s such a book that should be written now-

[00:24:46] Nathan: Right

[00:24:47] Dan: but rewritten, ’cause, like, all the stuff would be different, but he could d- he could do like- But

[00:24:50] Nathan: the, the premise-

[00:24:51] Dan: He could do The 4-Minute Work Week. Right. You know? The premise is so good.

[00:24:55] Nathan: Of extreme leverage.

[00:24:56] Dan: You could actually do it now. How do I use extreme leverage? Yeah. You know? Oh, yeah. You can actually do it, so I think, I think he has a huge opportunity if he wants it.

[00:25:02] Mm-hmm. But if you go read The 4-Hour Work Week right now, there’s so much translating that you have to do- Mm-hmm … from, you know, I read that when I was in middle school or high school. Yeah. There’s so much translating you have to do from that era to this that I can understand why it might not be as popular right now.

[00:25:17] Right. But I, there’s this other thing that’s happening, which is if people go to you for answers to questions, Google has changed things, so now there’s, like, the AI answers- Yeah … and they go to your site less, and there’s a whole thing around that. Things are always changing, especially in media, being a creator.

[00:25:30] Like, dis- some distribution channels are always working, some are dying.

[00:25:35] Nathan: Mm-hmm.

[00:25:35] Dan: Um, some m- monetization methods are always, are working, some are dying. You know, at every, we’ve always had, like, different legs of the stool on the business side, ’cause you just, like, lean to the one that’s, like, working, you know?

[00:25:46] Yeah. Uh, I, I think for us, the subscription has always been the core, and that’s the, that, that’s the thing you wanna compound. But sometimes, like, your courses or your ads or your consulting or, like, one of those things is ripping and it’s really nice, ’cause you’re like- Yeah … “Well, my subscription has been flat for the last couple months,” or whatever.

[00:26:01] So it’s n- it’s good to be, I think, diversified as a creator, and you’re always kind of looking for what’s the, what’s the pocket that I can grow into or move into? But I do think if you’re looking at that and you’re like, “My search, my search traffic is going down,” it can be demoralizing, and you can be like, “Holy shit, but I could do-” A hundred times more than I could previously Right Um, uh, just simple examples are you can, you can legitimately over a weekend, and this will be more and more true over the next, like, six months or so, over a weekend make an app for you and your community that’s very good by yourself.

[00:26:39] And,

[00:26:39] Nathan: and totally bespoke.

[00:26:40] Dan: Yes, by yourself.

[00:26:41] Nathan: Yep.

[00:26:41] Dan: Um, uh, do you know Johnny Miller?

[00:26:44] Nathan: No, I don’t.

[00:26:45] Dan: So Johnny’s a, a good friend of mine. Um, he d- he runs this course called, uh, Nervous System Mastery. Okay. Like very cool breathwork guy. Oh, I

[00:26:51] Nathan: have

[00:26:51] Dan: seen that. Nice. Yeah.

[00:26:52] Nathan: Yeah.

[00:26:52] Dan: And he, he made this app called State Shift that he was working with someone else who was technical for a while to, like, make it, but it, like, that, that didn’t quite work out, and then he just, like, rebuilt it over a weekend with Fable and it’s not technical at all.

[00:27:04] Nathan: Work- And it’s exactly what

[00:27:04] Dan: he

[00:27:04] Nathan: wanted … works beautifully.

[00:27:05] Dan: Yeah. Yeah. It works beautifully.

[00:27:06] Nathan: And the feedback loop is so much tighter. Yeah,

[00:27:08] Dan: yeah. So now you can make an app. I, you know, some of the things that we’re thinking about are, I think that feeds used to be something that you could only make if you were a social media company with, like, hundreds of millions of users and, you know, billions of dollars.

[00:27:23] Nathan: Uh-huh.

[00:27:24] Dan: Every creator or media company can, like, s- essentially have a feed that is full of stuff that every time someone in your audience checks it, it’s, like, something that is good, that is to your taste, that is, like, sen- probably something that you made.

[00:27:38] Nathan: So typically the problem with a feed is that you need, uh, volume, like typically from a marketplace- Yeah

[00:27:44] or social network for it to be interesting and off and- Yeah … off and off. You, you check the feed a couple times and there’s one new thing- Yeah … and then zero new things. Yeah. And then the habit dies.

[00:27:51] Dan: Yeah.

[00:27:52] Nathan: So what, what are you populating these feeds if you’re n- with, if you’re not-

[00:27:56] Dan: Well- …

[00:27:56] Nathan: relentlessly creating

[00:27:57] Dan: uh, this, we’re very early here.

[00:27:58] Nathan: Yeah.

[00:27:59] Dan: Very early, and it’s unclear to me who should own the feed. Mm-hmm. Um, is the feed something that I as a consumer put together for myself?

[00:28:08] Nathan: Mm-hmm. Oh, that’s interesting.

[00:28:09] Dan: That will probably happen. Um- The

[00:28:11] Nathan: answer d- is, like, should it be this or that, probably n- the

[00:28:13] Dan: answer- It’s probably both

[00:28:14] is probably yes. Yeah, it’s, it’s probably both. Um, eh, I think that there’s a big opportunity for the Clauts and ChatGPTs of the world to, uh, let anyone make, compose their own feeds that are, that are on in Claut or GrokBots or whatever-

[00:28:26] Nathan: Mm-hmm …

[00:28:27] Dan: but that you can talk to in English And just say, “I don’t want, like, X, Y, Z thing that you keep serving me.”

[00:28:33] Like, th- that’s the difference between AI, like this generation of AI and previous generations of AI is it’s not just, um, based on re- revealed preference, it’s also based on stated preference. Oh,

[00:28:42] Nathan: yeah. So

[00:28:43] Dan: I’m always gonna look at the car accident.

[00:28:45] Nathan: Yeah.

[00:28:45] Dan: But, uh-

[00:28:47] Nathan: You can say, “I don’t actually want to.”

[00:28:48] Dan: But I

[00:28:48] Yeah, but like, d- don’t send me this. Mm-hmm. And it, it won’t. And that’s one of the reasons why when you ask ChatGPT, like, what am I into, versus when you look at your Instagram For You page, like one is … My ChatGPT one is very wholesome and my Instagram one is not. Right. You know? And it’s like, yeah, you’re gonna look at the car accident.

[00:29:07] Nathan: Yeah.

[00:29:08] Dan: Um, so I th- I do think that there’s, uh, going to be a world where, uh, the, you know, your agent makes a feed for you or several feeds. Mm-hmm. And I do this all the time for myself, like I have a feed that I check regularly of, of the meetings that happened in the company, what are meetings that, um, there was an unclear priority or there was like a 50/50 decision that I should put my finger on?

[00:29:32] Nathan: Okay. And so you’re having it read Granola transcripts- Yeah … or whatever-

[00:29:36] Dan: Yeah …

[00:29:37] Nathan: and then score it based on something

[00:29:38] Dan: We record all team meetings in Notion.

[00:29:40] Nathan: Okay.

[00:29:41] Dan: And it’s publicly available to everybody, or, uh, it’s available with anyone in our Notion. With anybody. And then I have a feed and it just tells me, “Hey, like, m- uh, you know, there’s, there’s competing priorities here.

[00:29:49] You should, like, help.”

[00:29:51] Nathan: Yeah.

