Special episode: Apps vs Files with Gordon Brander, Peter van Hardenberg & Jess Martin.

December 31, 2024

This is a special episode which was recorded as a Twitter space conversation between Gordon Brander, Peter Van Hardenberg, and Jess Martin.
This conversation explores the ideas and trade offs of apps versus files.

Mentioned in podcast

Links:

Transcript

Introduction

00:00Welcome back to the local first FM podcast.

00:02We're currently on holiday break, but we've prepared a special episode for you,

00:06which was recorded as a Twitter space conversation between Gordon Brander,

00:10Peter Van Hardenberg, and Jess Martin.

00:13This conversation explores the ideas and trade offs of apps versus files.

00:17I hope you like it.

00:19All right.

00:19I guess we're going to get started here pretty soon.

00:21So my thinking on this would feel something like we've all been at a

00:25conference and we've heard an interesting talk, maybe enlightening talk even,

00:29where the person shared like five minutes of thoughts, but not even like

00:33a full, they didn't fully elaborate on everything they wanted to say.

00:37And the two of us said, Hey, you want to go sit down at a conference

00:39table and talk through this?

00:40Cause I feel like there's some interesting stuff here.

00:42And I have thoughts on it, and I would like to work them out, out loud.

00:46So that's kind of my framing.

00:48I love that.

00:49We are all, I think, building in one way or another, around this problem as well.

00:54Is that right, Jess?

00:55Yeah, more or less.

00:56Yeah.

00:56Do you want to maybe, like, recap what the motivating question was?

01:01Absolutely.

01:01the reason this is called File, Symbol, App, File over App is, Kepano from

01:07Obsidian posted a little, I guess a Twitter post that also he has, I

01:12guess there is a link on his website to a fuller article about this.

01:16And it's a philosophy that, Kepano holds.

01:18And then the tool called Obsidian, also embodies this philosophy.

01:23And, I'm not gonna go into the whole thing now.

01:26I'm gonna kind of assume that most people are, can like, go and read the, the post,

01:31and kind of get his thoughts, more fully.

01:34But I, the basic idea is that the tools that we use should store that data.

01:42In a way that we own the data, and he calls that files, and

01:47that we, and that the data lives beyond the application itself.

01:51So once we've, like, even if the app goes away, we've still got the file,

01:57and that file gives us some ability to, continue interacting with that

02:02data that we've spent such a hard time, or had such a long time creating.

02:07That's the, that was the impetus.

02:09And I think the reason that I wanted to chat about it and reached out to

02:14Gordon and Peter reached out as well, is because I mostly agree with what

02:18Kepano is saying, and yet there are some subtleties at which I, I think

02:22that probably we would, if Kepano was here, that we would, we would part ways.

02:28So I'm curious to kind of work out where the subtleties are and

02:32then talk about, you know, some real systems that embody this.

Files vs. Apps

02:35Well, let me, let me make the opposing case because I think it's important

02:38to have both perspectives here, which is files got their asses kicked.

02:43Right?

02:43Like, you want to email people your Google Docs?

02:47That, that's the case here?

02:48Oh, I get it.

02:49I get it.

02:49Yeah.

02:49You don't want to be able to share a link to a thing.

02:52You want to have to send it and then choose a file name for it and get it back.

02:56And then, something, something, merging, something, question mark?

03:00Make a small change, change the name of the file, send it back,

03:03final Underscore, final, March

03:057th, 24th, final, final, as submitted, dot, doc, dot, you

03:11know, word, x, dot, whatever.

03:13so I mean, files got beat because they're worse.

03:15What people want is places, not things.

03:17And so, okay, now I'm going to, I'm going to undermine my argument,

03:20but like the desirable properties of a Google doc are that a Google

03:24doc is a place, not a thing, right?

03:26So when you join a Google doc, you are given a key.

03:30To a place where you can go and collaborate.

03:34Now, everything that's wrong with Google docs also falls out of this, I think.

03:38But like, what's great about a file is also what's bad about a file.

03:41What's great about a file is that it's a thing that you have, you can't take it

03:44away from somebody else who has it, but also they can't take it away from you.

03:48Right.

03:48They have a copy.

03:49They can edit it privately.

03:50Right.

03:51They can share it themselves.

03:52They can decide what they want to do with it.

03:54But like, there are lots of power dynamics as well in terms of

03:57like, places versus things, right.

04:00Which is like, if you've given someone a thing, it's hard to get it back.

04:03But if you let someone into a place, you retain perimeter control.

04:06You can choose who's let in, in the future, if they're allowed back.

04:10Right.

04:11And while there's sort of a fiction about like actually retaining control over

04:14data, like You know, I can screenshot it.

04:16I can copy paste it.

04:17I can exfiltrate any number of ways.

04:20You know, that's like an extra step I have to take versus if I email

04:23you a PDF, you've just got it.

04:25So I think like, that's the, that's the beauty of the place is that it

04:28gives people, it gives the person who creates the data a lot more agency

04:32over its destiny, and it gives you a natural collaboration point,

04:38you know, kind of by definition.

04:39Yeah.

04:40And I think, I know that Gordon was.

04:43Wanting to talk about collaboration as well as like a natural place where files

04:46kind of fall down But I think one of the things that you're mentioning here One

04:50of the things that I think is included in the file over app that isn't mentioned

04:54explicitly is the file system, right?

04:57It's the it's the container for the file of like where does that file live?

05:02What is the place at which it actually lives because it does have a place.

05:05It's just not a Internet accessible place via URL.

05:08and I think that is, that's actually part of Kipano's argument is that

05:13the place where the file lives is a better place than the cloud.

05:17He's saying the file lives on a local file system on a device that you

05:21control, like a hard drive that is yours.

05:24And that makes it.

05:25in some place, in some sense is a special place or a better

05:28place than some of the cloud.

05:30And that's what you were mentioning as well, Peter, that like the cloud places

05:34enable, certain things, file system shaped spaces, places enable certain things.

05:40Like, how can we get the best of both of those worlds?

05:42Oh, well, that's a different thing, right?

05:44You're claiming like, also, this is a little bit of a category error, right?

05:47Like you're arguing, Like Apple over refrigerator.

05:51Oh, between file and app?

05:52You're talking about Kepano or the argument that I just made?

05:54Well, just the, the sort of framing of the discussion, right?

05:57Like, there are plenty of apps that work on files.

06:00Right?

06:00It's, it's not The problem here is, Like, apps do not imply the death of files.

06:06But there is a culture and a community set of values and certain

06:10sets of expectations, right?

06:12Like, are we talking about mobile apps or web apps?

06:15You know, there, this is a, like a, I think there's a, there's a lot of

06:18articulation points potentially here.

06:20For sure.

06:21And I think the only point that I was making to try and simplify

06:23that, that first point is just where does the thing live is whether the,

06:28whether the file or the app, is the file living on my local device?

06:33Is it living in the cloud?

06:35how do I get to it?

06:36Do I have to have a network connection to get to it?

06:37Can I get to it, locally without a network connection?

06:40I think those are all wrapped into Kipano's thinking about

06:44the philosophy of File Over App.

06:45Yeah, I guess, Kipano and, and all of us are kind of, you know, thinking through

Interoperability

06:51I think he's making a pretty strong rhetorical point.

06:54I guess I, I prefer to look at this in terms of design goals and

06:58kind of the qualities of different materials that you can work with

07:01rather than what's good or what's bad.

07:03Though I am personally, I'm very drawn to capturing my notes in files myself.

07:09You know, I've got 10 years worth of notes, you know, in a folder that syncs,

07:14and I've used plenty of SaaS tools over the years, some of which have gone out

07:19of business and I've lost those thoughts.

07:21So when I think about, you know, why files, why, why is

07:24Kapano interested in files?

07:26you know, there's some areas that I resonate with and I, you know,

07:29I think files, to your point, they live in a place that you control.

07:35the file system is a sort of user on substrate.

07:39It's old.

07:41It was designed in, you know, the sixties and seventies and didn't

07:44anticipate a lot of the things that we ended up doing with networks.

07:48It has a lot of problems around security and conflicts and many

07:53other things that we can get into.

07:54But I guess the one thing it does do.

07:57That the web tends not to do and many apps tend not to do is it does

08:01leave you in control of that content.

08:03And that's kind of interesting because it means often you can

08:06open that data in multiple apps.

08:10no one can

08:11tell you no.

08:13That the opening in multiple apps was the other side of this thing that

08:16I think is like, not actually true.

08:21Except in a few very narrow cases, which are the ones that

08:24programmers live in, but that much of the rest of the world does not.

08:28Well, you know, much of the rest of the world doesn't live on desktop

08:32machines with file systems anymore.

08:34I think, again, it's just, it's not about it is or it isn't.

