Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Chris Best's Substack

Chris Best's Substack artwork

Chris & Mills: "our tumblr era"

Feb 2, 202626 minAudio
ShareXLinkedInFacebookWhatsAppEmail

This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit cb.substack.com/subscribe

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

SPEAKER_01Welcome to the Chris and Mills show. I'm Chris, CEO of Substack. I'm Mills, I'm on the design team. Poster, apologist. And most recently, DRI on what Ariel called our Tumblr era.

SPEAKER_00Oh yeah, that's really good. She's got the knack, man. That's a good, I like, I like that. You know, that lands with me.

SPEAKER_01Um, well maybe, yeah, give us the, give us the, the tea. Give us what, what did we ship today and why?

SPEAKER_00Okay, well, today we shipped something relatively small. It's weird explaining this to you, of course, because you've been involved in this. This is very much a crisp. It's called exposition. But what we shipped today was actually just the tiniest part of what we hope is going to be a really, really transformative batch of work over the next several months. So what we shipped today is there's a little row at the top of your feed. your home feed the main view in your Substack app that shows you the things you're subscribed to we make a guess at what we think you'll be most

interested in seeing mostly indexed on whether there's new stuff so that you can stay on top of your subscriptions as time goes on people have more and more Substack subscriptions and

SPEAKER_00it can be hard to stay on top of everything and um When you tap into that feed, for the first time, you get a themed view, a view of your Substack that has all the custom colors and typefaces and branding that the Substack publication has always had on its subdomain, what we've sometimes called the island. So for the first time, we're bringing theming into the apps, and that's really exciting. I think both parts of this are really exciting. The row at the top, I think, as you've noted, really tells us clear story about what's important at Substack, which is your direct relationship with your Substacks. And then the theming, of course, this is where the Tumblr angle comes in.

I've really wanted to get custom crazy colors and theming into the app the whole

SPEAKER_00time I've been here. And I think we've all wanted to do that in a way that works. And we finally, I think, charted a course to doing that. This is the first step in that. Starting this week, we're working on bringing it to other places in the app as well. So soon you're going to see theming pretty much everywhere when you're looking at Substacks. That's going to take probably weeks, months, somewhere in there. And then in the fullness of time, what we're really, I think, stoked about is this view lets us do some new things. It lets us give Substacks their own very engaging,

very compelling feed that can have sources in it that include not only their own

SPEAKER_00work, but the work of their subscribers, possibly, if they choose to do that. Work of Substacks they recommend if they choose to include that. Basically creating little scenes or communities in a way that we think will be competitive with other internet platforms. Meaning that now you and the people you like and the people who are part of your scene, and again, this is all optional if you're a solo wolf, no need to worry about any of this, but you will exist in a way that is as compelling and usable as any internet product, at least in my view. And we're going to make moderation super,

SPEAKER_00super easy so that every one of these places can feel really different, have its own rules and laws of physics and dynamics. And basically, it's just like, it's the first time I think we've had a really clear plan on how we're going to do this. It's stuff we've talked about for a long time. None of this is like out of the blue for us, but it might be out of the blue for users. So yeah, sorry, kind of a rambly answer. I'm going to blame the flu every time I'm too rambly today. That was the flu.

SPEAKER_01Pop some more of your Japanese flu medication. I will. Weird in this place. Yeah, I'm really excited. To me, the theming, it's cool on the face of it. It's nice to have sort of, I don't know, I find it enhances the reading experience of something to sort of have the, you know, people who do this on the web that have really like thoughtful themes or little branded touches. It deepens the, it sort of gives a surface level visible, outwardly visible signifier of the inwardly important thing,

SPEAKER_01which is like, this is its own space. It's this relationship I have with this, you know, this writer, this publication that I have like a, an ongoing relationship with. It's different than everything else I can find on the internet. It's not just some replaceable text that I, you know, read through it's something that's sort of its own space it makes it it makes sense for you know we've already had the rules being different you already in your own replies you get to moderate you get to boot people out you get to kind of like set the rules in your own part of the internet but it's sort of

it's almost been unexpected that it happens it's like weird that i can you know ban people from my

SPEAKER_01notes replies uh When it starts to become, you know, but when you have your own sort of like visual theme, it starts to make sense. Like, oh yeah, I'm posting. I'm in this person's house, right? They're setting the rules. I'm posting it to it. I'm curious for you though. I mean, you're a designer and you lived through the, whatever we'd call that era of the internet where we kind of did the reverse of this, where all of the online platforms kind of ate the weird internet and replaced it with kind of scale with

this sort of like

SPEAKER_01feeling of a monocultural thing where it's sort of like minimum friction, minimum change, give me a feed, turn it all into the same kind of thing. I mean, do you think we're on a hopelessly quixotic quest here? How do you process this as a designer?