[00:29:51] Dan: And so, like, that’s a custom feed for me, or we have that for, like, sometimes for hiring for, like, certain job roles. Mm-hmm. I’m like, “I wanna see everyone who’s applying and try to pick out thing- people that seem like they’d be good.” I think that’ll happen on the, on the agent side or, like, in the, the big agent platform side, but I, I also do think, like, for us, for example, we’ve started to make feeds for…

[00:30:12] We have a frontier team and- Okay … the job of the frontier team is just play with models and-

[00:30:17] Nathan: Right …

[00:30:18] Dan: try stuff and whatever. And so we, we’ve started to create, like, a feed of here’s everything that we’re doing. Mm. It’s like it’s software experiments, it’s little, little bits and bytes of research, it’s, you know, it is, it’ll eventually, like, we’re now at, we’re at now at a point where you can really make very good videos that are essentially one-shot.

[00:30:40] It’s like you talking to a camera and then it turns it into, like, a really good, like, video for, like, two minutes. Especially if you’re a team of multiple people, it’s very easy, I think, actually, to make more content than any given person would even wanna consume in a day.

[00:30:53] Nathan: Right.

[00:30:54] Dan: Um-

[00:30:54] Nathan: Well, that’s interesting, you…

[00:30:55] It’s no longer the bottleneck.

[00:30:57] Dan: It’s not the bottleneck, yeah.

[00:30:58] Nathan: Yeah.

[00:30:58] Dan: Um-

[00:30:59] Nathan: And then especially if you’ve been at this game for a while.

[00:31:01] Dan: Yeah.

[00:31:02] Nathan: Right? You’ve written a book, you’ve done- Yeah … you know, you have- Yeah … 100 back catalogs.

[00:31:06] Dan: Think about, think about how Ryan Holiday works.

[00:31:08] Nathan: Mm-hmm.

[00:31:09] Dan: Everyone can do that basically.

[00:31:11] Um, but without the- You’re

[00:31:12] Nathan: talking about the, the note cards and the level of proliferation and all that or what do you mean? Yeah.

[00:31:16] Dan: I’m sort of talking about the, the… This is actually really interesting ’cause I actually wrote, this is going back to the first AI piece I ever wrote was in, I think, 2022.

[00:31:26] Nathan: Okay.

[00:31:26] Dan: And it was, um, AI in the age of the individual, and it was like the way that we will use these tools is essentially the way that Ryan Holiday works, which is he’s off doing his, like, you know, he’s reading, he’s reading Seneca or whatever, and turning that into, let’s say, a book. But he has a whole team of people who are then turning that into a, a, a blog post or a podcast script- Right

[00:31:49] or a video or, like, all the, all these different things. Everyone now has that at their disposal, but they don’t have to pay or be the level of writer that a Ryan Ho- Ryan Holiday is. The interesting thing right now is we’ve relaxed the bottleneck. This, this is happening for us right now. So we’ve relaxed the bottleneck on- creation.

[00:32:08] There’s so much in my Slack where every time I look at it I’m like, “Holy shit, this is fucking awesome. It’s so cool.” And we do not have, uh, good enough capabilities on the editing and curation side.

[00:32:24] Nathan: Okay.

[00:32:25] Dan: Every time you add AI to one side of a process, you need to add it to the other side. We’ve added it to the creation side.

[00:32:30] We’ve not- Yep … it’s not caught up on the editing curation, like discernment side.

[00:32:36] Nathan: Mm-hmm. This

[00:32:36] Dan: is actually the same problem as the agent swarms. It’s like we can set these things off. We c- just can’t tell when they’re going off the rails.

[00:32:43] Nathan: Right.

[00:32:45] Dan: It’s just a, it’s just a cap- a capabilities mismatch. But I do think over the next year or so what I expect us to do…

[00:32:51] So for example, the way that we work right now is when a model drops, we send it, an email out to everybody. Um, but also we’ve started to send, uh, what we call frontier alerts, where we have our, our Everyagent in Slack and, um, if you’re using the Everyagent, it knows a little bit about you and- Mm-hmm … the workflows that you have and what you care about.

[00:33:13] And when a new model drops, if you should pay attention to it, we say, “Hey, this new model dropped. It’s good for the kind of thing that you care about, X, Y, Z.” Right. “You should try it.” If not, we just don’t say anything ’cause, like, we, we basically want you to be able to, like, ig- ignore it and just- You

[00:33:28] Nathan: want a very high signal.

[00:33:29] Dan: Exactly.

[00:33:30] Nathan: Mm-hmm.

[00:33:31] Dan: Um, so but that’s a really interesting system to build because- We’re not writing a vibe check. We’re, we are writing a vibe check that’s a story, but there’s also, it’s gonna go out to people in this, like, custom way. It’s bespoke. So we’re building a system that has to know how to write for each person.

[00:33:49] Nathan: Mm.

[00:33:50] Dan: And s- and so we’re sort of moving up a level and then you, then your, your editors are, like, looking at, uh, okay, this message went out to this person. They had this reaction. Um, what should that teach us about how we should change the system or how do we, how do we make it better at writing in the future?

[00:34:06] So, so everyone’s moving up a level- Right … and the, what’s the output is you’re not necessarily looking at every single output- Mm … but you’re making sure that each output, like, carries your taste, and that’s the whole game. And it’s, um, you know, there’s a lot of analogs with having a team, um, and hiring people and making sure that they output stuff.

[00:34:25] Like, if you have a ghostwriter, for example.

[00:34:29] Nathan: How do you think about hiring right now? You know, you, in your interview with Sam Altman, you were talking about the sheer number of developers. Y- like, AI development is better than it’s ever been. Mm. It’s getting better every single week.

[00:34:41] Dan: Mm-hmm.

[00:34:42] Nathan: And so we’re hiring more developers than ever- Yeah

[00:34:45] which is the opposite of what the layman would expect.

[00:34:48] Dan: Yeah.

[00:34:48] Nathan: Uh, and so I’m curious for your take on that, and then also how you think about hiring within Every, Every.

[00:34:53] Dan: You would imagine if you’d told me, like, uh, five years ago that we would have a model that was as good as Fable or Astra- Mm … that I probably wouldn’t need any engineers, and we have plenty of engineers, and we’re hiring more.

[00:35:04] Nathan: Right. Same.

[00:35:05] Dan: Yeah. Um, and I think there are- th- really interesting reasons why that is structurally true. It doesn’t mean that, um … I do think probably m- macro y- you will be able to have companies that scale to larger sizes with smaller teams, but doesn’t mean that you don’t need engi- any engineers. I would project the job of engineer to be a job that, like, hiring increases over- Yeah

[00:35:37] like, in, in total in the economy. And one reason for that is now we live in this world where, you know, everyone can YOLO a pull re- pull request to the website. Who’s gonna be the one to make sure that that all works together well? Yeah. Like, yeah, you can have your super intelligent model and it might say no, but, like, the minute you say, “But yes, you should do this,” it’ll just do it and then- Right

[00:36:00] suddenly you have a whole hairball, you know?

[00:36:02] Nathan: Yeah.

[00:36:02] Dan: So you actually, when everyone can make stuff, you actually increase the demand for engineers who can, like, build systems to make sure that everybody, all the stuff that people can make is used well. And then also on the other side, um, the, the ambition, the level of ambition that you can go for is so high.

[00:36:24] Mm-hmm. It’s so high. Like, there are people on my team who are just, like, one-shotting real production level, like, pro software and it’s like cr- it’s crazy, and that’s just for themselves.