08:37It's about, there are different jobs that people are solving.

08:40I think, you know, there are actually a number of things

08:42that open up in multiple apps.

08:44They tend to have strong power laws.

08:46Like there's only a handful that end up.

08:49working out in this way, but I think that it's better when you have this

08:52sort of user owned substrate where those use cases can emerge, permissionlessly.

08:58Well, the thing is,

08:59like, I'm reflecting on like the file format wars of the nineties, right.

09:03And like the EU legislation to try and like force Microsoft to open,

09:08you know, the office suites file formats to other participants.

09:11And, you know, I don't think that was terribly successful, but, you

09:14know, that, that this is, um, just because you have the file doesn't mean

09:18that you really own it in a sense.

09:20Like there are certain properties of it that are quite desirable.

09:23And I think this longevity is, is really remarkable and a really

09:26excellent point to bring up, but yeah, there's a tension here between like,

09:31open file does not imply openness.

09:33Sure, though that can emerge retroactively.

09:36Like we saw that happen with PSD and with, you know, the doc format.

09:40Like PDFs were proprietary, right?

09:42Like they I was going to say PDFs

09:43the same way.

09:44So the thing that lets that emerge is that, that substrate

09:48that belongs to the user.

09:48If it was up to the app, if it was up to Photoshop, they would say, no,

09:52a competitor can't open up our file.

09:54but they don't get to make that choice.

09:55I can open it up and, you know, Affinity Designer or in.

09:59some sort of other app.

10:00And we saw the same thing happen around image formats, around

10:03video formats, around 3d formats.

10:06They tend to be things that are pretty low level and have like very broad

10:09expressive power so that there's a wide range of jobs to be done.

10:13That people want to, to do with them.

10:16I also think there's like a Lindy effect around these things.

10:18They consolidate around a few, like text, plain text has a lot of problems,

10:22but I fully expect it's going to be here in, you know, 10 years, 20 years,

10:26maybe even a hundred years, define it in whatever way you like, but I,

10:30I suspect whatever way you define it, it's going to last longer than, you

10:34know, the, the Notion API, like notion.

10:37com.

10:37Oh, certainly.

10:37Yeah, yeah.

10:39Yeah, I guess I'm just pointing out plain text.

10:40It's, I often think of it as a fixed point, but.

10:43You know, if you look back at, the history of plain text, right, we go

10:46from like ASCII encoding and having to, you know, I don't know if you

10:50were ever opened up like, fileid.

10:53diz from like Eastern European countries, which had different ASCII encodings than

10:57your, you know, Windows computer did, back in the like late eighties, early

11:00nineties, but you get these sort of,

11:02Yeah, plain text across Windows and Mac had, different file endings.

11:07That's right, yeah.

11:08You would hand a text file back and forth, and it would, you

11:10wouldn't have line breaks where you expected them to be, or you'd have

11:13additional line break characters.

11:14Control M symbols, yeah.

11:16no, that's, I thought, there was, there was a couple areas, like,

11:20I think, Gordon made the point that Kapano's making a really

11:23interesting rhetorical argument here.

11:25I love that he picks a fight here, and that he has a very, it's

11:28a really nice mantra, but he's definitely cheating a little bit.

11:31He cheats files as if they don't have an encoding, like, as if, like,

11:35plain text is some given thing, but plain text is an encoding.

11:40that we had to, like, as you come to adopt it widely enough that there are a

11:44large set of readers for that file format.

11:47And I think that when you get beyond, I think whether it's

11:51programmer usage or not, it's when you get beyond the document format.

11:54That's when things start to get, this argument is more

11:58difficult to, not as clear.

12:01because, you know, I, like canvases, for example, are everywhere now.

12:06And there's no single file format for exporting the contents of a canvas,

12:12whether it's Muse export or tldraw has the tldr file, all of these things, they're

12:17not compatible with one another, it feels like the early days of text files,

12:20like BDH was just mentioning, where you've got a bunch of different formats.

12:25And the weird part about saying file over app is that you still

12:28need a reader for the file.

12:30Like you have to have some code somewhere that reads that file.

12:34And so it saying files are better than apps or files or

12:37should be preferred over apps.

12:39It's like, well, a file without an app is literally useless.

12:42Like you have to have an app to, manipulate that file in some way.

12:46Can you define what you mean?

12:48Like the word app is, is carrying a lot of water here.

12:50Can you, can you unpack that word a bit?

12:53Sure.

12:54I think when I simplify file over app, I think what he's saying is data over logic.

12:58Okay.

12:59In other words, the, the data or data over behavior, the data stored at rest in some

13:05way that you can carry it into the future, own it, copy it from device to device,

13:09that's his idea for files, I believe.

13:11and he's making that, I think he's making, And then for app, he's saying,

13:18he's basically, I feel like collapsing, apps into databases, basically, of

13:24like databases that you don't own.

13:25you can't take your individual records from that database and,

13:29and do something with them.

13:30They're stored in this, mash of things.

13:33at least that's why I understood it is like data over the application logic.

13:36The term we use at Ink & Switch is tools, not apps.

13:39Can you expand on that a little bit?

13:41the concept is that an app in its sort of cultural construction, though

13:47not necessarily in its technical requirements, generally implies that,

13:51you know, it holds the data inside it.

13:53I think mobile apps and web apps both have different kind of habits here, right?

13:57Like, For the most part, a mobile app is a hermetically sealed box.

14:01Nothing gets in or out of it.

14:03You can't open your Apple Notes in Dropbox or, you know, Notepad

14:06or, you know, anything else.

14:08That's not necessarily true, but it is practically, culturally the case.

14:12And so the idea is that tools operate on some kind of material, right?

14:16And so Obsidian has strong tool culture.

14:19Right?

14:19Like one of the interesting things with Obsidian and reflecting on

14:22Gordon's comments is just that like Markdown and like plain, Obsidian

14:26doesn't operate on plain text.

14:28It operates on Markdown, but there's no such thing as Markdown.

14:31There's like as many Markdowns as there are apps that use some grammar

14:35of Markdown, but you get like a certain amount of interoperability

14:39because it's like these really nice, graceful degradation patterns.

14:43Yeah.

14:43Markdown is like an escalator.

14:45It can't break.

14:45It can only become text.

14:47Right.

14:48So like, it's kind of an inspired choice from that perspective, but I think your,

14:53your point is well taken like these formats, they tend to layer and it sort

14:59of necessarily like interoperability is an ecological condition.

15:02It's something that emerges over time.

15:04You don't design it.

15:05This is often, this, this is actually why for me, sort of permissionless

15:08substrates, substrates that belong to the user where their data saved

15:12seems important because it allows interoperability to emerge retroactively.

15:18But in any case, like these things, they tend to emerge over time where,

15:23you know, one person creates a format, another kind of app implements it.

15:28it doesn't happen often, but when it does happen, it tends to be very valuable.

15:32It's often oppositional though, isn't it?

15:34It's often oppositional and often happens in layers.

15:37Like I think that, you know, text is on top of, you know, bytes and

15:41then markdown is on top of text.

15:43And then we have like Obsidian layering their own sort of like bespoke

15:47markup forms on top of markdown.

15:50And, and these layers tend to move at different speeds with the lowest layers

15:54kind of necessarily moving slower.

15:56Like the part of the problem that's standardized and that people can

15:59interoperate around is necessarily the part that has to move slowly.

16:03Well markdown's not standardized, right?

16:05Like it's that's an emergent Like cultural phenomenon.

16:09It's de facto, right?

16:10Like there's sort of flavors that people converge on, like GitHub flavored.

16:14And you know, like all, all standards are sort of cultural

16:17constructions at some level.

16:18Yeah, you're right.

16:19It's, it's not like standardized in like the W3C or IETF sense.

16:22Or even in the Unix POSIX sense, right?

16:24Like files are.

16:25You know, these kinds of bytes and these kinds of orders, and

16:28this is how you end a file.

16:30But there, there are a couple of emerging standards, right?

16:32We're at the, we're at the situation where there's multiple plain

16:35text formats of Markdown, right?

16:37There's Common Mark and different parsers that you can adopt that give you.

16:41true, guarantees around the way that markdowns, but there's as

16:44you said, there's no one Yeah,

16:46this this actually seems important to me Like I think a lot of people think

16:50about standards and interoperability as these sort of platonic forms that we

16:53design in standards bodies And you know, while that's true that there are people

16:57who craft that kind of material it's like Typically done in retrospect and

17:02in real life cooperation usually happens in the sort of ad hoc or de facto way.

17:07And when our systems aren't designed to support that, it's not a huge surprise

17:12that, you know, interoperability doesn't emerge and then we can sort of

17:14point and say, well, nobody wants to interoperate and it's like, well, yeah.