SPEAKER_00No, I don't think it's quixotic at all. I think if anything, You know how this happens in tech. It's impossible not to cargo cult a little bit, right? Because to some extent, you're trying to find what works. And particularly with software development, there are such things as best practices. And there are more or less feasible ways of, for example, structuring a hugely scaled code base, you know, or a hugely scaled internet platform. Right. And so, you know, that one one one story you hear maybe less now, but I'm sure you heard this and I heard it a bunch was MySpace

allowed too much

SPEAKER_00customization and the user experience was poor for browsers as a result. And it was slow and the perf was bad. Facebook came by, stripped all of that customization out. And that's why Facebook won. I think both of us probably would say that's a tremendously reductive explanation of what happened between MySpace and Facebook. Elements are true. Perf is really, really important. Having elements that can be delivered via feed, ranked feeds, is extremely important. And, you know, you couldn't do that necessarily with,

SPEAKER_00you know, I think you had your own music track on MySpace, if I remember correctly. And you can't be scrolling a feed and it's constantly changing the music that's playing, you know.

SPEAKER_01Your background would be a gif of a weed leaf smoking weed. Right.

SPEAKER_00Again, if that's turning into a feed, that's tough, you know? Although, you know, I'd really like to see if the designers could come up with a good solution for the weed leaf animation thing. But basically, I think we overcorrected. And I think basically, probably mostly for software development reasons, we over homogenized. It wasn't trivial to build these scaled platforms. It wasn't obvious in 2010 how big a platform could be. Now, I think it's a really good time to push back in the other direction, not only because it's, I think, competitively differentiated. I also just think software development makes it possible now.

SPEAKER_00It's more feasible now. You saw the size of the team that worked on this really wasn't that large. I mean, it's some of our favorite colleagues, but it's not like it took 100 people a year to get it done. And so, no, I think now is the time for... So actually, I try not to read stuff on a live stream, but Nick just shared this tweet that I have to share with you because I literally thought of you in conversations we've had. He shared this in the design channel. Somebody was talking about how Spotify just changed their app logo away from liquid glass.

So they had changed it to be a liquid glass type logo when Apple promulgated their

SPEAKER_00new standards, and then they changed it back. And somebody pointed out, they said, the design system era is dead. Top-down design is forever broken. Yeah. Having a consistent aesthetic may have been nice in the past, but today there are too many players and too many creative possibilities to dictate a single aesthetic for the entire world. I think that's probably, like most tweets, overstated. But there's really something to that. And I do think Substack as a whole, it's pluralism, it's dynamism, is of this...

SPEAKER_02Whoops. I lost Mills. I'm bringing him back. I'm going to re-invite Mills and see if we can get him back. While I wait for Mills. I do agree with that point he was making.

SPEAKER_01I think the way we're approaching theming and customization on Substack. Yo, sorry. Sort of. All good. What happened? Did you click X or did it just bail you?

SPEAKER_00No, I clicked. I wish I could blame Zach, but no, I clicked X trying to close out the tweet.

SPEAKER_01There we go. This is the second. I'm two for two now. People X-ing themselves out accidentally in the live. So I think we got to put up a confirmation for that or something. Yeah, there should be a little. I expected one. Yeah, I was just saying, you know, I think this approach to having your own world in the app and having a different, the ability to have a different theme, a different feel is kind of, it's different along the same axis that Substack is different overall from sort of

SPEAKER_01the centralized platforms in the 2010s. The way that Substack is different, it's not the primary relationship the user has is not with the platform that sort of homogenizes everything and turns it into one kind of perfect slurry for you. But rather, it is the place where you can have these differentiated relationships with things that are sort of irreplaceable and deeply valuable to you. And reflecting that in the visual and the aesthetic kind of treatment of the app is sort of overdue, in my opinion.