[00:36:36] Nathan: Mm-hmm.

[00:36:37] Dan: And, uh, so that opens up all these different new things to do. But I think that’s why I’m not particularly worried about engineers.

[00:36:44] Nathan: Right.

[00:36:45] Dan: I think the job of engineer changes a lot- Mm … and the kinds of minds and thinking that are maybe re- rewarded might change to some degree- Mm … which is an interesting and important topic I think to, to dig into. But, um, there’s st- there’s lots and lots of demand for this.

[00:37:02] Nathan: If we’re diving in on things that change, uh, you have a take that is going, uh, trending on X at this very moment.

[00:37:09] And that is on the future of newsletters and the inbox, which is a topic that you and I both care about a lot- Yeah … because we’ve built our businesses- Yeah … around this very thing.

[00:37:18] Dan: Yeah.

[00:37:18] Nathan: So, uh, what’s the take, and we’ll dive in on

[00:37:21] Dan: it. It’s so funny, we were talking about this in the, in the pre-production and it’s like I fired off that take, I gave it no thought, and it’s like way more viral than, than this, like, 8- Than the piece you spent

[00:37:29] 8,000 word piece that I, like, slaved over for, like, three months or whatever. Um, so, but so it goes. Um- The spicy part of the take is, like, the newsletter is dead.

[00:37:39] Nathan: Mm-hmm.

[00:37:39] Dan: But what I

[00:37:39] Nathan: re- Says the person who has, what, over 100,000

[00:37:42] Dan: subscriptions. Yeah, 120, 130, yeah.

[00:37:43] Nathan: So why is that? Like, what are you seeing?

[00:37:45] Dan: The thing I mean by that is n- not

[00:37:50] And the way people took it, which to be honest, I was like, I knew what I was doing by, like, saying that. It’s

[00:37:55] Nathan: okay to be misunderstood- You know?

[00:37:56] Dan: Yeah … on the internet because it- It’s, like, good. Usually. You know? Like, I knew what I was doing. It leads

[00:38:01] Nathan: to more virality.

[00:38:01] Dan: Yeah, yeah. The thing I meant by that is for a long time- It has been the best advice and the conventional wisdom that You, all you have to do as a creator is get a person’s email

[00:38:19] Nathan: And then you, like, have that relationship.

[00:38:21] Dan: Yeah. You’re set. Like, you own the relationship.

[00:38:22] Nathan: Mm-hmm.

[00:38:23] Dan: And that has been sort of cracking and crumbling for a while. Uh, specifically the newsletters tab, the, and the promotions tab has been like- Yeah … a, a shot at that where, yeah, you’re really not actually guaranteed. And- Mm … and so now there’s all these workarounds where it’s like, you know, mark it as priority so it doesn’t end up in whatever.

[00:38:41] Nathan: Reply is

[00:38:41] Dan: m- making a big difference. Yeah, all that kind of stuff.

[00:38:43] Nathan: Yep.

[00:38:43] Dan: And I think we just saw … We’re, we’re seeing a new bump, a new, like, level in that. Uh, it’s, like, escalated a lot with, you know, dots or Grok bots or, you know, all these other tools where I, I really don’t look at my inbox that much.

[00:38:58] Nathan: Yeah.

[00:38:58] Dan: I really don’t.

[00:38:59] I look at Slack a lot, but, uh, mostly my, my dot handles everything and then I go into the, my inbox every once in a while for, like, I just need to do something really quick.

[00:39:09] Nathan: Mm-hmm.

[00:39:10] Dan: And so I just … I’m not even encountering my newsletters tab. And if I’m doing that now, it’s gonna take a while for everybody to do it.

[00:39:20] Nathan: Right. Yeah, someone in the comments was like, “Yes, but you’re at the .01%.” Yeah. And you’re like, “Okay, great.”

[00:39:26] Dan: Yes.

[00:39:26] Nathan: So what are we setting as the time horizon? Yes, exactly. It, it’s not like, oh, so it’s not gonna happen. Yeah. It’s like, so just when?

[00:39:32] Dan: Yes, exactly. That’s- I,

[00:39:33] Nathan: we’re not saying three to four weeks from now.

[00:39:34] Dan: Yeah.

[00:39:35] Nathan: You know?

[00:39:35] Dan: Over the next couple years.

[00:39:36] Nathan: Yeah.

[00:39:37] Dan: Um, and, and not only that, but the models will be so much better that, like, for sure, I think the medium of email-

[00:39:43] Nathan: Mm-hmm …

[00:39:43] Dan: will still be around. Like, you’ll still wanna use, like, SMTP or whatever to, like, send texts back and forth, but you will absolutely not be i- in general looking at a list of, uh, uh, last in, first out ranked list of- Right

[00:40:00] emails that you have to deal with.

[00:40:01] Nathan: Mm-hmm.

[00:40:01] Dan: Right? You’re gonna have your assistant be like, “Hey, like, three things came in. What do you want me to do?” And you go like, “This, this, and this,” and then- Yeah … done. The thing that that misses, obviously, is … I mean, my, my dot has never been like, “Hey, this newsletter came in that you should read that you would like.”

[00:40:16] Um, I, I now have it, I say every time every publish is, like, send me something.

[00:40:20] Nathan: Right.

[00:40:21] Dan: Um, ’cause I’m, like, experimenting with this ’cause I think there’s, that’s a new playing field that I think is gonna be really important. But I think very few people are thinking in that way. Mm-hmm. And it’s gonna be a rude awakening in two years when you’re like, “My open rate is zero,” or it’s 100%, but, like- Right

[00:40:40] no one is interacting with it or with my stuff.

[00:40:42] Nathan: It’s not, it’s not the humans behind it.

[00:40:43] Dan: Yeah. If you’re a writer on the internet, you are … You’re, you’re writing stuff you think is great, and you are also thinking about what subject line is gonna work and, like- Mm-hmm … all that kind of stuff, and the rules are gonna change a lot because the first thing that’s gonna see the email is it’s not Google’s promotions filter, it’s Grok bot.

[00:41:03] Nathan: Mm-hmm.

[00:41:04] Dan: That’s very different.

[00:41:05] Nathan: Yeah, and you could tell it, like, “Hey, when this …” So I had Seth Godin on the podcast, and, uh- He was like, “How do you, how do you become a creator who is so good that you’re missed when you don’t publish?”

[00:41:17] Dan: I agree. I think that that’s like, that’s the, the true size of your audience is, um, if, if you’re somehow you always get into the promotions tab, like, who goes and finds you again.

[00:41:29] Nathan: Yep.

[00:41:30] Dan: You know?

[00:41:30] Nathan: Like, yeah. Yeah. Those, those are the people. And so that’s the thing where someone i- they prompt and they’re like, “Why, why hasn’t Dan written anything recently? That’s so weird.” And it’s like, “Oh, because my Grok bot or my- Yeah … DOTs has been filtering all of that out,” and so then I’m prompting it saying, “Hey, I actually do wanna read that.”

[00:41:45] Dan: Yeah.

[00:41:46] Nathan: And so the bar, it’s kind of like a dragging from the promotions to the inbox, but it’s like, it’s on… It’s another level- Yeah … of it.

[00:41:56] Dan: And, and I think it’s so early that right now I’m like, cool. Every, I read every… Or, you know, my team writes it or-

[00:42:03] Nathan: Yep …

[00:42:04] Dan: whatever. Um, I want, I want that in my inbox. I haven’t done that with anybody else.

