Collective intelligence

Back

Often when you ask an expert who's accumulated a large amount of experiential data around a problem area, they're fabricating an answer. They actually have way more information than they could possibly convert into verbal symbolic language. And the inability to articulate something doesn't mean that there isn't knowledge there. Taste is real and experience is real. And you can have a lot of knowledge that can be extremely difficult to articulate.

Hello and welcome to Metamuse. Muse is a tool for deep work on iPad and Mac. But this podcast isn't about Muse the product. It's about the small team and the big ideas behind it. I'm Adam Wiggins here with my colleague Mark Granigan. Hey Adam. And joined today by our guest, Connor White Sullivan of Roam Research.

Thanks for having me on.

And Connor, I happen to know you have a dog companion. There's a Husky, right?

He is. One thing I like about them is that they're not bred to be obedient dogs because he didn't want somebody who was an inexperienced sled driver to drive the whole team out onto thin ice, so the dogs sort of take light suggestion, which is one of the reasons they're particularly hard for first time owners.

I feel like they take light suggestion also is a good training for being a manager of software engineers and designers.

Or maybe a parent too. But yeah.

Yes, a parent of a toddler. Absolutely. And I think our audience probably knows who you are and knows about Roam because you're definitely a notable figure in the tools for thought scene that we consider ourselves part of. But for those that aren't familiar, maybe you could give us a brief introduction.

How would you introduce Roam?

I consider it having created not only the kind of modern phenomenon of tools for thought, which obviously that concept extends well back in time, indeed, Mark and I did a whole podcast on it. But in terms of popularizing it in kind of the last few years, it's really, I think, opened the aperture for a lot of tools, including us and others to say there's more to productivity software than, I don't know, email and note taking and calendars. And that's what I think of as the collective kind of tools for thought scene.

And then, yeah, the specifics of the product. I think it really is all about the value of linking thoughts together and bringing things that I think of as being part of obviously the Internet, part of things that have been in our knowledge tools in different ways over the years, but putting them together into this kind of notes and personal memory and personal thinking space in just a new way that really struck a chord with people, indeed, to the point that I think it's been widely copied now.

And I would say you basically invented or at least pioneered a whole new category of software, which is quite a special thing to do in one's career, I would say.

The thing that is interesting to me is that part of my frustration in the last few years is that none of the folks who have supposedly copied us have copied the things that I think are actually important or are even indicative of the direction of why I built Roam or what we're aiming for. I think of writing as a tool for thinking. We've talked about this in past discussions, just one on one. I don't have a great extended working memory.

I've worked with people who are actually geniuses, who are able to visualize complex systems in their head, who are able to recall any piece of information they need, but I have a hard time just laying out all the steps of the problem and trying to think through all the variables that are there and just trying to keep my head straight, especially around things like software design, let alone systems design or building a team or any kind of complex decision.

So Roam, what you see right now as a product is something that did largely evolve as a sort of cognitive prosthetic for me, largely to handle my ADHD and trying to learn as an autodidact all of these things that I needed to do to be able to build Roam. I'm a self-taught engineer, self-taught designer, self-taught manager, maybe not good at any of these things, but had to learn how to fundraise, had to learn how to do marketing.

Like I studied none of these things, had no formal training in anything, and I had to figure out how to get good enough at a lot of things at the same time, more or less, or in various sequences. So I built Roam as a tool for helping me to organize my own learning and also just to I've had very severe ADHD for my whole life and it runs my family. But it is not. I think, Mark, I might have heard you say it on a podcast. Maybe it was some other colleague of yours that was on saying that they were characteristically unemployable or something.

Well, I was fortunate for startups to exist because I don't think I could have held down like anything, even remotely resembling a white collar job for any amount of time if I had not been able to to build my own companies where I couldn't get fired.

So a lot of Roam was built as a tool for me to be able to just organize my own thinking as I was thinking. So I think of it first and foremost as an extension of my working memory so that I can zoom in, eliminate all the extraneous things, have a clear workspace, but then at any point I can pick up pieces from, I can break problems down into smaller chunks and know that I will have the relevant information available the next time I'm able to pick it up, which might be some indefinite point in the future.