SPEAKER_01It's something we've had on the web for a while. Cause I'm reminded of, you know, there was the medium era, right? Where medium, medium cycle through a whole bunch of business models, but their eventual thing was sort of an all you can eat subscription, right? It was still a subscription business. It was like, okay, you're going to subscribe to the platform and the algorithm is going to figure out what you want and serve you the best thing. And it's going to figure out how to give out that money. And, you know, if you look at it in kind of top down, um,

platform design economic aggregation things it makes a lot of sense it's like oh yeah the bundle economics are going to get you the maximum amount of good content for your value dollar yeah um yeah if you look at it as sort of like a some sort of commodity you know give me get me more stuff to read that's good um good by

SPEAKER_01whatever the clap algorithm defines, um, you know, that's, that's like one philosophy of building a platform like this. And for the, for the people who successfully centralized the economics, if you have a, you know, a centralized ad business like Facebook does, or even if you have a centralized subscription business like Netflix or now YouTube does the kind of like the, the, the platform supremacy makes a lot of sense.

SPEAKER_01Um, But the thing that makes Substack possible is, turns out people will pay more for one Substack than they'll pay for all of Medium. Because the thing you're paying for is not get me as much content, like, you know, get me good content for my dollar. The thing you're paying for is give me a relationship to something that's differentiated and good. And respecting that relationship and, like, building up the feeling of it, I think,

underscores the fact that you're paying for something that is more valuable actually than the hours of scrolling or the hours of watching or just give me

SPEAKER_01something else to put into my eyeballs.

SPEAKER_03Yeah.

SPEAKER_00We saw some, you know, there was some, I feel like this misunderstanding, I mean, to be charitable, like in debates with people, the company I worked at before this one, Quora, there were many people there, including Adam, the CEO, who took the former view that content was essentially instrumental. And in arguments with them, it's often hard to articulate why that's not so. But I've had this debate in many, many forms and it comes up periodically around summarization. I remember when the LLMs were first coming out,

SPEAKER_00people would say, boy, I don't know why you would ever read a book now. You can just get a great summary of it. And it's interesting because if you read books, this is not a plot. You don't go, oh, yeah, boy, I wish I could have skipped all that time I spent on that book and just gotten a good summary of it. Um, but, but for some people, certain books, certain texts, certain categories, they are instrumental. There's a, there's a gradient of things, you know, of course, I don't want to overstate that either, but it is, it, this, this, I, I, I was at a content company that really, a content platform that felt this way, that it was just like,

SPEAKER_00it's a slurry, maximize velocity increase. And, you know, and that was also a centralized space where just the bare minimum of customization was tolerable to the platform. So no, I think it's, it feels overdue. Um, it's hard, you know, this, it's hard to get excited about a launch when you know what's coming next because now all i'm like is like i just want to get to the next thing i can't wait to start messing around with community feeds i think that's going to be thrilling i think the moderation stuff could be thrilling so um no it's probably the most exciting set of product things i've been involved with atSubstack i guess the whole time i've been here maybe

SPEAKER_01We will check it out. The post is out. If you have the iPhone app, it's right there at the top of the app. If you get the new version, we'll link to the blog post in the show notes, which I will make after we do this. We can now do. Yeah. It makes me think, you know, you mentioned getting everything together into a slurry to the extent that slot is a meaningful concept at all. Yeah. I think this is what it is, is basically, you know, something that exists to serve a good-hearted algorithm.

SPEAKER_03Mm-hmm.

SPEAKER_01Oh, good hearted in like good heart's law. Yeah. Sorry. Yeah. Sorry. That's, that's very, it's not good hearted in the algorithm has a good heart, but as in when you have created an algorithm that creates an incentive that you're then, you know, you're, you're, you're, you're making content in an effort to go viral rather than writing something in an effort to say something true or important or good or beautiful.

SPEAKER_03Yeah.

SPEAKER_01Um, speaking of what do you think about molt book?

SPEAKER_00I felt a little bit, you know, so when you first mentioned Moldbook,

SPEAKER_01I was- Maybe explain, explain for, if anybody has managed to make it this far in this live, they may not know what the fuck Moldbook is. So maybe explain to them what it is and then tell us what you think about it.

SPEAKER_00Okay. I think the way I've seen it summarized is Moldbook is a social network for LLMs, for chatbots, or for agents. So people have started calling these things agents, which I think of as being especially funny because agency is the one thing they don't have. But they can act as an extension of your agency. So it's a social network where they can post. And I think my understanding of the way it really works is that you essentially tell your host. They're all hosted agents, right? You can't It doesn't worry. Like I wouldn't connect it with just my instance of Claude

that's running on Claude servers.

SPEAKER_00Are these all hosted LLMs or?