[00:42:10] Mm. I will probably over the next five months decide Nathan’s got a great newsletter. Mm. Like, also send me, but then okay, what, what’s gonna, what’s gonna happen? The o- the obvious thing is gonna happen, I’m gonna start getting a bunch of notifications from my bot that’s like, from stuff that I don’t really want.

[00:42:26] Nathan: Yeah.

[00:42:27] Dan: And then I’m gonna try to tune it, and then OpenAI or Claude is gonna, like, come out with an inbox. Right. But it’s gonna be different. Mm. It’s gonna not look like an email inbox in the same way. It’s gonna be, “Hey, we like…” I mean, we’ve been doing this for a long time. We have this, um, we have this email agent we, we built over the last two years, it’s called Quora- Mm

[00:42:47] that summarizes all your, all your emails. It, like, looks at all the stuff that you need to… Looks at all the stuff that comes in, it filters out, uh… It, it makes sure you see anything timely in your inbox, and then it just summarizes the rest, so you just, like, read one thing and you’re done. They will absolutely have that in ChatGPT and Claude.

[00:43:06] There will have to be… Well, A, the game for creators will be how do I make sure that I’m included in that- Right … and rank well on that so that people click on it and then read my stuff? Which means how do I write it so that Claude thinks it’s important?

[00:43:21] Nathan: Mm-hmm.

[00:43:21] Dan: Um, and then I think on the OpenAI and Anthropic and, you know, Google’s side, it’s gonna have to be- How do we, h- how do we create a new playing field or a new platform that works for people?

[00:43:37] Because people wanna read, they wanna read from the people they like.

[00:43:40] Nathan: Yep.

[00:43:40] Dan: Uh, but the physics are all different here, so they need to be figured out.

[00:43:43] Nathan: So I’m thinking about that in particular. If w- I, I obviously have my, the last 13 years of my career built around email and newsletter specifically. So I always think about, one is whatever’s coming next is going to kill email, right?

[00:44:00] Because it’s been, uh, forecasted many, many times. It

[00:44:03] Dan: won’t kill it.

[00:44:04] Nathan: It … Yeah, and so but then trying to break it apart and, like, what are the component parts of it? So the way that I’ve thought about it is, um, email right now is a unique identifier, you know, of like, okay, I can reach you at your email address and, and that sort of thing.

[00:44:19] And for whatever reason, that’s become the thing that we place in all these spots around to be like, “Hi, this is who I am.” Yeah. It’s a, as you mentioned, like, it’s a communication protocol of, like, SMTP, um, and it’s a few other things, but those in particular I’m like, okay, what actually gets replaced? People would say, like, oh, SMS, um, or I guess RCS, right, is gonna replace it or, you know, WhatsApp or Signal or these other things.

[00:44:44] My, my take and I- I’m curious for your thoughts on it is I, I watch, like, uh, Zuckerberg and Altman and whoever else, they like … Like, I can’t see OpenAI being like, “You know what? Let’s bet everything on WhatsApp as the communication protocol and let’s run all of our-

[00:45:01] Dan: No, no, no …

[00:45:02] Nathan: things through Zuckerberg.” Like-

[00:45:03] Dan: They

[00:45:03] Nathan: won’t do that

[00:45:03] it’s not gonna happen.

[00:45:04] Dan: They won’t do that. Um, your agents will connect to email. Mm-hmm. The … I think the, a better- It’s the

[00:45:09] Nathan: inbox place that’s gonna be-

[00:45:10] Dan: Yeah. I think a better analogy w- might be When you meet someone at a conference and you bump phones- Yep … you exchange your phone number.

[00:45:19] Nathan: Mm-hmm.

[00:45:20] Dan: But you’re not leaving them a voicemail.

[00:45:22] Nathan: Right.

[00:45:23] Dan: You

[00:45:24] Nathan: know? And actually, the phone number is not even relevant- Yeah … ’cause we’re gonna, we’re gonna text over

[00:45:26] Dan: iMessage. You’re gonna text and then you’re gonna FaceTime or whatever.

[00:45:29] Nathan: Yep.

[00:45:29] Dan: And so I think the, the n- the email, email as a network, as a communication substrate- Mm-hmm … is obviously very stable.

[00:45:37] It’s gonna stay- Yeah … stay that way. But, uh, who is sending and receiving those emails?

[00:45:44] Nathan: Mm-hmm.

[00:45:44] Dan: Who? Um, and then what is the, what is the f- like, product surface that the eventual email owner is looking at is gonna be very different.

[00:45:54] Nathan: Mm-hmm.

[00:45:55] Dan: And those will have very different mechanics to them that right now, if you’re a newsletter writer, you have not built your business for that world ’cause it hasn’t existed.

[00:46:05] Nathan: Right.

[00:46:06] Dan: But it will exist. So, uh, y- y- there will be some value, for example, to having a person’s email because their agent’s gonna read it-

[00:46:16] Nathan: Mm-hmm …

[00:46:17] Dan: for sure. But it’s not the same guarantee that the human is gonna read it that it used to be. Mm-hmm. You’re gonna have to do other things to make sure that that happens.

[00:46:26] Nathan: Yeah. What do you think, like, how are you positioning Every? Like, the, ’cause the newsletter’s always been at the core of what you’re doing. Yeah. And then now you have this, this insight, somewhere between hot take prediction and insight. You know? Of you’re like, okay, this is where it’s going, and as an entrepreneur, our job is to have those sorts of insights and then make decisions- Yeah

[00:46:46] that result in monetary value around that.

[00:46:48] Dan: Well, this is, uh, A, this is, like, pretty- Uh, new- Yeah … for me. Um, B- Hot,

[00:46:55] Nathan: hot off the press.

[00:46:56] Dan: Yeah. B, I do think we’re already doing a bunch of stuff that’s like, okay, we have frontier alerts.

[00:47:02] Nathan: Mm-hmm.

[00:47:02] Dan: You know, we’re- Right,

[00:47:03] Nathan: so you’re, you’re, you know, the same way that we, you know, we used to do, like, push notifications in iOS, you’re like, okay, why have push notifications into your Slack-

[00:47:11] Dan: Yeah, into your Slack

[00:47:11] if

[00:47:11] Nathan: you have the Slack

[00:47:12] Dan: integration. We’ve got frontier alerts. We’re gonna have, uh, we’ve got feed- we’ve got a feed on our site, so maybe you’re, you’re checking our site. I mean, I one-chatted an app last night that, like, maybe we should have an app. You know, like- Mm … we’re gonna start exploring in all these different places, and we have

[00:47:25] We build soft- we build, like, real software that, um, a lot of people use and a lot of companies use. So I’m not worried for us. I, but I am … Uh, the way that it, it changes my mind strategically is I’m like, okay, our quarterly goal right now is how many emails we get. I think that’s fine for this quarter, but if I’m thinking about next quarter, is that the number that I- Mm

[00:47:51] is that the number I want? Or, for example, should I focus more on our AEO SEO? Right. Because if you’re ranking really well for AEO right now, there’s a really good chance that that’s going to be a big signal that goes into whether the agent, like, lets you through-

[00:48:08] Nathan: Mm-hmm …

[00:48:09] Dan: eventually. I’m just … I’m making stuff up right now.