So Roam is a tool for writing, but it's also, and I'll talk a little bit about, it's a little hard to fully explain, especially what we've been doing over the last few years, if you don't know the context of why I started Roam and what it's trying to get to and why I even got interested in software in the first place, but I don't want to tangent too far yet. So yeah, it's a different medium for writing and thinking and trying to organize your brain so that you can think thoughts.

The way I said it before like this, there are things you can't see with the naked eye that you can see with the telescope, and there are things that you can't hear, but you know, if you've got a powerful microphone you can hear them. And I think that there are thoughts that we can't think unless we've got some sort of cognitive aids. And Brett Victor has talked a lot about this. I know you guys are probably fans of his work.

I'd love to chat a little bit about some of those ideas, but I think that a lot of our diagramming tools, mathematical notations, programming languages are all cognitive prosthetics that allow you to think thoughts you couldn't otherwise think. And Roam is, it's also a programming environment. You can write code and execute it in code.

We're trying to create a whole new kind of medium for expressing your thoughts, first to yourself, but then eventually be able to create a communication medium that can allow for a different kind of coordination and knowledge transfer and a new kind of collective action, collective thinking, collective intelligence. And that's the real thing that has been motivating me for at least the last 15 years, which kind of leads me into the questions that I want to ask you guys. Well, please do.

I have plenty to say about what you just said.

It's very inspiring, especially because in many ways you're not talking about the specific features or exactly the way that, how does this writing slash thinking slash note slash memory tool differ from what comes before, but this underlying why, which is exactly as you said with Brett Victor, I think Andy Matuszek talks about this a bit in his work talking about, for example, Roman numerals versus Arabic numerals and how that allowed us to essentially do new things, think new thoughts, do new kinds of math.

And the computing medium obviously has all this potential to open that up, but to date, even as far into this computer thing as we sort of are, in many ways, we're just transliterating, you know, okay, I've got a sketchbook. Okay, now that I've got an iPad, let me make a direct transliteration of what's on paper. I've got a typewriter. Let me turn that into a word processor and so forth.

And I would say most notes programs, even pretty sophisticated ones, I don't know, you know, take Evernote and Prime 10 years ago, can obviously do a lot of things that like a paper filing system can't do, but in the end, it kind of is just that on a computer.

And it seems very clear to me that there's so much more potential if we truly embrace the dynamic medium of the computer, and there's probably a thousand different experiments we need to do and different people will need different things to your point about what exactly is the right thinking prosthetic for you probably is also for a lot of other people, but maybe not everyone in the world. Different people need different ones. And that's why I think it's so important to experiment and break out of our established categories.

But I felt like a few years back, you couldn't get past the like, again, productivity software, just kind of like notes, email, word processor, spreadsheet, and happily the tools for thought seen that you really helped seed, I think has opened our minds to like, okay, let's do some innovation here.

Even the idea that you could have end user customization where people could actually write code. I mean, like, we got so much pushback when we let people run arbitrary JavaScript inside. And I mean, rightly so, because it's also a multiplayer tool. Obviously there's some security concerns, but my entire thesis is I want to give people power.

Right.

And I know I'm extremely neuro atypical. And I know that a lot of systems which worked very well for plenty of other people worked horribly for me, right. Schooling being the sort of most obvious one. So I know the feeling of being put into a box and the box not being extremely constraining and wanting to do more and needing people who do not want to give you their permission. I hate asking for permission. And so that's one of the reasons that first I was like, well.

Wild West, you want to run JavaScript, we will give you the ability to completely break everything in your graph. Like if you want to really mess yourself up and just like grab some code that you found off the internet and put it in there and like maybe you'll lose, you know, all your notes because you've got some random, I don't, especially in the early days, it was still a small amount of attention there. I have a very different attitude than folks with the security mindset of, well, what if hypothetically somebody might be trying to steal my notes?

I'm like, you've been using the product for a day and a half, like, I don't think this is a hard target yet, but I get ahead of myself. I have been really excited to see that proven out, you know, that people now are trying to do something that was pretty common in things like text editors and for Emacs and Vim and for professional programmers are very used to the idea of being able to modify their tools. And if you work in the trades, like my recreational activity is doing metalworking, you know, I like welding for fun.

Right.

And one thing I like to do is like making my own tools and making jigs. And like if you're doing any kind of carpentry, you know, oh, you don't have the exact right tool for it. Well, if you've got an angle grinder and you've got a welder and you've got some scrap metal, like you might be able to Jerry rig something up that might be able to serve the purpose of what you're trying to build a one off tool. And we haven't had those for knowledge workers except for in the domain of computer programming.