17:17Cause the material doesn't support it.

17:19Like, like there's a kind of chicken or the egg problem here.

17:21Right.

17:22I think both in your work and Jess's work and our work, right?

17:26We're all exploring.

17:28Different ways of exposing the substrate in more interoperable ways.

17:33Yeah.

17:33And it's not so much like locking the door open as like, right now, I think

17:39like the political economy of choosing files over these other things, right?

17:43Like if it's so much harder and worse for you as a developer to choose, Like a

17:48file like thing over an API like thing.

17:52You're probably going to pick the API like thing.

17:54Indeed.

17:54Some of this is like technical capability and some of it is cultural norms.

17:58Yeah.

17:59Like files are just the thing that's there.

18:01And they're actually in many ways, an odd.

18:03odd shaped thing for the problems we're trying to solve, but they're

18:07just, uh, at least in the app store and on the web, there aren't really

18:13good, viable alternatives yet.

18:15Like in the browser, we're starting to get things like the file API and I, you know,

18:20on, on iOS, we're starting to get like a little bit of a sort of file system.

18:25So it was very charming when Apple excitedly announced

18:28files like they'd invested in.

18:29Yeah, totally.

18:30Right.

18:30But I, I, it's actually kind of telling that like, why, why is

18:34it actually that we want this?

18:36It seems like a lot of knowledge work that

18:41needs to cross boundaries of multiple apps.

18:44and this is just the only tool we currently have to do it.

18:47Mostly because there haven't, I think, haven't been good

18:49commercial incentives to build.

18:51Right.

18:52Interoperability layers that, that makes sense on the network, at least not

18:56yet.

18:57Well, you still need a data container, right?

18:59Like it's one of the nice things that files make is like this, you know,

19:02relatively, this thing that has boundaries that contains a certain amount of data

19:07that you can then pass around via a whole bunch of different, units of

19:11sharing.

19:12Yeah, exactly.

19:14So I, I'm actually curious.

19:15So for people who are listening, who don't know, Peter's work, he's working

19:20on a, CRDT, which is, basically a data structure you can think of it's called

19:24Automerge and Automerge actually calls their, unit of, merging a document.

19:31So.

19:32Peter, do you guys think of documents, Automerge documents, as files in this

19:37file like way we've been describing, or do you think of them in a different way?

19:41The specific stated intention of Automerge is to invent the

19:45new file system of computing.

19:48Like, the CRDT thing is just like an implementation detail.

19:51The problem is, We don't want places and we don't want things.

19:54We need properties of both, right?

19:56There's a wave particle duality here.

19:58We don't want to give us the good things about files, things that we have,

20:02things that we can hold, things we can maintain, things we can keep for as long

20:05as we want that we can use in whatever way that we want, we can pop open the

20:09hood and peer under the covers, right?

20:10Like all that stuff is, is I think vital, but like, it's so apparent that the POSIX

20:18metaphor of the 19th, literally 1970s.

20:22Right.

20:22The series of bytes, it assumes single access, right?

20:25Like, God, the amount of my career I've spent trying to figure

20:28out how to lock a file system.

20:31There's so many, so many tricks that, that we use in like professional

20:36contexts to cope with shared

20:37mutable state, shared mutable state.

20:39And the whole idea behind Automerge was to say like, look, what if you could.

20:43Really think about this, really take your time, really sweat all the details, not

20:47be in the rush that a startup's in, not, you know, spend years with Martin Klepman

20:52proving the algorithm correctly, then spend years with, you know, Orion and

20:56Alex Good getting the performance right.

20:57And people are like, Oh, Automerge is this, Automerge is that, Automerge

21:00is just getting started, right?

21:02Like this is a, this is a 10 year project.

21:04It's a 50 year project.

21:05It's, it's a thing that you just keep working on until it works.

21:09And I think at this point a lot of the pieces are in place and we're

21:11starting to see that pay out and if anybody's following Ink and Switch's

21:14patchwork project right now, right, like what we're seeing is like this

21:19ability to do ad hoc, like new forms of real time and asynchronous version

21:25control on top of this substrate because it has the right primitives.

21:30And of course, this is a research project, so in the process of working on this we're

21:34discovering the ways that the APIs are insufficient or where we need to add new

21:38kinds of indexing to improve performance or, you know, all those other kinds of

21:42things, but it's sort of like we, We've talked to so many people, you know, I

21:46was talking to, someone over at Scratch the other day and they were saying like,

21:49oh, I did this in my thesis, but like, it wasn't actually possible to build

21:53because we didn't have the material.

21:54Right.

21:55And like a former collaborator, Blaine Cook, who was a coauthor on

21:58the Upwelling paper we did last year, you know, he was saying like, it's so

22:01amazing getting to work on this stuff.

22:02He did a startup five years ago or something.

22:05Where you tried to build, you know, text editing tools, but like

22:07they didn't have the material.

22:09And so it's about trying to figure out what the shape of that material is.

22:12And you know, I, I hadn't brought it up before because I don't want this to

22:17be about a particular like concept that I have or that we have about a single

22:21technology, but it's about like really thinking about what the opportunities

22:25are and the potential, but I think a lot about this question, right?

22:28Like I don't want my computer to be a mall where I go in and I take

22:34products out one at a time, you know, and bring them home and consume

22:38them and put them in the trash.

22:39I want my computer to feel like a wood shop where I have my stock of lumber and

22:44I've got my tools and if I want to make something, sure it's going to require a

22:47little more thinking and expertise, but I can make anything, I can do anything,

22:51I can work with other people, right?

22:53Like it's, it's a different kind of productivity.

22:56Like a kitchen is very much the same thing as well, right?

22:58We buy groceries, we buy staples, you know, and a handful of like basic

23:02tools, cutting board, a heat source, a knife, you can do almost anything in

23:07the kitchen with a few extra things.

23:09You can do remarkable things.

23:10Right.

23:11And like, you know, we, we reject the idea of unit task devices.

23:15Often we make fun of people who buy the one off gadgets for the

23:18kitchen, but then we live in a world of these one off gadgets on our

23:23devices, you know, on our computers.

23:25I'm trying to think like, if, if we're gonna invent tools, we also need to

23:29figure out what the stock material is.

23:31And like Yeah.

23:32Text files are cool 'cause they're old and, and like what's old is good.

23:36Like I don't mean that in a dismissive way.

23:38Like it, it is cool because, and it stood the test of time.

23:40Exactly.

23:40It's, and it will stand the test of time.

23:42It'll still be here in 50 years if it was here in 50 years.

23:45For 50 years SQLite is amazing too.

23:47Simon Wilson's super inspiring.

23:49They're working on, Like data set and these kinds of ideas where

23:52it's like, yeah, like a database is a file like that's awesome.

23:56Yeah

23:57It's more or less written by actual monks as well It's it really does feel

24:03like a kind of cathedral of technology.

24:05That'll be here in a thousand years Yeah,

24:10I think that's, I love that woodshop metaphor and it's often one that I

24:16think about, it, it's so evocative because if you've worked in a woodshop

24:20and you understand all the things that you're, the, the ways that

24:24you're able to modify your shop and your tools in order to get work done.

24:28And then you just put that same lens and look at your computer.

24:32You realize how incredibly rigid, the computer is, even when you're a software

24:36developer who has, an ideological commitment to using open source

24:41software, where you have the source of every single application that you use.

24:46The difficulty in modifying those applications to your own

24:50custom specifications is, is astronomical compared to working

24:54models and metaphors.

24:56Mm

24:56hmm.

24:56Yeah, the new materials you're talking about.

24:58It's great.

24:59I was reading, uh, Brian Kernighan's biography.

25:02it looks like the C programming language, but it's his, his self published thing.

25:06Unix, a history and a memoir.

25:08And, um, it's a funny book.

25:10But there's some really great parts in it where he talks about sort of like the

25:13invention of Unix pipelines and how it was sort of like somebody would ask how you

25:18could do X with a computer, you know, or with this new pipes thing they'd built.

25:22And then, you know, him or Richie or somebody would go off and, like,

25:26write the next little piece, you know, sort or whatever, so that

25:29they could answer that puzzle.

25:31And I just love this sort of, like, they discovered the Lego bricks, they

25:34you know, they built the shape and then they just filled in a few pieces and

25:37then after a little while it was kind of like, yeah, all the major pieces

25:40are there and that was good enough and they stuck you know, it's still

25:43out there and you can still use it.

25:44But it's not really the best way to do spellcheck anymore, you know?

25:48Such a glorious vision though.

25:50You know, I guess for me, like files are not the thing, but what they are,

25:54at least for particular formats is like a kind of Lego dot and producing Lego

25:59dots like this that allow you to compose things together, like Unix pipe is

26:03another kind of Lego dot this strikes me as important in like a laws of physics.