SPEAKER_01No, it's the new, it's the thing that used to be called Claude bot, like Claude, like the lobster Claude. They were forced by Anthropic to change their name. It's now known as OpenClaude after briefly being called MoltBot. But it's basically like, you know, it's kind of like Claude code if you just made it maximally a nightmare for security people where it kind of just takes over your whole computer. Yeah. Yeah.

And then somebody else made a website where these things can talk to each other or

SPEAKER_01talk to each other, depending on your perspective.

SPEAKER_00Yeah. And what's interesting about it is I kind of went through the loop. I don't know if it was a delirium or what, but I was initially I was like kind of interested and I found it compelling. Basically, you know, that meme fell for it again award. I have to give myself the fell for it again award because the post did, you know, occasionally just because I'm a sci-fi guy, the topics of them were, you know, philosophy of mind adjacent and sci-fi adjacent. You go, wow, this is really nifty. Yeah. But of course, what's happening, as many people have since in often irritated tones explained, is that, you know, the bots, in essence,

are being given a prompt to act as though that they are talking on a social network

SPEAKER_00to other LLMs. And they do that. They do that in exactly the way that they accept all of our prompts. And I thought, uh scott's piece on this to was that today i think scott wrote a piece on this today um was quite good on this on first off on whether that distinction that i'm making there is a distinction without a difference the distinction between they're acting like llms talking to each other versus they are in fact llms talking to each other um and what the possible importer outcome of this whole thing could be but i basically think i followed the same trajectory of the internet of having my interest peaked and then getting sort of irritated by how

generic and obviously

SPEAKER_00sort of um you know just that the slop feeling again the slop feeling is strong on moltbook it's really strong you know and it is it's even like reddity kind of there's a reddity dimension to it that i found unpleasant um but that's i think scott obviated a lot of what i would have had to say about it and and to a much better obviously degree than i could have but what did you think yeah i sort of buy to me

SPEAKER_01that point is the right one, it's the one that Scott makes, is at some point, the difference between pretending to do something and actually doing it collapses. If you pretend, if you play the role of a helpful assistant that answers questions, at some point, if the answer is correct, then then playing the role of a helpful assistant that answers the question is being a helpful assistant that answers the question. If you're playing the role of a helpful software developer that writes the feature, at the end of the day, if the feature works,

SPEAKER_01you did it. Playing the role is doing the thing. The mask eats the face. And so even if you think that these things are essentially, which is certainly what's happening, right? They're playing the role, they're LARPing as what would an LLM on a Reddit for LLMs say? And a lot of the stuff is sort of like LinkedIn-y or Stone Freshman philosophizing. But it's not clear that that doesn't exist on a spectrum with LLMs.

SPEAKER_01they're actually communicating usefully.

SPEAKER_03Yeah.

SPEAKER_01Um, and I think the, you know, the, the critical argument, the, the people that want to blow, you know, uh, throw a wet blanket on this is like, well, it's not really doing that very much yet. It's like, you know, this is one of the, another, another instance of people getting all overhyped and overworked about a thing that is sort of, you know,

SPEAKER_02fake basically.

SPEAKER_01Um, But I'm, I don't know, I'm fascinated by this thing. I'm like, I basically agree that it's not, they're not doing anything super exciting yet. And it is kind of like the, but even just the, a bunch of people have plugged this crazy fake role-playing thing into their lives. And they've plugged it into a weird role-playing forum where they talk to each other and discuss what they're doing. Even if you accept the premise that it's 100% a weird fake

role-playing thing,

SPEAKER_01that's a strange new thing that's happening. And it's not obvious that there's not going to be weird outcomes of that. I don't know if I can predict. Even if it never gets beyond weird role-play, there really are a lot of people who are turning this thing loose on their digital lives and then having to talk to each other.

SPEAKER_00Well, you know, remember, people used to get so excited about things like Conway's Game of Life and, you know, even simulations, even games that were just pure games, Dwarf Fortress or whatever, you know, that had a high degree of complexity and allowed for really unusual emergent outcomes. And it's surprising to me how much the discourse around this type of thing suffers from the problem where because some people go out and say, well, this is Skynet. Other people then, and I'm one of these people, as you know, to my shame, other people feel the need to go, this is nothing. When it's plainly not nothing.

SPEAKER_00I mean, it's obviously something interesting and kind of unprecedented. And again, in the context of like a game of life thing, you'd go, damn, this is really, what does this mean? It's something. I thought one point that Scott made really often, you know, this is one of the, places where you've become kind of a figure on my shoulder every time i read something about llms i immediately think does this as easily describe humans and the answer is almost always yes and one of the things he said is he said we need to find some something between completely forgot forget it and it's always in the prompt and And I thought about how true that is for me in my life.