And I think that people who do other kinds of work, it's been very exciting to see so many like folks who are doctors who have never written a line of code in their life. And they're able to learn in the weekend enough to build some functionality into Rome that like is not my priority. I don't care about it. It would never occur to me to make it, but it's ideologically important to me that they not have to get my permission to make the tool do what they want to do.

So here's the question I've got for you guys, which is when did you guys start caring about computers at all? And what was it that made you care about them?

Yeah, I usually pegged myself for eight years old, and I think it was an Atari with one K of RAM, since I'm old enough that that was the kind of computing. I think we had one of these at our entire school, the elementary school I was in. I don't know what drew me to it. Maybe this is just a classic young nerd thing that you can't identify it. But I like to at least post hoc rationalize it that I saw the potential for creativity and I immediately want and all you could really do with program computers back then was program them, right?

Like I think it used Logo, maybe later Basic, and you'd get in there and just the same way that a kid just wants to pick up that piece of paper and the crayons and start drawing scribbles and that it's this form of expression. I saw the same thing in the computer and just was endlessly fascinated with it. What about you, Marc?

Yeah, similar story for me. I did not have the experience of programming very young. I didn't do any really substantial programming until I was in college and the specific impetus for me was I was studying economics, among other things, and I wanted to do agent-based simulations to test out some economic ideas. And so, OK, I got to teach myself Java. And I remember in retrospect how completely terrible that Java program was, just the incredible amount of copying and pasting. You wouldn't even believe it.

But anyways, at that point, I got into that track that Adam was describing where it's an incredibly powerful and accessible medium for creating. I always like creating things like I did model airplanes and other stuff like that. But there's actually a pretty narrow set of things you can actually do that's both powerful and accessible. Maybe you are into welding, but as a 19-year-old in rural Maine, it's kind of tough, right?

Yeah.

But you can get a computer and do whatever you want and you don't need to ask anyone's permission and the sky's the limit. So it's pretty cool.

I want to touch on both of those. Adam, you're talking about being a nerd. If you can imagine it, I was such a nerd, I didn't have enough friends to play D&D with. Let's say that. Like, I used to play this single-player D&D type book. It was like a choose your... I don't know if you guys might have been actually maybe too old to remember the R.L. Stine choose your own adventure books. Those were like really big in the 90s.

Oh, yeah.

There was a game called Quest and it was individual paragraphs, each with a number. And it was like, oh, if you go down the right hallway, go to 232. If you go down the left one, if you fight the goblin or whatever. But I remember playing these games all the way through and then I actually made them multiplayer. I did have two friends who were nerdy enough to indulge me in this for like a couple recesses before they were sick of it.

But I played the games all the way through and then continued the rule set for the game and just started writing paragraphs at the end of the book to try to like keep the game going. Because they were originally supposed to publish like 10 of these game books, but only two of them got published in the US. So and in some ways, actually, there's something reminiscent of Rome in that sort of backstory. You're talking about agent based simulation for economics, Mark. Here's the next question I've got for you guys.

What's the first problems or bigger problems that you remember unaware of or even caring about?

I'm going to give you sort of a half answer here. So there's certainly problems if I go back in my memory when I was a very impressionable kid, you know, whatever. The third grade science teacher says, you know, all the turtles are dying. So everyone goes home and clips all the six pack plastic things and stuff like that. But something that's still sticking with me is when I was working in computers and originally I had this very unalloyed excitement about the cloud coming out of college. I was like the cloud services in particular.

This is before I was even at Roku.

It's just so powerful to be able to have a hosted service that does everything for you.

And the end game is everything moves to that model. And obviously, I still think there's a lot to that. But it was only with the experience of living in a society that has embraced that model. You realize some of the really tricky downsides of it.

Something I'm still grappling with is someone who works in computers.

So we could do a whole episode about that. But that's one that I've definitely thought about.

Our follow up question. You guys know the phrase, you can't solve the problems you have. It's attributed to Einstein. You can't solve your current problems with the thinking that got you into them. Do you know the exact quote?

We can't solve the problems by using the same kind of thinking we used when we created them.

Yeah.

Yes.