26:08Kind of way.

26:09Like, one of my favorite books,W.

26:10Brian Arthur's the nature of technology.

26:14I always get it confused with Kevin Kelly's book, but the, the nature of

26:17technology, he basically makes this, this very sort of, mathematically grounded

26:21argument that innovation at its core is a combinatorial process and not just

26:26like technological innovation, but, you know, evolution, language, art, like

26:31these are about taking pieces, composing them together to produce something

26:36new in a sort of recursive process.

26:39And like early computing was all kind of like, I don't know if this was

26:43intuitive or chance or what, but all of these early pioneers of computing

26:47designed their systems in this way.

26:49And we saw this kind of Cambrian explosion of really interesting

26:53new things emerge very rapidly.

26:55Like actually we've had mobile phones, for Longer than it took the browser to be

27:01invented after the introduction of DOS.

27:03So it's just like, like a lot happened in like a very short

27:06amount of time in that first wave.

27:08Right.

27:08And that's part of me kind of wonders that if, We sort of reduced the

27:12combinatoriality of technology to actually slow the rate of innovation.

27:16not necessarily like there's someone scheming in a room, but sort of the

27:20purpose of the system is what it does.

27:22And like, like big money, but also kind of governments and people.

27:26It's eternal September, right?

27:28Yeah.

27:28It's like

27:29if

27:30you have a bunch of silos and those silos do one thing, like on Instagram,

27:34you post photos and on TikTok, you post short videos and on, like you, There is

27:39no such thing as combining those into some new medium because they don't let you.

27:44And so you kind of have these known quantities that you can regulate

27:47and control and also extract 30 percent taxes on pretty efficiently.

27:52So there's this kind of like striation that's happened of the space.

27:56Some of this is probably inevitable, but I do kind of want to click the dial a couple

28:00notches in the other direction, just by

28:02nature.

28:03I kind of think this is, the way you describe that is, to me, a lost cause.

28:08Or maybe lost cause isn't actually the right term for it.

28:11What I mean is that, actually, a television is great if you

28:15want to watch Love is Blind.

28:17Is everything necessarily a television?

28:20No.

28:20Well, this is, this is the point.

28:22And so what I think though is like,

28:25I don't think we should be trying to convince the mainstream systems

28:28to change their minds or solve the problems we want to solve.

28:31It's that we need to, you know, like a woodshop is not the right way to build

28:35an IKEA, you know, assembly line, right?

28:38Like, but we need the existence of this other kind of computing.

28:41It is, it is not.

28:43a replacement for, it is an alternative to, for the people

28:47who want it and need it.

28:48Yeah, we don't disagree.

28:49Like there's always going to be Facebook, and there's always going to be like the

28:52Big Bang Theory or whatever, but I would love for there also be like a world where

28:57like there can be, you know, I don't know, independent films or hell, like,

29:01uh, Dune II is like something that's, kind of in the air and kind of strikes, that

29:05balance in a nice and interesting way.

29:07It just feels like Like what we want is diversity.

29:10We want like a technological ecosystem that can actually support

29:14a plurality of, of different ends and not just like Walmart.

29:19Yeah.

29:19And well, and the political economy is the challenge here, right?

29:21Which is that right now it costs so goddamn much to write a piece of software.

29:26It's like completely untenable to build or operate software at all.

29:29And I say this as somebody who has done quite a lot of it at all

29:33scales, you know, in production.

29:35It is just insane.

29:37We think of software as an informational good, right?

29:39Like that's sort of the like conceptual model.

29:42Oh, it's just some bits on, you know, floating around, but it's not, it's

29:45like a whole, like massive amount of extremely expensive infrastructure that's

29:50operated by a, you know, a priesthood of like DevOps people who sacrifice

29:56themselves to get up at three in the morning to figure out why the transaction

29:59wraparound on the Postgres has happened again, you know, like it's hell.

30:03It's crazy.

30:04Yeah.

30:04The strange economics of, software, I guess, support that

30:08kind of, like, rising floor.

30:10Because if you do have a hit, it tends to make a ton of money and

30:13then all of that's a rounding error.

30:15Yeah.

30:15It's the

30:15network

30:16effects.

30:16The complexity of software is determined by the economic value that

30:19can be extracted from the software.

30:21Yeah.

30:22Yeah.

30:22And that, that sort of sets the floor as well.

30:24Right?

30:25Like if you're, if you're making not thousands, but billions on a success,

30:29then you can afford, you know,

30:31Yeah.

30:31To burn out a bunch of, software engineers and, you know, leave them

30:35staggering around, their hometown, gazing bleary eyed into the morning light and

30:40wondering where their life went wrong.

30:42At the same time, like these economic incentives, they create individual

30:46choices for like, individual developers.

30:48And I think that's, I think that's where a lot of the slowdown has happened is

30:53when you, there aren't, there weren't options if you were into computers in

30:58the sixties to go make whatever the sixties equivalent of 450, 000 a year.

31:04and kind of at a cushy job where they.

31:06You know, bring food to your table and you can get up and take

31:09foosball breaks whenever you want.

31:10like the, the cruise ship mentality, I think, lures a lot of people away.

31:16and I think it's really hard to kind of wake them up to the fact that

31:21software doesn't have to be this way.

31:22We might have to do a little work.

31:24We might have to be a little uncomfortable for a little while.

31:28You got

31:28to make the right thing.

31:29The easy thing.

31:30That was the lesson we learned at Heroku.

31:32If you make the right thing, the easy thing, people will do it.

31:34All the, you know, all of the principles and ideals, like, yeah, there's a

31:39certain amount of cultural affinity and there'll always be some, like, weirdos

31:43and early adopters, but ultimately if you want to win the day, you have to

31:45make the right thing the easy thing.

31:47But you need some people doing that, making the right thing, the easy thing.

31:51And I think that's what I'm talking about is like, I want to find that 2 percent

31:55or 3 percent of the developer population who actually wants to do the work and

31:58make the right thing, the easy thing.

32:00and that, and for that, you need to talk to people.

32:02You know, you need to have, these wonderful mantras that, Kepano and

32:05that, um, Peter and, Gordon, you guys have honed over the years.

32:10actually, Gordon, I, to turn the conversation a bit to your work,

32:14because you are working on a, network tool, network thinking tool that

32:19captures those thoughts in documents, files, one might say, and, You

32:24have not chosen to use Markdown.

32:27You've chosen, you've made some interesting technical choices.

32:30So I'm curious how, when you, when you look at file over app, are you like,

32:34yes, that, that aligns with my, The decisions that I'm making right now

32:39for the tool that I'm building, or how are your thoughts different from kind

32:43of what you've seen Kapano lay out?

32:45Yeah, that's a good question.

32:46I guess what we are building is a sort of notebook app called Subconscious

32:50powered by this decentralized note graph protocol, and the protocol is multiplayer.

32:56So the notion is like, what if you had your Roam graph or your Obsidian graph,

33:00but it could sort of freely network with other graphs in this cloud of graphs.

33:05But that cloud of graphs doesn't live on my server where I can sort of, you know,

33:10extract a 30 percent text or whatever.

33:12It exists on this decentralized network.

33:14I actually, that's not quite right either.

33:16The protocol is local-first.

33:18so basically it syncs your slice of the graph, like your content plus the content

33:22of people you follow onto your devices.

33:24So you have your files.

33:25and then this was clear,

33:27right?

33:27They're, they are actually files.

33:29They're actually files.

33:30If you

33:30run orb on a, or what you guys used to call it, orb.

33:34Yeah, it's called orb.

33:34Yeah, no, it's called orb.

33:35There's, so there's a couple of things, right?

33:37There's the app, but there's also like a command line interface.

33:41You can use if you're a developer, that's like git.

33:43And, and actually it works a lot like git.

33:45The whole thing uses kind of git like, versioning semantics.

33:48the difference is that it syncs not just your code, Quote unquote

33:52repository of notes, but also kind of the, those are the people you follow.

33:56So you have your own little internet archive, I guess.

33:58Right.

33:58You like Git, uh, it's like Git sub modules or that's a terrible example,

34:01but it's like, No, don't say

34:03that.

34:03That's mean.

34:03It's not like that.

34:04I promise.

34:05It's, it's, it's all, it's much, much better than that.

34:07It's all versioned together.

34:09I mean, I, yeah, without getting deeply into the details, it's

Data Ownership

34:13This is part of, I guess, what's.

34:15That's what's a little bit different from something like Git or why you can't

34:18just use Git, but right, you know, I guess a lot of what we're doing, yeah,

34:23at some point during your usage of this, you will end up with actual

34:27files on your device, like, yeah, very different than most other systems that

34:32we interact with right now, Obsidian being one of the ones that you do

34:35end up with files on your device.