[00:48:11] Yeah. But, like, n- that’s the kind of stuff that I’m like, okay, I need to be experimenting with what is the, um, what are the new form factors that people are gonna wanna consume us in, in new channels that are being opened up, Slack, our own feeds, apps, you know? Uh, okay. A b- a b- a big one is everyone’s working in Claude and ChatGPT.

[00:48:35] Now we have an MCP server. Can I get them to install our- Yeah … MCP server and, like, tell them when a new article comes out? And these are all gonna be, like, very early and kind of rough and not gonna work, but in a year it’ll be very different. Yep. And I’ll, I’ll have kind of created … I’ll have done enough experiments to understand what the shape of the right solution is that we can double down on.

[00:48:53] Nathan: Yeah, that makes sense. Can you push content through an MCP?

[00:48:56] Dan: Yeah. Well, well, so MCP servers are, like, the, like, Claude or ChatGPT would pull, would pull the server. Okay. So, like, every day it’s, like, you know, seeing is there anything new? However, really exciting is OpenAI recently introduced this thing called MCP events.

[00:49:13] Nathan: Okay.

[00:49:13] Dan: Um, so you, you can actually push to a person’s ChatGPT. Mm. And I think that’s a huge deal.

[00:49:21] Nathan: Do you think that the other platforms will- Absolutely. Like, if, if it takes off, they’ll adopt it. They’ll

[00:49:24] Dan: do it. Yeah,

[00:49:24] Nathan: yeah. Yeah. That’s such an interesting thing of, like … I mean, it’s fascinating that OpenAI created the closest thing to- magic-

[00:49:33] Dan: Mm

[00:49:33] Nathan: that we have ever seen.

[00:49:34] Dan: Mm-hmm.

[00:49:35] Nathan: And it has almost no mode.

[00:49:36] Dan: Yeah.

[00:49:36] Nathan: Right? Because it’s like, “Oh, this model kind of annoyed me, so let me from a dropdown-” Yeah, yeah. “… select another thing.” You know, and you’re like, “This is the greatest invention that society’s ever had,” and it’s like, “I don’t know, man, I think I’ll just switch over to Gemini.”

[00:49:47] It is

[00:49:47] Dan: a very interesting race to be a part of and to s- and to watch unfold.

[00:49:52] Nathan: Yeah. It’s wild. Before we move on, so in the, thinking about the future of newsletters and all of that- Yeah … is there anything else that you think creators should be doing today to get ready for that?

[00:50:01] Dan: I mean, my, my general advice is Use the tools.

[00:50:06] Nathan: Mm-hmm.

[00:50:07] Dan: Probably don’t use them because you’re afraid. Like, find things- Yeah … to do with them that you’re genuinely like, this- think this is awesome.

[00:50:14] Nathan: Like, what have you always wanted to exist in the world? Yeah. And it’s like, okay, then try to make that happen.

[00:50:17] Dan: If you’re a creative person, now is so cool. And I know that it, you know, I, I love writing.

[00:50:26] Nathan: Mm-hmm.

[00:50:27] Dan: And I love just putting one word after another.

[00:50:31] Nathan: Mm-hmm.

[00:50:31] Dan: And I’m not saying you should just, like, one-shot all your writing and then whatever. It’s like, no. And when I’m five hours into a writing session and I’m like, “Fuck, I have to reorganize this section,” and I can just be like, “Go do it,” that’s really great.

[00:50:45] Nathan: Right.

[00:50:46] Dan: That’s really great, and it means I can do way more and way more interesting things. Like, the level of ambition of this piece is very high. Uh, this c- you know, how Codex happened piece. Right. It’s like, you know, it’s 8,000 words. We have, you know… It’s, there’s all these custom designed elements on the page.

[00:51:01] It’s full-

[00:51:01] Nathan: They animate as you scroll.

[00:51:03] Dan: Yeah. It’s fully agent native, so not only do you have the piece that I wrote, but you have access to all of the underlying interviews that I wrote and all of the like-

[00:51:10] Nathan: Oh, wow …

[00:51:11] Dan: uh, all of the, like, quotes are linked so you can click and see where the, uh, uh, w- you can see the point in the interview where the person said that.

[00:51:19] Um, and then you can have your agent read it and be like, “What did Dan leave out? What did he miss?” So, like, what are the context things that I’m missing? Mm-hmm. Like, all that kind of stuff. Or you can say, “Hey, like, uh, there’s this really good lesson in this piece. Like, can you help me understand it more or bring it back the next time I’m making this decision?”

[00:51:33] Right? So-

[00:51:34] Nathan: Right …

[00:51:35] Dan: there’s all these things now that, that I can do that, uh, A, help me be better at the thing that I love to do and, and are becoming their own craft, but also there’s new opportunities for formats and mediums because, yeah, there’s… Whoever thought of, like, an agent native article? Like, that’s- Yeah

[00:51:53] that has not, never been possible before. There’s all this opportunity to do video and games and it’s just, it’s really cool.

[00:51:59] Nathan: Yeah. Something else that I think creators should focus on is just the highest quality content.

[00:52:04] Dan: Mm.

[00:52:04] Nathan: Like, the kind of thing that will stop the scroll in the feed or, like, this is so good.

[00:52:08] Something that stands out to me about Every as a company is, yes, you have 120,000 people on your newsletter. Um, I bet there’s very few people, like very few creators that you would trade your newsletter list for.

[00:52:21] Dan: Mm.

[00:52:22] Nathan: Who, like, I can think of maybe a couple people that have the quality of a list-

[00:52:26] Dan: Mm …

[00:52:26] Nathan: uh, on par with what you have.

[00:52:29] Dan: Mm-hmm.

[00:52:29] Nathan: Um, we recently turned on subscriber signals for- Yeah … uh, for Every, and going through that with you, it’s just like, I w- I won’t name the specific names, but it’s, like, the CEOs of all of these companies, the top writers, all this. Like, you’ve built a, an insanely high quality audience. Thank

[00:52:45] Dan: you.

[00:52:45] Nathan: And it’s because you’re talking about a topic that everyone is ob- is obsessed with.

[00:52:51] And then you’re also doing it to a level of detail and quality. And so I bet even if the inbox changes or goes away in that way, people are still gonna be seeking out your work in whatever format. So that’s the first thing I would say is, like, focus on, okay, how can you… Like, quality matters more than ever.

[00:53:09] Dan: I totally agree. A- and, and most of what I’m talking about is about for g- for sure, if you are mostly relying on tactics to- Mm-hmm … like, get distribution, like, all the tactics are gonna change. Yeah. Right? But, like, if you’re a great writer, people are gonna find you. Mm-hmm. Um, and even if you’re a great writer, you’re still have to pay attention to some of these shifts- Mm-hmm

[00:53:31] because that’s how people see you, and that’s just sort of- Yeah … the way the game is played.

[00:53:34] Nathan: Hey, if you wanna master the skill of making money, make sure to pre-order my book, The Ladders of Wealth. It’s gonna help you learn exactly what skill you need next to go from earning a living to building wealth.

[00:53:44] The book comes out March 2nd, but if you go to ladders of wealth.com right now and pre-order the hardcover,

[00:53:50] Dan: I will send you the full e-book and the audiobook for free, and I’ll get them to you early. So go to laddersofwealth.com to learn more. I’d love to get you early access.

[00:54:00] Nathan: Are there any opportunities that have come up from having, like, building an audience?