And I do think there's maybe a positive spin on that. One is like we had dumb thinking and then that led us to be in a bad situation and we need to be less dumb. But another way to put it might be that in moving yourself or your group or society forward with better thinking, well, that creates new problems like the cloud version there that Mark mentioned. And now you need to solve those new problems. But on net, you're probably better off than where you were before.

It's just that the idea that anything is going to bring a panacea, a utopia where all your problems are solved. And now we don't need to have new thinking and new solutions and be aware of the downsides of the world we've created. That will basically never happen.

I think you could even generalize it and say, even if there's not progress, there's change. The world is different.

Yeah.

There's no going back. You know, that's the way it is. The only way out is through.

So I thought that's the potential hope is networks. I politically became really alive when I read Yochai Benkler's The Wealth of Networks. And, you know, saw Clay Shirky's organizations institutions. My life plan was to sell John Deere tractors in Africa because it seems like the coolest job I could do. I was planning on like doing a few years in college and then like going and being like a heavy equipment salesman in Africa because I wanted to travel. And that looked like a job that would pay for me to travel to really crazy places.

And I thought, you know, excavators and tractors were cool. So I was like, that probably do that. That was like my freshman year plan. And then I was like, oh, but actually we might be at a period of history that is as important as the printing press.

Part of the thesis, I guess, of The Wealth of Networks is that the creation of this network society through the ever increasing communication capabilities, the Internet being the kind of at least to date, the ultimate manifestation of that creates a moment of opportunity to have an impact, to change the way the world works. Again, that's certainly where the startup world sees itself as in this highly dynamic, you know, early stage thing where you have the opportunity to maybe have more impact than you would as an individual.

So was it that part of the book that sort of inspired you to think, OK, that...

Well, at least a couple of things. It was the idea of nonrival goods. So first, the idea of I'd make something and it costs zero dollars for there to now be a million versions of it.

Right.

And that because the goods are nonrival and they're post scarcity, like they have a different kind of economic pattern to them. That was one aspect of it. And so he sort of had a four part quadrant that he was sort of laying things out. He was thinking about the state, the firm, the market and the network. So a state would be something which is public goods like, you know, they're trying to manage resources that cannot be sort of carved up into small pieces.

You couldn't have property rights on things like clean air, you know, so places where there's lots of externalities and like one person could hurt the commons. But there isn't less private incentive for people to maintain or protect the commons. So the state historically has used coercion for the governance of the commons. So the state would be centralized management of a commons. The firm would be centralized management of private resources. The market would be decentralized management of private resources.

And the network is decentralized management of public resources. So like it allows for the creation of new kinds of commons, particularly information commons.

And so here we're thinking, what, open source or the way that like DNS works where there's no...

I mean, I also was interested in Ray Kurzweil at the time. So I was thinking general like techno utopian post scarcity, like what happens when we can 3D print organs? And the more we can get to actually we might be on the cusp of technology that allows you to take things from the digital world into the physical world. And this could...

would be potentially somewhat revolutionary in terms of if I can get any medicine that I need by like downloading it and if someone can make an open source version of the medicine that I need, like that was the kind of one aspect of what I was thinking, that was that book. But the other thing, I think it was mostly just I got some hope that like, hey, there's, you know, Lennox. I also got then disillusioned, but I did a bunch of open source stuff.

My undergraduate thesis was on trying to create a way of Wikifying a local government and actually like making a more direct democracy type approach under the assumption that, you know, people have a ton of tacit knowledge. Like there's voices that are not heard that have expertise that is not like recognized. And you need culturally relevant solutions.

I was coming from anthropology background, so I was thinking a lot about like the thing that is going to work in a rural village in Ghana is like not going to work necessarily in Boston, Massachusetts, and even the thing that's going to work in Southie is not going to work in Jamaica Plain maybe. Like you need to tap into the resources and the culture and like the actual life and local context, lived experience of people who are in a community.

Yeah, well, I'm a huge fan of being close to the problem lets you, like you said, have a tacit knowledge, understand it in a way that you just can't, but yet as our societies get bigger and literally this is just a scaling the number of humans thing that exists, which is governments are going to naturally get further away from the people, right? The government of the United States, 200 years ago, when the population of the United States was a tiny fraction, you know, it's much closer to those people whose problems it's hopefully trying to solve.