34:37But that's a choice you guys made.

34:39The files are, I guess you could say they're a lens over

34:42the data on the protocol.

34:44And, for me being able to get a live synced, instance of those files

34:49on a device is sort of proof of, of what, what we call credible exit.

34:53So this is like our minimal definition of what Sort of user ownership of data should

34:59mean for our tools for the way we choose to build for it It's like you should be

35:03able to take your stuff and leave or bring it into another app or whatever Now I

35:07think Obsidian actually has credible exit as well And they do this through through

35:10files, but I think they have some maybe tricky edge cases around multiplayer And

35:16this is because, you know, f files, one of the shortcomings of files is they don't,

35:20they're, they just don't do multiplayer really well on, on two axes, actually.

35:24One is, most file formats are not designed to sort of freely merge updates

35:30from a bunch of different sources, like Peter edits the file, I edit the file,

35:35how do we resolve those conflicts?

35:36The other sort of axis around which file systems are not really well designed

35:42for multiplayer is around security.

35:44so once you have multiple players, it's sort of an adversarial environment,

35:49or you have to design for that anyway.

35:51You have to predict

35:52the emergence of an adversarial environment.

35:54Yeah, because, you know, imagine you're reading and writing these files and

35:58then you just connect this thing to the internet, what if 4chan suddenly

36:01has write access to your hard drive?

36:02Like that's the, that's the worst case that you need to the mistakes

36:06Scuttlebutt had too, right?

36:07Which is they did this like replicating friends of friends thing and then you

36:10only had to compromise one person's friend graph and all of a sudden

36:14you were, you know, you could sort of sneak someone into the graph and

36:17then have everybody else back up your objectionable content for you.

36:21Right, which is, and then this became a major problem because everything was

36:24append only and everything was replicated and everything was encrypted and all of

36:29a sudden you're downloading gigabytes of like terrible hate speech or other

36:33kinds of nasty stuff because, you know, somebody made like a face account,

36:38you know, infiltrated a community and then like, you know, linked in all

36:42their like hateful people that they want to follow and now everybody else

36:45is like stuck mirroring their content.

36:47I guess, yeah, you know, again, this, this system of replicating

36:54friends of friends content.

36:56Is very easily gamed by an adversary.

37:00Yeah, I guess it's important to kind of define your, Risk profile.

37:07Yeah, yeah.

37:07The threat model.

37:08Yeah.

37:09Like is your threat model that, like about replicating friends of friends content?

37:13I like, I think, browsers for example, are even coping with this in a more

37:17pedestrian, not a decentralized way, but like they're introducing the, the

37:21file API, which actually will allow you to save files out from the browser in a

37:25sort of, so you could go to photoshop.

37:27com, edit a photo and that saves it out to your disk.

37:29And this is something I worked on when I was at Chrome a

37:32little bit, and it has a lot of.

37:35Thorny issues.

37:36Actually, most of the browsers only partially support it right now or

37:40support, you know, certain parts of it because the way the web secures content

37:46around domains is very different.

37:49sort of comes in at an odd angle to the way file systems think about security.

37:54And so anytime you have these two sort of security models budding up against

37:57each other, you have this turbulence zone where all sorts of funny, scary things

38:01can happen and you need to deal with that.

38:03So this is actually, this is a pretty tricky problem.

38:05You know, it's one that we've, taken a stab at for our use case.

38:09but I think often you'll run into this dilemma that isn't immediately

38:13obvious, but you can sort of, typically you can either have.

38:16sort of files or, or, or data ownership or multiplayer.

38:21but not both, unless you start picking apart some of the security problems

38:27that the web has, you know, sort of left us with some of that security stack.

38:32Gordon, when you say, unless you pick apart some of the security

38:34problems, what I hear you saying is like, you have to solve for sort

38:38of the identity slash network slash authentication authorization piece in

38:44order to get files and collaboration.

38:46Like it implies that there are multiple, you know, Writers that have different,

38:51that are different, you know, and, and you have to somehow allow them to

38:54communicate with one another, allow them to identify one another, and then

38:57also, allow them to secure content, set up the rules for who can write

39:02when, how, all of that kind of thing.

39:05Is that?

39:06Yeah, I think that's right.

39:07I've personally become convinced of the, the value of and this is rather

39:12low level stuff, but signing everything with a key that maps one to one to you.

39:17And I think, you know Hold on, everything?

39:18What does

39:19everything mean?

39:19Because this is a specific technical problem that we're, like, quite

39:23deeply investigating right now.

39:25Because, like, you sign every keystroke on a 200, 000, keystroke

39:30essay, like the Embark essay.

39:31And what you have is mostly signatures.

39:33I like the way that you, you, you love mip mapping, don't you, Peter?

39:37Like going one level more, granular with definitions.

39:41Well, it's tough because like this is this, like you have to constantly

39:44be moving up and down these hierarchies of abstraction when you're

39:47working on these problems, right?

39:48So like, you know, Newsphere is cool because, you know, you can operate

39:53at this very coarse, chunky grain.

39:55But when you say everything, this is why I'm asking, like, you

39:58know, like, does this exclude that kind of lower real time thing?

40:02Maybe this is a solvable technical problem.

40:04Just curious what you think about that.

40:06Well, and I suppose you might be able to sign at the

40:08granularity you care about, right?

40:10I guess what I mean is, the web anchors security to domains.

40:15So the way the web thinks about security is, it's guaranteed by a domain.

40:20if you want to start making attestations about, people or individuals or actors

40:26on the network, then you start, maybe wanting to give those actors keys.

40:30In some capacity.

40:31And now I think there's, you know, sort of the cypherpunk view of it, which is

40:35that these keys should be self sovereign.

40:36And I, I sort of think that that's a possibility that a protocol

40:39should allow for, but a lot of people are going to want to escrow

40:43those keys or have them recovered.

40:44It doesn't, a detail that users shouldn't have to care, care about most of the time.

40:47Right.

40:48And an important one too.

40:49Yeah.

40:50But, what signing with keys does is.

40:54It allows for what a cryptographer called non repeatability.

40:57Like you can basically say this person definitely signed this at this time,

41:00or at least the person who had control of these keys, and if you keep those

41:04keys secure, you can actually sort of compose, trust at pretty high levels

41:08in pretty useful ways, mathematically.

41:11So in any case, I think where the file system and the network tend to run up

41:17against each other at odd angles is the, the internet or the web rather tends to

41:21think about security in terms of origins and, of domains and actually apps by and

41:26large adopted that security model as well, like, like iOS apps, Android apps, and the

41:30file system tends to think of, Security in terms of who has access to the device.

41:38And these, these two ways of thinking about security don't, don't mix

41:42very well without some sort of trust boundary or sandbox boundary

41:46that mediates that relationship.

41:49Can you, can you talk more about this?

41:50Cause this is, I've had many conversations with like browser heads

41:54and like, I get the idea of an origin and let me see if I can say this in

41:58a way that like a non browser head.

42:01I mean, maybe I'm a browser person now, too, by, by fate, but like an

42:05origin is basically like somebody owns the SSL cert for this DNS.

42:12And so we're just going to trust everything within that domain

42:15that's signed by that cert, period.

42:17But we're, but we're going to prevent mixing of that data.

42:21You know, unless it helps us sell ads with any data from any other origin period.

42:28I mean, I'm a little bit snide when I say that, but it is a very real concern

42:32and it is very empirically how ads work and, you know, however you feel about

42:37that, that's just, it is the case.

42:39That sounds like a pretty accurate description to me.

Decentralized Protocols

42:44Do we want to keep going?

42:46Well, the reason why I think it's relevant to the files conversation where

42:49we started and the apps conversation is that origin model is like so Like, many

42:56people I meet who work on these kinds of problems They get really frustrated by

43:00like cores and like same origin policies.

43:04And the browser keeps ratcheting these tighter and tighter and tighter

43:07over time, you know, yeah, exactly.

43:10And so to me, this is actually a really fundamental problem.

43:14Where it's like, well, but it's my computer.

43:16I want to decide what I do with my data.

43:18And then in the JavaScript console, it's like, browser says no.

43:21Yeah, actually it's the, um, right.

43:23But like the, the most secure web would be one where each website is hermetically

43:28sealed and you can neither interact with it nor receive output from it.

43:32Just little pocket universes.

43:35So, I mean, there's a sort of like reductio ad absurdum to this, where

43:38like we started with a pretty simple, it was like a reasonable solve right at

43:42a time, especially in the early days of the web when they were like, I think

43:46the mental model was that you might run a, a server out of a closet or perhaps

43:50your school would run a server, like it was this very utopian kind of flat

43:54Notion of what a network would be.