[00:54:03] And I’m thinking about, of a bunch for you but are there any opportunities that have come up from building an audience to this quality that you’re just like, “I can’t believe this person is reading or this opportunity is…”

[00:54:13] Dan: I mean, the whole, my whole life is from this, basically. Right. It’s fucking, it’s cra- it’s really crazy.

[00:54:18] Um, and I get to meet people that are, you know, are heroes or that I’m- Mm-hmm … like, I can’t believe you wanna, you’re asking me for my input on this thing. Right. Like, I’m, I just write a little newsletter or whatever, you know? Um, I feel like there’s this crazy important technology change happening, and I’m just, like, sitting right in the middle of it.

[00:54:41] And not just sitting in the middle of it, I get to, uh, explain it to people who are a little bit further outside, and also interpret it for the people who are even deeper into it than I am, ’cause they need someone to do it. And, like, that’s, it’s an amazing role to play as a writer. That’s, like, what you wanna be doing is, um, as this new world is, is- Unfolding to help people understand how to see it and think about it.

[00:55:09] Nathan: Mm-hmm.

[00:55:09] Dan: And that’s partly a, a, an active observation of just, like, being like, “Here’s what’s here,” and partly a creative act of, like, what could it be or should it be, and how do I see it? And I, I, you know, I think it’s the best.

[00:55:21] Nathan: Well, I think one of the things that I am both wildly excited by and struggle with simultaneously is I feel like I’ve been given these tools that are so powerful that I’m, I’m not even thinking about them in the right way of what’s possible with it.

[00:55:35] Mm-hmm. And so I’m like, “Oh, it can do all of this.” Things that I used to have five engineers work on building or all of this, I’m like, “Oh, it just kind of works right out of the box.”

[00:55:43] Dan: Mm-hmm.

[00:55:43] Nathan: Like, an example is, uh, I always do… You know, I meet creators, I always wanna know what, what newsletter platform are they using?

[00:55:49] What, what’s their tech stack look like? And so, and that’s a hard problem to solve because every good newsletter platform works really hard to prevent bots from signing up for it- Mm … right? To prevent list bombing and all of this stuff, including Kit. But now like, without me even telling it, I didn’t, did not even create a skill for it to do this.

[00:56:08] Um, I mean it helps that, like, GroqBot has, uh, access to our CRMs to be able to see who’s a Kit customer. But it will go on its own, it will go check the sign up, the code on the website. It will check the DNS records to see if there’s an MX record tied to MailChimp or whoever- Mm … or Kit or whoever else.

[00:56:25] And so it just, it knows all of these things, and I feel like I’m constantly, like, asking it to do, like, 201 level tasks. Mm. And, you know, if I, if it was a person, it’d be sitting back and be like, “Bro, you have no idea what I can do.” Like, like, push me to some limits, like- Yeah,

[00:56:41] Dan: yeah …

[00:56:41] Nathan: you know, be way more ambitious and take on some way bigger things.

[00:56:44] Dan: Yeah. Well, I, I get that all the time. We have a, um, we have this, one of our most popular articles, um, it’s called The Eight Levels of AI. It’s by our head, our head of evals, Mike Taylor, and it’s just, it just lays out here’s- your baby prompter to your orchestrating like- Mm … a team of agents. And what people tend to do is just take that and put it in their Rock Bot or their Claude and be like, “What level am I at?”

[00:57:10] Nathan: Mm.

[00:57:11] Dan: And then every week I want you to tell me what are the things I might try to like get to the next level, and works really well.

[00:57:16] Nathan: That’s probably the most important shift, and I’m … if I’m honest, I’m only halfway through making it.

[00:57:20] Dan: Yeah.

[00:57:20] Nathan: The, the, the trap that I tend to fall into is thinking what should I be doing?

[00:57:25] Oh, let me try this.

[00:57:26] Dan: Mm.

[00:57:27] Nathan: And the simple reframe is just saying, “What should I be doing?” Mm. And just ask the AI- Yeah …

[00:57:32] Dan: what

[00:57:32] Nathan: to do. Just

[00:57:32] Dan: ask. Yeah. It’s, it’s like it’s so unsatisfying, because people are like, “How do I do that?” And I’m like, “Just ask.” You know? Yeah. That, that’s why we made this like prompt and this thing, ’cause it’s like- We

[00:57:41] Nathan: were texting about something.

[00:57:43] I can’t remember what it, what it was. And you said, “Did you try asking the godlike-” I’m sure … “super intelligence?” And I was like, “Oh,

[00:57:50] Dan: yeah.” Yeah. Yeah. It’s just like the main skill is just realizing I should just ask this thing.

[00:57:58] Nathan: Right. Which is so bizarre.

[00:58:01] Dan: Yeah.

[00:58:02] Nathan: Um, I wanna dive in a little bit on how Every works as a company.

[00:58:04] Sure. Um, so the business model, you make money in a few different ways. Yeah. As I understand it, you’ve got, uh, the subscription, that’s the main one. Yeah. People pay what, 20 bucks a month?

[00:58:14] Dan: 30.

[00:58:15] Nathan: 30 bucks a month and get everything. Everything. That’s in the name. Yeah. You do some brand sponsorships and- Mm-hmm

[00:58:22] work like that. Uh, you do some consulting.

[00:58:25] Dan: Yep.

[00:58:26] Nathan: What am I missing in there? Is that … Th- those are the main two buckets? We have

[00:58:28] Dan: courses.

[00:58:29] Nathan: Okay.

[00:58:29] Dan: Um, I think that’s most, I think that’s most of the business.

[00:58:33] Nathan: Okay. Yeah.

[00:58:34] Dan: Uh, like the, the core of the business is subscription.

[00:58:36] Nathan: Mm-hmm.

[00:58:37] Dan: And that’s the thing that we try to grow.

[00:58:39] Um, the other stuff matters and has been a really nice cash flow, ’cause we’re like- Mm … we have not raised a ton of money.

[00:58:47] Nathan: Yep.

[00:58:47] Dan: I look at it as it’s … it helps us fund the rest, the subscription flywheel. Yeah.

[00:58:53] Nathan: Something that I’ve been really curious about is like your product studio approach.

[00:58:57] Dan: Hmm.

[00:58:57] Nathan: And you talked about this, um, at Lenny’s conference, and this idea of pirates versus architects.

[00:59:04] Hmm. And I’m so curious how you approach this product development. Um-

[00:59:08] Dan: Yeah, so we, uh, we have a pr- a product team. Um, and they’re responsible for, for example, building the agent. Um, and that’s a product that has customers that like we’re trying to improve all the time, that is … needs to have a stable roadmap.

[00:59:22] Nathan: Mm-hmm.

[00:59:22] Dan: The frontier’s always moving, and our job is to be at the frontier. And for a while I was like, “It’s everyone’s job to be at the frontier,” so we all always have to be experimenting and, oh my God, this new thing came out, like let’s push it into the agent or like let’s push it into whatever, whatever other product we have.

[00:59:41] And that is very chaotic. There’s something, like, important about that, which is if we never incorporated the new stuff, we’ll wake up in six months and be like, “Uh-oh.”

[00:59:52] Nathan: We don’t have a business. ”

[00:59:53] Dan: That’s bad.” Yeah.

[00:59:54] Nathan: Yeah.

[00:59:54] Dan: Um, but it can’t be, like, everyone’s job to be constantly playing with new stuff ’cause n- nothing will get done.