Do you guys remember your first ideology?

Baby's first ideology?

Baby's first ideology?

Yeah, I mean, the classic thing you have with, let's say, university students is, yeah, they get really into environmentalism or something like that. And it becomes almost the purism of it, right?

Do you remember the first thing you were ideological about? It's easier to call it an ideology if you're like post, if you've left it in some ways.

Yeah, probably open source, actually.

Open source? How old were you?

And Linux specifically. This is the year of Linux on the desktop.

This year is Linux on the desktop?

Well, being the kind of lover of open source belief in what that could bring and thinking, OK, commercial software's days are numbered. Eventually we're all going to be running, you know, things that are developed in the common for the common good. It's just a matter of time. So, yeah, I think that was probably one of my first in the late 90s.

I was definitely in that ideological camp until I tried to run an open source project. And then I realized it's a lot easier if people actually can make a living and do the thing full time.

Yeah.

What about you, Mark? Do you remember your baby's first ideology?

Oh, I was going to give you another half answer, which is I've always been more of an is than an ought person. I associate isms with ought. You know, the world ought to look like this. And there's something to that. And of course, people having ought notions feeds back into what is. But for me, I keep myself busy just trying to understand what's actually going on and the dynamics that are unfolding. As you understand things better, you certainly develop notions about how they might be different, how you might want them to be different.

But I try to keep a real close eye on how the world actually is, because just understanding that is quite hard. To give you a concrete example, you talked a little bit about technological determinism, just understanding how the various technologies that have and are evolving, what that means for us is incredibly not obvious, even something like computers or even networking. Is networking going to be decentralizing? I don't know. It's right in the name.

Shouldn't it be or is it going to be highly centralizing as for example, Samuel Burge has argued in one of his pieces that we can link to? It's not obvious to me.

All right, well, I mean, we can take a second on this because I think the best kind of prophecy is forthtelling, right, where certain kinds of things like the most interesting predictions are self-fulfilling predictions, right? Something where because you imagine a thing to be possible and you believe that it's worth your energy to try to make that thing possible, you can make the thing possible, right? I mean, both of you guys have made startups happen out of nothing. Like nobody makes anything happen out of nothing.

But like the early stages are crazy because it's not a Ponzi scheme per se, but like you need to convince investors that you're going to be able to convince engineers that you're going to be able to convince customers. Like it is a crazy balancing act where you have to make a vision into reality, even just in the assembling of the early team and the raising of enough capital for people to quit their jobs for long enough to get the proof points to convince more.

Yeah, I think there is a faith-based element to it. Yeah.

I mean, my thinking is there are two ways. You know, the idea of a map territory conflict?

Yeah, absolutely.

If the map I have in my head doesn't match the territory, there's two ways I can change things. I can try to update my map or I can get a bulldozer and I can try to change the territory, right?

Yeah, there is something to that. The power of the ideological person and often the world are changed by young people that have sort of like an unrealistic vision because they aren't stuck in the status quo and they are willing to take that bulldozer in. But it has to be balanced somehow by pragmatism.

My first company was trying to make an online town common. The idea was if you could get the for any local issue and every issue could be made into a local issue was the sort of the hope, right? If you could accumulate political capital online.

So if you could confirm that all the people on this little forum thing were actual registered voters in this town was the mechanism that we had for trying to accumulate political capital, then you could sort of force a more responsive local government and start to sort of decentralize the place where people are most likely to influence and get some power. And the idea at the time, I was hoping, you know, oh, if you gave people that like then we could actually have democracy experiment everywhere, you know, like. But it totally didn't work.

It totally like I didn't even want to use it because I realized I didn't care that much about, like, you know, the property taxes and like the paving of the roads and like what to name the new library, like just the local politics issues were so boomer and I was like this little 19 year old, like libertarian socialist, like we're going to have an Internet anarchism revolution and all of my users for that product were like 60 plus. And I was so glad that I had no power to coerce anyone to do anything because I just didn't understand the world.

So I became even more pro startups because there's something beautiful where you have to be right about making something people want. You have to both have a vision, but that vision does get tested against reality of like, will it blend? Like, can you ship it like when you ship it? Will anyone care? Like if you build it, will they come? Right.