43:56And so the server was assumed to be the agent of the, you know, let's say,

44:01uh, Gordon's email would be like a computer that was like in my , my closet.

44:07And so like, yeah.

44:07Gordon com Right.

44:09Was a, yeah.

44:11And so if that, if that server is signing things, you know, you can sort of say,

44:16Oh, well, Gordon signing things, but that's not really the world we ended

44:19up with because of network effect, because of Facebook signs, everything.

44:24Yeah.

44:25Now.

44:26Keys are tricky, like, because there are these funny little things

44:29that you can't lose, just like in real life, or you'll be locked out.

44:33Except there are no locksmiths.

44:35There are no locksmiths, yeah, at least unless the NSA has its way, you know, but,

44:41No, but I do, I do have some hope that we're going to see some, some interesting

44:46kind of new calculus emerge around this problem with, the advent of like trusted

44:50computing modules, like secure enclave, on the iPhone, or I forget what the Android

44:54version is called, which makes me a bad former Googler, but you can sort of stash

44:58keys on these secure pieces of hardware, and then you have a pretty, you know,

45:03maybe your threat model is the NSA has your phone and is trying to crack into

45:06it, but that's less the world I'm trying to solve for, and more the world of like.

45:10It'd be kind of nice to know if,

45:12this I think Bonsy Buddy is a, is a, is a really credible threat model, right?

45:16Like, that's sort of the core threat model for, that browser

45:18vendors think about, right?

45:20Which is, you've got some, like, hostile agent on the user's computer.

45:23They've installed a search, ostensibly a search toolbar,

45:26or some kind of browser plugin.

45:28Which is now like extracting their credit card information or their bitcoins

45:32or just other private information.

45:34So so why does this matter?

45:35Well, so like this this origin security model inevitably results in silos

45:40like by definition data silos Like the the silo is the website the the

45:45whole security models that the data can't make it out of the website

45:48And so you just sort of end up with,

45:50Well, it does at the server, right?

45:52That's, that's the funny thing.

45:53The server.

45:54That's right.

45:54So they run an API so you can come in from other places there, but the

45:57user can't do it for themselves.

46:00Yeah.

46:00and then this puts the server in a very privileged and powerful

46:04position on the network because an API can, can cut you off.

46:08so data that might be notionally yours, is in practice, it's not really.

46:13I don't know.

46:14If you want to try to solve for that problem in some capacity, like multiplayer

46:18and some kind of data ownership or the ability to take your content with you.

46:23I think there's only three ways to solve this.

46:27one is with a decentralized protocol.

46:30Without going too much deeper into it, the protocol also has to

46:32have certain security qualities.

46:34I think the other is actually with local-first data, and this is, Peter's

46:38working on this problem of CRDTs.

46:39The reason local-first data gets you this is because you know, a server might cut

46:44you off, but it's like, screw you, buddy.

46:46I already have my content, right?

46:47It was already synced to my device.

46:49No, but that's not true.

46:51Like I wish, I appreciate the compliment implied there, but I think you still

46:55require the first thing, right?

46:56Cause like when you get the data, like this idea of local-first data,

47:00if it's happening in your browser, it's still being written into an

47:03indexed DB or an OPFS or whatever else that's owned by the origin, there is

47:08the ability to write real files, but it sucks and doesn't really work.

47:12And I see some folks here who have, like, built apps, like, no one is making a

47:16user click, like, allow access to this folder every time they open an app.

47:21Gosh, yeah.

47:21Uh, so that's what the, that's like the file access API.

47:25Yeah.

47:26Right, like, so it is, it, all of that data is still trapped inside the origin.

47:30Yeah, I'm sure that's true.

47:31I, so, Peter, you were saying, I have, wait, I have a third, third point.

47:36Yeah.

47:37So, and actually I do want to say CRDTs are still important.

47:39You're talking totally, totally, but it

47:41doesn't solve the, it

47:42doesn't solve

47:42the origin thing.

47:43And that's the part I don't have a good,

47:45let me circle back to this.

47:46I just want to throw out one time over coffee.

47:48You told, you told me.

47:49You don't need to care about which protocol you get the data from if

47:52the file knows how to merge itself.

47:54And I thought that that was a wonderful provocation to sort of like

47:57argue in favor of the CRDT thing.

47:59And, and all of these are satis satisficing, right?

48:01Like a protocol, a CRDT, these are not going to give you perfect exit in any

48:06case because, you know, in the worst case you might be shut out from the app.

48:09That lets you use this stuff in a useful way.

48:11But I do think they give you the ground conditions for things like

48:14adversarial interoperability to emerge.

48:17But the third thing, and the one that I think is underexplored, and I haven't

48:20even thought through it fully, is, something like a legal contract that

48:25guarantees the user access to their data along some axes, as long as a

48:29service continues to be maintained.

48:31You mean like a, like a

48:32UCAN or like an actual, like.

48:34Like TNC.

48:35Like something like, so, and this is the thing, nothing like this exists today,

48:40but there's no reason you couldn't write something like a EULA where

48:43the valence goes the other direction.

48:45Instead of the user signing away their rights, the user is granted

48:49certain rights over their data.

48:51Hey, I'll take that To my lawyer and yeah, no, actually I've been

48:55thinking, I've been asking for

48:57this.

48:57I sent some, I sent some irony there, but if you're interested in, I was

49:02thinking about actually going to a lawyer and see, like asking, is this crazy?

49:06Cause like open source would have been crazy at one time, but you

49:10know, it's interesting, right?

49:11Like we actually can bend contract on some interesting directions to solve

49:15using Social technology is things that are difficult to solve with, you

49:19know, bits and, atoms technologies.

49:23Right.

49:23Well, the challenge here is not even, could you write a

49:26contract that would do this?

49:27The question is like, how would this affect your ability to fundraise?

49:31Or what would happen when you told your VP of sales what you were doing?

49:37Yeah, maybe so.

49:38I think you run into a lot of these same problems when you say, Hey, we're going

49:41to give users control over their data through a protocol or through a CRDT

49:44anyway, so, you know, it's like in a lot of ways, like, this condition of things.

49:50Not interoperating and users not owning their data, like, an investor might look

49:55at that and say, that's a moat, right?

49:57It is a moat.

50:00Yeah.

50:00I tend to think about this in like longer terms that, you know, it's like in some

50:04ways, like Google doesn't have like a lock on search, like it keeps trying with

50:09things like the open graph, but I, like, I think it's perfectly possible to build

50:12a defensible business without data lock in, but it is, it's something you get for

50:17free with the current software paradigm.

50:19And so we're sort of stuck in this metastable state where extra work.

50:24This is where I think things like your protocol or Automerge or other kinds

50:27of similar projects have potential.

50:30If you can crack the, like what you need is enough benefit from

50:34participating in the ecosystem that allows you to get like a network style

50:39returns from interoperation over time.

50:42Yeah.

50:42You need a low enough activation costs that you can bootstrap

50:45the ecosystem around it.

50:46That has to be enough value from adopting the technology to see that

50:50kind of like a feedback on the returns.

50:53And I think the other thing that we haven't talked about that, but that

50:55is implicit in all of this is like the political economy of the software.

51:00That is available to us is determined by like the market economy of the

51:04labor market and the VC funding market and everything else.

51:07And like, it's all well and good to like hand wave that away, but it is empirically

51:11the case, you know, like information may want to be free, but my rent isn't.

51:16I think I look at this as in like a kind of scrappy, maybe

51:22like entrepreneurial way.

51:23I'm looking for the cheap hat because I don't want to wait, abolish, wait

51:26for the abolition of capitalism to improve software somewhat.

51:29And I do think there are these kinds of judo moves, these interventions, like

51:34CRDTs that can change the calculus.

51:39Yeah, maybe I'm wrong, right?

51:40and, and maybe you're wrong.

51:41Maybe we're all wrong, but I, I don't think so.

51:43I, I think that it requires great ingenuity and creativity to find

51:47the strategic angle of intervention.

51:49But yeah, you know, people do this all the time.

51:51They do it, for, they, They apply strategy to find an entry into a market in order

51:57to gain, because incumbents don't want new entrants in the market either.

52:00And I think we have to put that hat on actually, if we're going

52:02to, to change the, the balance of power, that users have.

52:07How do you think about that with what you're

52:08doing?

52:08Oh, sorry, Jess.

52:09No, it's okay.

52:10I was going to mention, I think there's, so I, I love your three things, Gordon.

52:14The, there's three ways you can solve the problem.

52:16Like some decentralized protocol that, Or two, local, distributed local file

52:20state, like distributed local state.

52:22And then the third, sort of a legal contract.