[01:00:01] It’s, it’s chaotic, it’s confusing. Um, and some people love to play, play with new stuff, some people don’t. Both people are valuable So what we did is, uh, I established a frontier team. Frontier’s job, just play with new stuff all the time. Constantly be testing new models, building experiments. Um, and then we have a product team who’s like, “We’re gonna make the best possible fucking product.”

[01:00:27] And what we’re trying to do is move things from the frontier to the product team as they get more tested out and more, like, proven. And so the frontier team stuff, it always just starts as experiments that basically… Or they’re, they’re content, um, that tell us about, like, is this new little feature something that we might wanna do?

[01:00:48] So, like, a, a good example is, um, there’s this model that came out called JEV, which- Yeah … it just does really, really fast classific- classification. Um, and the, the frontier team immediately… Like, the day it came out, we had, like, this big vibe check and, like, a bunch of pe- a bunch of people on the frontier team made feeds and made all these, like, demos and whatever.

[01:01:06] And, um, then someone on the agent team was like, “We should use that to make frontier alerts.” So, like, you just sort of-

[01:01:14] Nathan: Oh, to decide if- You,

[01:01:15] Dan: you, like, push it … who it

[01:01:16] Nathan: should go to.

[01:01:16] Dan: Yeah. We, we just sort of… We say, “Here’s, here’s what’s possible now,” and then the product team can be like, “Ooh, that actually works really well for what we need.”

[01:01:23] The

[01:01:23] Nathan: product team, like, gets to select from the frontier team-

[01:01:25] Dan: Exactly …

[01:01:26] Nathan: of, oh, that rises- Yeah … to what

[01:01:27] Dan: should go to- Yeah. And, and, you know, there’s a thousand things that the frontier team has done that the product team has totally ignored. Which is fine. Yeah. That’s, like, part of it, you know? And I think that gives you a little bit of the best of both worlds of we’re exploring, but we’re also exploiting, and we’re not being distracted over here, but we are, like, creatively, like, on fire over here.

[01:01:46] So

[01:01:46] Nathan: before you were taking each person and saying, “I need you to carve out 25, 50% of your time- Yeah … I need you to live in both worlds.” Yeah. And you switched that and you split it up and said- Yeah … “You’re over here, you’re

[01:01:56] Dan: over there.” Yeah, exactly. And then within the frontier team, when you’re exploring something new, I think there are two main archetypes of people.

[01:02:03] There’s pirates and architects. Pirates are like, “I’m just gonna, like, go do a bunch of stuff. That’s me.” You know? Yeah. Um, and architects are like, “Okay, you have something that’s, like, kind of working, but it’s, like, extremely messy and un- unfocused. Like, I’m gonna try to, like, help you narrow it in into something that, like, works well.”

[01:02:19] So architects help you move it up the, up the frontier, up, up the idea pipeline stack from something that’s, like, totally messy, completely vibe coded, to something that, like, you could look at it and see, oh, maybe more of the team should be using this.

[01:02:35] Nathan: Right.

[01:02:35] Dan: Um, and then the product people can look at it and be like, “Oh, maybe we should release to customers.”

[01:02:40] Nathan: Mm-hmm.

[01:02:41] Dan: Yeah.

[01:02:41] Nathan: Yeah. I think that’s really interesting. Something that I wanna do more of i- at Kit is build some of these, like, one-off features, ideas. Part of the reason we built our, our app store is to make the product totally extensible.

[01:02:54] Dan: Mm.

[01:02:55] Nathan: And so now you can vibe code. You would not wanna vibe code the core app that sends 3 billion emails a month and has, uh, you know, hundreds of thousand plus active users.

[01:03:06] And we can’t let every random person into our code base. But with the App Store, you know, we have all of the, the templates and the tutorials and all that, so anyone can vibe code an app. And they’re like, “You know what? I want this functionality. It should exist.” And so I, I prompted it and it does exist now.

[01:03:19] Um, but that’s allowed us to test and iterate a lot more. But it’s very easy to do the the, like, pirate that work. And we’re quite good at the scalable work. Mm-hmm. But the in between is hard, and so that’s like, that architect role makes a lot of sense to me. But what I think about is the ongoing support- Mm

[01:03:42] of each one of these things. Yeah. And this is a problem that I haven’t solved, is like- Yeah … oh, I could easily… Like Kit does not have an iOS app.

[01:03:47] Dan: Yeah.

[01:03:48] Nathan: We should.

[01:03:49] Dan: Yeah.

[01:03:49] Nathan: But I’m like, “Oh, now I have this steady stream of- Yeah … tickets forever.”

[01:03:53] Dan: Yeah.

[01:03:53] Nathan: And I don’t know how to solve that.

[01:03:54] Dan: We have the same thing. Um, the, the way that we have solved it, and it’s not perfect, but the way we’ve solved it is anything we launch as an experiment, we don’t support.

[01:04:06] Nathan: Mm.

[01:04:06] Dan: Like, you should know.

[01:04:08] Nathan: Like it’s in the labs.

[01:04:09] Dan: Yeah. We’re launching this, we’re never gonna support it. Mm. Don’t submit a ticket. Um, ’cause we’re just gonna, like, we’re just trying stuff, and people know that. Yeah. And that’s the deal, and it’s, like, clearly marked. And then when something feels proven out enough that we’re gonna invest, that’s when we’re like, “Yes, we’re gonna support this.”

[01:04:23] Mm. It’s, like, an ongoing part of the business and all that kind of stuff- Mm … and we’re not gonna, like, uh, just let it, let it

[01:04:28] Nathan: ride. So the

[01:04:30] Dan: deliberate classification of- Deliberate classification …

[01:04:31] Nathan: this is an experiment or in the labs.

[01:04:32] Dan: I do also think that in the next six months for apps that you want to keep just, like, kind of working- Mm

[01:04:42] in maintenance mode, the cost is gonna be very low.

[01:04:44] Nathan: Yeah.

[01:04:45] Dan: Um, it, you know, it already is pretty low, but there’s a lot of, there’s a lot of work you probably need to do to, like, set that up. But six t- six months to a year, it’d be pretty low to just, like, go from email about a bug to an agent has triaged and fixed it, and then emailed- Yeah

[01:05:01] the customer, um, for, you know, simple apps. Uh, so-

[01:05:06] Nathan: You’re right. At that screen size it does have that issue.

[01:05:09] Dan: Yeah, exactly. Yeah.

[01:05:11] Nathan: That’s fixed now.

[01:05:11] Dan: So I, I, I don’t think the ongoing maintenance costs are gonna be that h- high, but the more of these things you have, the more messy it is. Like, things break- Right

[01:05:20] even, you know, so it’s, it still will be something that you’re thinking about, and the best way to deal with it is to be deliberate about… well, make a distinction between we’re releasing this and this is just an experiment, versus we’re actually gonna support this, and be deliberate about when you change something from an experiment to something you’re supporting.

[01:05:37] Nathan: Right. Yeah, that makes sense. I’m gonna play with that more because I mean, we’re already doing that inside of Kit where, you know, so many of these pull requests, it will… Like, when you, the engineer sits down to start on a task, especially if it’s a bug fix or something like that, there’s already a PR waiting for them.

[01:05:54] Yeah. And they just have to decide, is that how I wanna fix it- Yeah … or something else? And often they’re like, “Well, that’s 90% there. You know, here’s a little tweak. Go for it.”