You kind of have to believe it in order to build it, but reality will test you there, which is one of the reasons I like startups and it's also one of the reasons why I'm hopeful for more diversity of political entrepreneurship or things like that. Like it's one thing I really do share in common with biology and the hope that there will be more micronation someday in the future and actual entrepreneurship and meaning bounded communities or something, something like that. Utah is a great example.

Somebody put out a vision and a bunch of people with the same kind of ideas. Utah is the original network state.

Certainly makes me think of charter cities, which is certainly another kind of libertarianism type sphere idea, but yet is that idea of it's not just about self-governance and getting to choose, but also to let a thousand flowers bloom. We have to try a bunch of stuff because, as you said, ideas have to be tested in the real world and we can sit around and debate them. And indeed, people do.

But until you can try it at scale over time, see how it actually impacts people's lives, do people really want to live in that place that does change society in some fundamental way?

Yeah. So here's another question I've got for you guys, which is like, and I'll give my answer first, but I've been thinking recently about just beliefs that sort of lodge in your head that end up propagating into all sorts of other things. And you don't necessarily go back that often to re-examine them. So I'll give one, which is the idea that creative work can't be coerced.

And I think this is part of why I've been so pro-voluntaristic type associations and trying to figure out networks for mutual aid and ways for people to help each other, where it is a very opt-in system. But I think it might also be just directly related to me having a pretty oppositional, like, low-agreeableness personality where I really don't like to be coerced in anything. So like, I just, I assume that good work can't happen under real coercion because I won't work under coercion.

Therefore, you know, I don't know if anything jumps to mind for you guys in terms of like little beliefs like that that might color.

The core of critical thinking, I think, is trying to examine beliefs that are in your mind and how do they get embedded in the reality is it's rarely a thing. I encountered a new idea, fact-checked it carefully, and then decided to make it part of my worldview. It's more you get exposed to something a lot over and over again, and it just, through osmosis, sinks into how you see the world.

And I always find it funny to stumble across little beliefs, even just things like, you know, should you keep this particular food item in the fridge versus is it okay to, you know, sit on its shelf stable? And sometimes there's just something I picked up when I was a kid from one of my parents or something, and I didn't realize until I was an adult that actually you can stick that on the shelf.

Wait, wait, wait.

I just never examined it, right?

What did you keep in the fridge that you didn't need to keep in the fridge?

Yeah, I'm trying to remember what, maybe it was potatoes? Was one that was like that. It does slow down their like budding or something like that. But I don't know, I think my mom always stirred potatoes in the fridge. And then I had a roommate that was just like, I'm just going to put them on the shelf, we don't have room in the fridge. And I'm like, wait, you can't do that? You know, they'll spoil. But then I'm stopping and thinking, well, wait, how do I know that? Or why do I think that? And the answer is, you know, it's just something I absorbed.

What about you, Marc? Do you have any things you've noticed that were like, it's the most general question. Do you have any unexamined beliefs? What a terrible question, Mike. I apologize.

I'm not sure if this is exactly what you were asking, but there are some lenses I keep in my pocket. I'm always putting them up and using them to look at the world. So, one lens is the lens of trade-offs from economics. That's very easy to speak in absolutes or to speak in terms of improvement or degradation. The reality is almost always one of trade-offs. Another one that I use all the time relatedly is distributed information processing. This kind of is related to your idea of mutual association.

The world is so complicated that there's no way for it to be understood centrally, especially when you consider that a lot of the things that are important to understand are matters of personal preference. So, it's not physically possible to bring that information into one place, compute, and spit out results about what ought to happen. So, it has to be done in a distributed way. And it's so easy to fall into the trap of, you know, what if we just brought all the information in one place and figured out what to do? It just can't be done.

And when you remind yourself of that all the time, you come across many cases where you see people trying to do that, to try to extract the information and put it through an explicit machine and turn out an answer. And you have to instead just let it be out there and let the network process the information and decide what should happen.

Can you go into more detail?

Well, this is a whole key tenet of Austrian economics or Hayekian economics. People can look up those things and read about it. There's a famous, I think it's an essay written about the manufacture of a lead pencil. And something as simple as that, there's actually no one in the world who knows how to manufacture a lead pencil. It has to involve many different people from around the world and they all have their own test of knowledge and understanding of what kind of wood is right.