52:25And I think the one, Peter was pointing out that the market economics right now,

52:29what is easy is also very profitable and therefore the label market and the funding

52:34market were, it, it drives against the, any of these values and it supports sort

52:38of the data modes that we were talking about before with the same origin policy.

52:42And I think the biggest funding mechanism that I see, Peter's not going

52:46to love this, that is alternative to VC right now, is there is a large lump

52:51of money in the Web3 blockchain world.

52:55And you do see some of those projects putting that to use

52:58to try to solve this problem.

53:00I see.

53:01Famously anti capitalist distributed wealth system of

53:06the cryptocurrency market.

53:08Known for its transparency, honesty, and, Fair outcomes for all involved, yes.

53:14You're not wrong, and also there's a lot of money there.

53:17And so this is kind of like, it's basically the flip of

53:20what you were just saying.

53:21It's like we have to acknowledge the reality of the market economics.

53:24Well, the reality of the market economics is that there is billions of dollars

53:27in these decentralized protocols that are basically payment rails for

53:31Distributing tokens and things like that.

53:33And a lot of shady, crazy stuff is happening there for sure.

53:36but they are, tell me about the

53:38stuff that's not, cause I think that's interesting.

53:40So here's the one that I wanted.

53:41This is the reason I brought it up is I don't know if you guys have been

53:44paying attention to the Farcaster protocol, but they're distributed

53:48social media protocol and they have been iterating on this thing called

53:52frames and frames are like a micro app.

53:57Within a tweet, if you could embed.

53:59And the, the interesting thing about it, they're opting for Gordon's

54:04option one for distributing state and, access and, and all of that.

54:09Yeah, they actually,

54:10their, their protocol is not too different from BlueSky or even our own, Neosphere.

54:15but it does make a couple of choices around key rotation that,

54:19and they use a blockchain for this.

54:20I could get into the why, if we want to later on.

54:23Well, here's the more interesting thing.

54:24Just today, they showed off, frames that allow for, money tra uh, sorry,

54:30token transfers to happen inside of it.

54:32So basically they built into their protocol, micropayments.

54:36And the reason I think this is really interesting is because I think

54:40one way out of this, the current.

54:42uh, world that we're in with browsers and same origin policy and the sort

54:46of, the dead end, but it's a happy dead end where the web is, is a, is

54:50building a payment protocol into the, the primitives of the network itself,

54:56such that people can renumerate each other directly, and You know

55:00the web was meant to have this?

55:01There's an HTTP code for payment, like pay per view.

55:05It just never got, yeah, yeah, yeah, yeah, yeah.

55:08This is like Tim Berners Lee just stuck it in there.

55:10There's it's, it's along there with like, uh, what is the, there's

55:13another code for I am a teapot.

55:15Um, there's like these funny, like reserved codes for just in case that Tim

55:18Berners Lee stuck in there and payment was one, uh, we just never got it.

55:22Yeah.

55:23I don't know.

55:23I don't know what you're going to ask with regard to Web3.

55:26I think the idea of like digital payments is not crazy on its face,

55:30and it would be interesting to think about ways to do that with a protocol.

55:33I think my stance with regard to Web3 is sure to piss literally everyone off

55:38on both sides of this very kind of like divisive issue, which is that, uh, I

55:43think that, I look at a blockchain as a really interesting technological

55:45artifact that does solve some problems that are very difficult to solve in other

55:50ways, but the number of people who want to solve those problems is small, and

55:53nevertheless, they can be high value.

55:57So these are things like key rotation, which Farcaster uses.

56:01There's been this, there's always been a challenge with key rotation

56:04where you need like a third party to, to basically handle that rotation.

56:09And that third party is kind of like a point of failure if someone, co ops it,

56:13but in any case, you know, I, here's where I'm going to piss everyone else

56:18off, uh, on the crypto side is like.

56:21I think all of the problems of crypto are not crypto problems.

56:24They're, market problems.

56:25And if you look at them point for point, like, you know, tulip mania,

56:28wildcat banking, crazy token pump and dump, these are literally all things

56:33that we had to go through with regard to, capitalism in general, markets

56:37in general, for the last 200 years.

56:39And then we accrued this enormous edifice of regulation.

56:45Around markets in order to get them to function.

56:47So my personal attitude toward this stuff, which probably a lot of people

56:52are going to agree, disagree with for one ideological reason or another, it's like

56:55markets are kind of like nuclear reactors.

56:57They can generate a lot of interesting outcomes.

56:59but they also have a tendency to melt down, without cooling rods.

57:03And, and this is just kind of what I see happening with, with crypto.

57:06And I think it'll continue to happen up until the point that it gets

57:09some sort of enlightened regulation, which may never come because it

57:12seems like nobody's actually invested in that outcome on either side.

57:16so there's Gordon's unpopular take with regard to Web3.

57:21So Web3 is speed running 200 years of financial market regulation.

57:25Yeah, and, and actually,

57:27especially To me, like, proof of stake blockchains are a really interesting

57:31technology that solve, like, a interesting problem of, like, decentralized,

57:35totally ordered, sort of, source of truth that's credibly neutral.

57:39These terms all have to be qualified because, you know, they're all, all

57:42of that is also kept afloat by, like, basically, The financial function

57:48that a chain is supposed to serve.

57:52but I think the product market fit that chains have found right now is

57:54basically Vegas and, you know, crypto just recently skyrocketed back up again.

58:00So I've come to the conclusion that Vegas is actually anti

58:03fragile and you can't kill it.

58:05but that's also something that is, Vegas is not personally my favorite place to be.

58:09so.

58:10If I engage with it, it's, it's, you know, kind of, with, with care and

58:14with a little bit of self protection.

58:16Yeah, I mean, I, I think Jess, has certain expectations about my

58:21perspective on these things that are not entirely unfair, but I, I would

58:24say I'm not anti crypto conceptually.

58:27I'm just opposed to getting dragged into the kinds of communities that I've seen.

58:33You know, I just, like the ideas are reasonable.

58:37It's just, the outcomes seem.

58:39Odious.

58:40So I'm just like,

58:41so many things when,

58:43when there's like a healthy, high functioning, cryptocurrency for like

58:47micropayments that isn't like a, like an insane pump and dump, a penny

58:53stock, gambling, whatever, right?

58:55Like, yeah, like, but that would be cool.

58:58I, I always ask if anybody has one in mind and nobody does, but in the

59:02meantime, I also think it's a little bit irrelevant cause like, Oh man, use PayPal.

59:06Use, like there's lots of, use Square Cash.

59:09Like there's, use a freaking credit card.

59:11Like there's, payment is, is, yeah, there's more friction

59:14than, than you'd like.

59:16But like Stripe is pretty good.

59:18There's alternatives and Wraparate.

59:19Like these are hard problems, but they're so much less hard than they used to be.

59:24And they're, you know, Apple Pay is pretty great.

59:26Like, you know, like we don't need to invent a whole new

59:31financial economy to do payments.

59:34But I do think broadly, like we've got a little bit off

59:38track here, which is like files.

59:40Well, no, it's files over apps.

59:42And I think again, like we have to ask why, why are things the way they are?

59:47And so things are the way they are because you can make a living

59:49building software that way.

59:51You can get investment from other people to build software.

59:54And if you can't get investments and you can't make a living, then you're

59:56not going to have the ability to attract, attract top talent, right.

1:00:01To bring in the investment it takes to build a team in advance of revenue.

1:00:05Right.

1:00:05And like the problem with crypto money is that it always ends up being predicated on

1:00:11like the promise of crypto style returns in sort of the, you know, idealized case.

1:00:16And so what you see is like ecosystems aligning around trying

1:00:19to generate one of those crypto pops, because that's the expectation

1:00:23that comes with the investment.

1:00:24And I think, I don't want to call anybody out.

1:00:26I'll just say that there have been very interesting and promising

1:00:29projects that have over time had to realign themselves to generating

1:00:33crypto Interest style returns.

1:00:37And that, that, that has very frequently been to the detriment of their mission.

1:00:41Right.

1:00:42Yeah.

1:00:42Token pumping tends to be an unstoppable force.

1:00:45Yeah.

1:00:46And getting that doom loop.

1:00:48Yeah.

1:00:48And I wasn't necessarily bringing up web three to say that that was

1:00:50the solution, but I do think the acknowledging there, there's two things

1:00:55that I think are interesting here.

1:00:57One, how do you fund the.

1:00:59The research into making, making easy what we want, what we believe to be the future.

1:01:04and one of those funding sources that is non VC is a pool of cash that is Web3 that

1:01:10has all sorts of other challenges with it.

1:01:13Right, but look at what's happened there, right?

1:01:15Like, like really specifically look at what's happened there.

1:01:18Well, I do actually, I want to take Jess's, perspective on this for a minute.

1:01:22Cause I, I sort of gave my, you know, like very boring, ambivalent take on it.