[01:06:03] Dan: Yeah.

[01:06:04] Nathan: Okay. You said before we sat down that you have some hot takes. I, we, we got one of them- I

[01:06:09] Dan: spent my hot take . And the- Gemini’s a great model

[01:06:12] Gemini’s a great model. Gemini’s a great model for writing.

[01:06:14] Nathan: Okay.

[01:06:14] Dan: Yeah, great model for writing.

[01:06:15] Nathan: Yeah.

[01:06:16] Dan: People will find it to be silly that we were worried that AI would take all of our jobs and that- … humans would not be use- useful for anything anymore.

[01:06:26] Nathan: Mm.

[01:06:28] Dan: I think that would be very silly in hindsight. Not to say that it won’t change jobs or, you know- Right

[01:06:33] historically, lots of technology changes have, like, eliminated certain job roles. Uh, for example, if I was, uh, if I was an Uber driver, this is … We’ve been saying this for a long, long, long time. Right. So it’s gonna take a long time, but going in a Waymo in San Francisco, I’m like, “That thing’s pretty good.”

[01:06:49] Nathan: Yep.

[01:06:49] Dan: Um, and, uh, and I think that’s very important that it’s talked about and dealt, and dealt with, um, because it’s a big part of people’s live- livelihoods. And it feels obviously silly to me that we’re, we’re like, “Oh, yeah, only, like, three people will have jobs and everyone else will be the permanent underclass,” and like, whatever.

[01:07:10] Like, it’s just not… That’s just, uh, uh… W- We will look back at that and be like, “Oh, we were so naive. Like, those models weren’t even that good,” you know? Right. The models we have now, though, Those are so good.

[01:07:23] Nathan: Well, I think a, a fascinating thing is if you look at Andrew Carnegie’s story-

[01:07:28] Dan: Mm …

[01:07:29] Nathan: so his father was a professional weaver-

[01:07:32] Dan: Mm

[01:07:32] Nathan: in London. Mm. Or in, in England.

[01:07:34] Dan: Mm.

[01:07:35] Nathan: And, you know, like, a tradesman doing all this stuff, and then the loom gets invented and he’s out of work. And so they go from the middle class-

[01:07:43] Dan: Yep …

[01:07:43] Nathan: to, uh, quite poor. Yeah. And they end up immigrating to the United States.

[01:07:47] Dan: Mm.

[01:07:47] Nathan: And so you get someone who’s, like, at the, uh, like, gets the worst end of a technological innovation- Yeah

[01:07:53] and then says, “Cool, so I’m gonna take what I learned and I think I’ll just become, like, one of the richest men in America.” Does that work? You know? And you’re like… And so, like, often th- these disruptions and all of that- Yeah … result in some pretty crazy changes.

[01:08:06] Dan: Yeah.

[01:08:07] Nathan: All right. I love those hot takes. So final thing that I wanna ask is what are you reading?

[01:08:12] You’re someone who is on the cutting edge of everything. Mm-hmm. And then I also know that you, you’ve always loved philosophy. Mm. You’ve always loved writing and the craft of it.

[01:08:22] Dan: Yeah.

[01:08:22] Nathan: What’s, what’s the reading that’s fueling your future?

[01:08:26] Dan: I’m currently reading Wind, Sand and Stars by, um, Antoine de Saint Exupery.

[01:08:33] He’s …

[01:08:33] Nathan: Oh, yeah?

[01:08:34] Dan: Um, the Little Prince guy.

[01:08:36] Nathan: Yep.

[01:08:36] Dan: Very good. It’s his memoir.

[01:08:38] Nathan: Okay.

[01:08:39] Dan: And I think it’s, it’s about his experience being a pilot for the mail, like the- Mm … Aeropostale, like the French mail service in, in the Sahara in Africa. Um, and, um, I mean, there’s a lot, there’s a lot in it that is about-

[01:09:03] I have this, I’m in charge of delivering the mail and this new technology that people think is, uh, separating me from nature and the problems of man- Right … and actually it is plunging me into, like, nature and the problems of man. Yeah. And I think he has that, that sensibility, so I re- I really like that.

[01:09:19] Nathan: I’m gonna have to read that.

[01:09:20] Dan: It’s, it’s, it’s pretty good. It’s pretty good. Um, other than that, I’m in, like, a l- in a little bit of, like, I’m in between a couple things. I just finished Les Miserables.

[01:09:29] Nathan: Okay.

[01:09:30] Dan: Um, which I thought was actually … It was really good. I- Hugo is a little bit, um … He’s pretty over the top.

[01:09:37] Okay . And he spends, like, 40 pages, like, talking about French slang, French c- criminal slang, which to be honest, I could get on board with, and it’s a l- it’s a lot. Yeah . And I think also his, a lot of his, like, love scenes are a little … I think French, maybe just French love scenes are a- Yeah … are, are kind of a lot.

[01:09:58] But, um, but it’s, it’s, it’s quite, it’s quite a beautiful book also. Um- That’s awesome. So I try to, like, mix between philosophy, fiction, memoir, kind of-

[01:10:08] Nathan: Yep.

[01:10:09] Dan: Yeah.

[01:10:09] Nathan: All of that. Uh, I love y- your approach of being on the cutting edge of everything, and then saying, like, “Hey, let’s, let- let’s also stick to the classics.”

[01:10:16] Dan: Yeah, that’s what I’m- That’s sort of the kind of new and old thing that I was, I was trying to point to earlier, that- Yeah … they’re both very important to me.

[01:10:23] Nathan: I like that.

[01:10:24] Dan: Yeah.

[01:10:24] Nathan: Okay, if you were to recommend one article that you or the Every team have written, someone wants to dive in, if a creator listening wants to dive into-

[01:10:32] Dan: Hmm

[01:10:33] Nathan: your world, where would you start?

[01:10:34] Dan: I mean, you gotta do the Eight Levels of AI. Okay. Yeah. Um, so just search. We have a guide. Um, and you maybe put it in the, in the- Yeah, we’ll put it in the show notes … in the show, show notes. Um, and then I wrote this piece called After Automation, which is, like, why will there be really great human work to do after automation?

[01:10:51] Um, what is, what do we see for, for ourselves? And then why is that structurally true? That I think is, uh, a good thing for you if you’re, like, trying to get into the mind space of, oh, actually these tools can help me do my work better.

[01:11:05] Nathan: Yeah. I love that. Okay, so check out those two things. Go check out every.to.

[01:11:10] See everything that you’re doing. And then, uh, probably follow you on all the social channels.

[01:11:16] Dan: Definitely.

[01:11:17] Nathan: I think X is … Is that where you’re spending most of your time? That’s the

[01:11:19] Dan: big one. Yep.

[01:11:19] Nathan: Okay. Just Dan Shipper on X? @

[01:11:20] Dan: dan shipper.

[01:11:21] Nathan: Perfect.

[01:11:22] Dan: Thanks so much for coming on. Thank you.

[01:11:23] Nathan: In the episode, you heard Dan say that he writes by just talking to AI.

[01:11:28] If this is something you wanna learn more about and possibly do yourself, check out episode 101 with Jay Singh. Jay helped our team at Kit automate workflows with AI voice agents, and in that episode, he builds a working voice agent from scratch live on the show. Like the video if you enjoyed it, hit subscribe on YouTube or wherever you’re watching, and I will see you next week.