And they know about the quirks of the machine and how it's always off by one degree. So you've got to counteract that. That sort of thing that it seems so simple, but even something as basic as that can be known centrally and needs to be distributed. By the way, not even to mention how many should be produced, at what price, where, what materials, just an incredible amount of complexity that can only be computed on and distributed away. I just find that a handy idea to go back to often.

Can you think of examples besides the market? The first thing that you were making me think of was, I feel like I've only in the last few years gotten language for thinking about why it makes sense to listen to emotions so much. Thinking of emotions almost as like Bayes net massive information compression systems, where you're just getting a vibe about like, oh, this feels off. There's an essay called the limits of legibility, or it's like a less wrong post that I like.

But often when you ask somebody, especially an expert or somebody who's like accumulated a large amount of experiential data around a problem area or around a skill or something like this, like, well, why do you think we should do it this way or that way? They're fabricating an answer. Like they actually have way more information than they could possibly convert into a bit stream that is compressed into verbal symbolic language.

And so if you treat the answer that somebody gives you of why as if it's actually meaningful, many people actually treat the why, especially if they want to argue about it, as though that's the real thing, rather than a tiny symbolic representation of what in that moment they were able to generate, which might not even be the real thing. And the inability to articulate something doesn't mean that there isn't knowledge there. That is like such a like taste is real and experience is real.

And you can have a lot of knowledge that can be extremely difficult to articulate. I found this to be extremely challenging when I was trying to introduce sort of counterintuitive cultural norms into the company. And I was bringing people who were used to working in. Since I finally found a moment where I actually made a point, Rome is not a normal company because I think normal companies are what got us into the situation we're in. I wouldn't want to work at Google. I wouldn't want to work at Microsoft. I would have no interest in being there.

They're not building products for people like me. And also they kind of are. But the thing that I'm interested in is trying to figure out a different way of thinking together and in a bunch of different ways. But especially as I was having folks coming in who I'm trying to communicate certain practices, especially practices around how I work with Rome that had just evolved over time.

I've got a whole very different way of using the tool than your average user and trying to communicate why I do things a certain way or why I was even asking somebody to do something a certain way. It was very hard to do if there wasn't trust that there was some intuition there and that the words that were going to be used as the explanation for why we're trying a thing were not the actual only reasons. Like I could come up with a hundred reasons for why we might try to do the thing.

Yeah. It's such an important point. And we've mentioned it on the podcast many times, but I think it's worth reiterating that experience and judgment and expertise, they're incredibly multidimensional. Millions and millions and billions of dimensions. And there's no way to compact it down either in terms of the model itself of experience, say, or the answer to some discrete symbols, as you were saying. And furthermore, when you get discrete symbols out, they're often just back-solved.

Like this huge multidimensional model spits out an intuitive answer, but then it's unsatisfying to convey that so the brain just like finds a way back through symbols it knows to convey something that sort of ends up at the destination. And it sort of plausibly sounds like a quote unquote argument or quote unquote reason, but it's just totally back-solved. That's not how they actually got to the answer.

Now I'm going to turn this around because this is an idea that I've embraced in my own thinking, but what does that mean for a tool like Roam or other tools for thought, which are inevitably collections of discretized pages and links and things like that? How do you reconcile those two worlds?

TLDR, my first startup, I started as an open source project, could not recruit anybody to actually work on it, was somehow able to pitch it as a business plan competition, like at business plan competition. It's got 10k, suddenly could get actually better engineers and the ability for them to work full time. And so I found having an open source political project, plenty of people who were interested in the idea, but nobody could actually help me build the thing.

A lot of the talent, as soon as I framed it as, oh, well, I guess we'll make it a media startup and maybe we'll sell the data. It's disgusting to think about the idea of selling political data now. But at the time I was just trying to figure out how do I win this business plan competition so somebody will give me some money. I worked construction before I got into tech. That was my summer job, but it was a terrible business.

I got to give some TEDx talks and go to the Aspen Institute and we did end up getting acquired by AOL, but it was not a good business model. We were selling software to newspapers to get high quality use generated content on super niche issues that they were like civically important for the mission of newspapers. It was just a bad, bad business. And I was trying to solve so many problems at once in terms of how do you build a user interface for collective intelligence?