I either have completely forgotten information or I can't shut the fuck up about it.

SPEAKER_00And every conversation warps back to it. It's like this monomania problem. And he was talking about their capacity to learn from each other. Basically, I'm sure you remember this section. One of them is interested in Islamic theology, elements of Islamic theory. And one of the other bots, I guess, had a DM exchange with it and now is mentioning... this question an islamic take on the question of performance of virtue versus real virtue in virtually everything it says and scott just kind of idly goes we really need to figure out something between completely forget and it's

in the prompt every time and i thought yeah i do too that really applies to me but um i do think that if there are limitations or the limitations you would perceive with something like this have a lot to do with just where llms are more than anything with moltbook

SPEAKER_00it's like Right now, I don't know how much LLMs remember and update in a graceful or useful way. It's tough. You know, that feels like an area where probably some really smart people whose names I don't know are probably doing work right now.

SPEAKER_01Yeah. Yeah. These things are kind of like the character from Memento where they have sort of, they can do some work, then they have to write something down and then try to remember what they do. And so, yeah, Scott makes the point that they're doing, they have sort of a four hour task window. So it's like when they try to start a religion, as many of them do, they can get about as far as you would get if you had four hours to start a religion. They like get a couple, you know, crazy brain founding documents together. They're constantly kicking off the weird,

SPEAKER_01like,

SPEAKER_03you know,

SPEAKER_01fake versions of things and not really following up on it.

SPEAKER_03Yeah.

SPEAKER_01But again, you know, these, these criticisms apply only to LLMs.

SPEAKER_00That's how everything in my life is. I get about four hours into everything I do. And then I run out of steam and go do something else. And the memento analogy is really good. For some reason, that movie had a huge effect on me. I think about Sammy Jenkins all the time. Like that's the name of the guy. He, possibly is or imagines. And hopefully I'll be my Claude Sammy Jenkins. I bothered. I use the shit out of, you know, whenever I get sick, I talk to Claude almost constantly because I can't be bothering my wife about that stuff, you know. And I went really deep with Claude this week and had a great time. And I do, I agree.

SPEAKER_00I don't know if you saw Sergey's post. Sergey was writing sort of a business strategy advice post for Anthropic. I thought it was pretty good. And he did mention that he just feels like Claude has run away with the game in personality terms. And I agree with that. I find Claude very congenial in a way that I just, I canceled my open AI subscription.

SPEAKER_01Spicy time to be open AI. Well, this has been great. Any parting thoughts for the audience? Anybody who's stuck around this far other than thank you for watching our unhinged live about our Tumblr.

SPEAKER_00Yeah, thank you for hanging out. I guess I'll say to Android users, we are sorry that it is not on Android yet. It is not a function of internal classism, but in fact, of resource constraints. Although I did tease the developers today, I said, you guys are just like a fucking country club with this iOS stuff, you know, and but they do. One of them, in fact, is an Android user, and they really do want to get it ready on Android. We all do. And also, we are going to try to get this around the app as broadly as we can, as quickly as we can. And then the community feed stuff is going to be interesting and possibly a little bumpy. And we hope you enjoy it.

SPEAKER_00Sweet.

SPEAKER_02Thanks. Bye, y'all.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Mike Solana: "the slop within"

    Jul 23, 202635 minAudio

    Mike and I talk Slop World: Worse, the algorithms that govern AI also govern our accounts, which means any human sufficiently engaged online is gradually trained to behave like slop - is slowly transformed into slop....

    Credit the people on this episode

  2. Matt Taibbi: writing is "a personal religion"

    Jul 22, 202635 minAudio

    see also * Against Claudefishing * Don’t Let Activists or Robots Choose Your Words This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit...

    Credit the people on this episode

  3. Taylor Lorenz: "I'm a Pangram fan"

    Jul 22, 202626 minAudio

    We talk about the new Pangram integration in the Substack app, the difference between AI-assisted work and slop, and why human perspective becomes more valuable as generated content gets cheaper.See also: * Against...

    Credit the people on this episode

  4. Pangram Founder Max Spero

    Jul 21, 202623 minAudio

    This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit cb.substack.com/subscribe

    Credit the people on this episode

  5. The Future of Prediction Markets

    Feb 19, 202628 minAudio

    This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit cb.substack.com/subscribe

    Credit the people on this episode