1:01:27But like, I think actually there, if you, if I remove my moral valence,

1:01:33like there are a lot of things that seem crazy on their face, that generate

1:01:36unexpected outcomes and like, I see a lot of money getting dumped into

1:01:40things like ZK proofs, which is a general purpose cryptographic technology

1:01:44that has nothing to do with chains.

1:01:46Work has been mostly funded by, blockchain maxis.

1:01:50I see interesting work on elliptic curves.

1:01:53you know, like, uh, like Zuko Wilcox's work on elliptic curves,

1:01:57that's being done around this.

1:01:59they use IPFS.

1:02:00IPFS ecosystem has this new kind of, I guess you could think of it as like

1:02:04a zip format for blocks that, you can stream in and it's all sort of like

1:02:10signed in the way that you can sort of streaming, uh, stream this stuff in

1:02:14and, and cryptographically verify the blocks as they come in is it's using

1:02:17this elliptic curve technology that was, you know, it's ultimately developed by

1:02:21Zuko Wilcox for ZK stuff, uh, Zcash, I

1:02:25believe.

1:02:25And, uh, UCANs, from the Fission team are another great example.

1:02:28I think they've taken, you know, a lot of research funding

1:02:31from those kinds of sources.

1:02:32I think that's good stuff.

1:02:34Yeah, I think

1:02:34we're also, we're seeing like Farcaster and BlueSky and Nostr and various sort

1:02:39of stabs at decentralized, protocols that actually don't really have a

1:02:43crypto component, but I think are benefiting greatly from the mental

1:02:47models developed by crypto researchers, cryptographic researchers, as well

1:02:51as, Cryptocurrency researcher.

1:02:53So I think, you know, it's, it's like, there's this quote I really

1:02:55like from this, evolutionary biologist who says, containment can lead to

1:02:59contagion, and a degree of isolation is of, great, Importance in sort

1:03:04of new forms of being a false.

1:03:06So it's like the island ecology thing Like if something's off to the side

1:03:08and it's under a set of different selection pressures You could be like

1:03:11those selection pressures are perverse.

1:03:13They're terrible.

1:03:14Maybe they are but you're gonna get some weird and interesting critters

1:03:18that emerge out of that primordial soup for good or ill, right?

1:03:22Like I definitely think we're seeing some things emerge out of the crypto

1:03:27landscape that wouldn't emerge out of the incentives of like the web 2.

1:03:320 aggregator kind of landscape.

1:03:35Well,

1:03:37we're, we're about a little over an hour and let me see if I can try to

1:03:40land the plane here and kind of recap the conversation we've had so far.

1:03:43So we kind of jumped in with this files over apps and kind of talking about each,

1:03:48some of our different work around this.

1:03:51And then, there was a lot of history discussion around sort of how

1:03:55did we get to the current system where apps could even be used in

1:04:00a sentence like file over app.

1:04:02and what does that, what does that mean for where we are today?

1:04:05And then I think the, the most recent exploration has been

1:04:07how do we get somewhere else?

1:04:09How do we get to a place where files to open, to echo Peter's opening

1:04:13comment, where files didn't get their butts kicked and where we can actually

1:04:17rely on some of the file like goodness that we get, where we can get this,

1:04:22a combination of files and places.

1:04:24And, Yeah, there's been some really great work, that, Gordon and Peter and

1:04:30certainly, I've been trying to chip in as well on, moving us in this direction.

1:04:34Really appreciate Obsidian and Kapano for kind of, being a

1:04:37working piece of software that embodies these concepts as well.

1:04:41we've talked a lot about, local-first software as well.

1:04:45that's another term.

1:04:46If you're not familiar with it, you should definitely look for the original

1:04:50article from Ink and Switch that defined the term local-first software.

1:04:53And if you want to kind of hear more conversations like these,

1:04:56there's a podcast by Johannes, which I noticed is sitting in

1:04:59the audience, called localfirst.

1:05:02fm.

1:05:02And that'll kind of take you on the journey of, local-first and which

1:05:07really does, not 100 percent overlap with the file over app, philosophy,

1:05:11but there's a lot of overlap there.

1:05:13And there's a lot of technology that's being developed that can

1:05:15hopefully get us from where we are.

1:05:17And sort of this current ecological conditions to a new place.

1:05:22but Gordon or Peter, do you guys have any closing comments before we

1:05:26wrap up?

1:05:27I do, I do, which is, I liked Peter's put that we need to make

1:05:31the right thing, the easy thing.

1:05:33I think it's actually extraordinarily difficult right now to build an app

1:05:37that, puts the user in control of their data in any meaningful sense.

1:05:41Almost literally, no matter how you define user ownership of data.

1:05:44And, you know, the work Peter's doing in CRDTs, and I think the many decentralized

1:05:49protocol projects, including our own, are all attempts at trying to lower the floor

1:05:55so that more people can, can build apps that actually do put users in control.

1:05:59I think that's, that's necessary.

1:06:01cause otherwise it's going to continue to be prohibitively

1:06:03expensive to make that a reality.

1:06:07Yep, absolutely.

1:06:09Absolutely.

1:06:09Peter, anything to wrap us up?

1:06:11I don't think we have an answer yet.

1:06:14Like, I don't mean us here, though, also that.

1:06:17But, like, I think that, there's a lot more work to do and a lot more to say.

1:06:21And, you know, I have plenty of like pet theories and, and ideas as I'm sure

1:06:25everybody here on this call, does, but I think, I've tried not to, I've tried

1:06:32not to fall into solutioneering in this conversation because I think it's, it's

1:06:36easy to get, jump into the why don't we just, kind of mode, but I think that

1:06:40like a good, clear understanding of the opportunity, the challenge and the

1:06:45benefits and like a, uh, Like I just feel like very often people sort of

1:06:49lament the way things are, but I think Gordon's approach to kind of analyzing

1:06:55the, not just the technical landscape, but like the political and cultural

1:06:59landscape as well is part of why I so enjoy these conversations that we have.

1:07:05And I really, thank you, Jess, for, for allowing me to

1:07:08jump in and be a part of it.

1:07:10Likewise.

1:07:11It was great.

1:07:12But yeah, I'm, I'm, I'm very, uh, Excited to see folks like yourself, you know,

1:07:18like actively engaging kind of one rung closer to sort of market forces.

1:07:23And, you know, I don't think there's a wrong thing to try.

1:07:27Like if you're doing things, I wouldn't, I'm doing things you wouldn't.

1:07:30I think that's great.

1:07:32I'm not here to be a bigot for an approach or a technology.

1:07:35I'll speak my mind and tell you what I think.

1:07:36And if your thing works, I will celebrate, you know, that's, I hope that we all

1:07:41have that mindset of like, You know, it's not about arguing for ideological

1:07:46purity or like the one true solution.

1:07:50I think it's just we got to get out there and try and make things that work

1:07:53for people and then See what people say when they try them and then keep, keep

1:07:56going, keep iterating, keep trying again.

1:07:59So if you're sitting and listening to this, I don't

1:08:01know how many people are here.

1:08:02There's a bunch of people wandered in.

1:08:03yeah, like go, go make things, I guess.

1:08:06Or try some of the things that we've made and complain about

1:08:09the ways they don't work for you.

1:08:10That's, that's honestly, that's honestly a pretty good, pretty good contribution too.

1:08:15Such a gift.

1:08:16Such a gift.

1:08:17All right, guys.

1:08:17Well, thanks.

1:08:18And, yeah, look forward to the next time I get to chat with both of you.

1:08:21I see it, Jess.

1:08:22Take

1:08:23care, y'all.

1:08:23Oh, Hey.

1:08:23And if anybody out there wants to see, I think you're going to the

1:08:27local-first conf in person, right?

1:08:28Yes.

1:08:29I will.

1:08:29Yeah.

1:08:30I'm going to be there.

1:08:30Gordon, are you coming?

1:08:31I'm so sad to miss it.

1:08:33Yeah, I will not.

1:08:35Yeah.

1:08:36I think it's going to be glorious though.

1:08:37It's, it's going to undoubtedly be an interesting cross section of people.

1:08:41I feel like this is a, an idea that's just breaking into

1:08:44sort of broader consciousness.

1:08:46and the people who have been investing in it for a long time,

1:08:48I'm sure are all sure to be there.

1:08:49And it's, it sounds like it's going to be a real party.

1:08:51Yeah.

1:08:52local-first conf I mean, not everybody will be there, but hey, that's the idea.

1:08:55local-first conf, it's in Berlin, end of May.

1:08:58yeah, you should look it up.

1:09:00Berlin is amazing too.

1:09:02That should be like excuse enough.

1:09:04Oh my goodness.

1:09:05What a place.

1:09:06Alright.