How do you think about the political dynamics of like, okay, what people are excluded if you're using real names and you're using local like voter registrations? The problem of political coordination plus how do you crowdsource from a large body of people the actual best ideas from a broad perspective so that you don't have to read every comment? I was trying to solve a bunch of things at once and I found that I actually had to do some sort of science and the fact that I couldn't isolate any variable was like just blah.

So after that company was bought, I was still interested in how do you build a better way for groups of people to in a weird way centralize their decentralized knowledge? So maybe the Hayek point is it's not even perfect execution. This is just a bad idea. I've wasted my whole career. Maybe like if I just read Hayek, I'll be like, okay, the collective intelligence stuff isn't going to happen, but I wanted to simplify the problem. And so my first thought was if I am able to be as a single player able to take the best writing that was done.

So like one problem we had in the first company was how do you get critical mass for a social network and then how do you create a ecosystem that actually inspires people to be as articulate as they possibly can be about what their position is or like why you should do the thing? How do you actually get people to give really, really high quality content? And then how do you from a large mass of users identify the best content from a diversity of perspectives?

Because instead of just having people vote down ideas that are good articulations, but they happen to disagree with, how could you actually get the best ideas from many different perspectives and see this sort of multi-dimensional object of like any kind of question. But we were particularly starting with these local political questions. And my thought was the simplified version of this problem is well, one, we had only been able to launch in places where we had critical mass for my first company, which was called localocracy.

And so I was like, the tool has to work without critical mass. It has to work as a single player tool. And if I can start with the best writing throughout all of history, and I can be the one who's aggregating it, and I can figure out how to like map these different perspectives together. Well, now I've just.

isolated a bunch of variables because I don't have to worry about getting the best articulation of an idea. I've got the entire corpus of human history. I can just pick out what I think are the best articulations of the idea. And I don't have to worry about critical mass because as long as I'm interested in the problem, I can do this or as long as anyone's interested in the problem. And so there was a book called The Syntopicon. They'd spent like 50 million bucks.

It was from Encyclopedia Britannica and it was a great book's course of all the best ideas of Western history and the first two sections of the book are an index of these ideas where they sort of summarize it and they point to the paragraph number of where the idea is articulated by Descartes or Hegel or Marx or Kant or Plato or whatever. So I was like, well, if I can make a digital version of this and I can make it for a single player and then I just charge money for that.

I don't have to worry about selling software to newspapers who then run it to advertisers.

And I have to convince a bunch of other people if I can just find people who want to organize thoughts and I just sell it to single players, then I can maybe get the iteration cycles that I'm going to need, I can basically keep this company alive long enough to run through all the iterations to solve this potentially impossible UX problem of how do you actually create these high dimensional objects that represent many different perspectives around a single sort of truth thing? Like how do you build a truth engine?

How do you build a system that actually allows you to sort of create this Bayes net so that your beliefs could propagate?

Well, I see the breadcrumbs now. You start with kind of collective action and you're thinking in terms of governance, but you're also thinking in terms of networks and how to bring together sort of computing and some of the open source and maybe kind of more freedom oriented ways of organizing ourselves. You tried to do that with software for kind of participation in government.

And that was a total bust. But it leads you into the like it was so I mean, it's like imagine selling software to government. Now, imagine that you have to sell software to government and to the newspapers that are going through like massive decline at the same time. And the subject matter that is going to be discussed on there is like extremely boring. You guys familiar with Michael Nielsen's Reinventing Discovery?

That was the book I was looking for for forever after my experience with localocracy and then trying to work on because when I ran a labs group briefly and poorly at Huffington Post, because after we were bought by AOL, we ended up in the editorial division for HuffPost. And then I just was able to spin out my own little labs group for about a year, focusing on kind of collective intelligence, crowdsourcing knowledge, figuring out ways of doing new stuff.

And anyways, the book that I found that was sort of one of the better textbooks on thinking about the problem of collective intelligence is that one, and he talks about things like the problem of a conference where even if you have all the experts in the same place, you're not necessarily routing the right people for the right conversations.

You have to worry about when you're making everything synchronous, whether people had enough coffee or whether they're distracted by something like you want to be able to allow for a certain kind of serendipity to be more predictably happening and remove the sort of constraint of they have to be in the physically right place at the right time. They have to just happen to bump into each other. When you run into somebody, you don't know that they know something that you need to know in order to solve the problem that you're working on.

But you don't know what the name of their knowledge is. And there's certain kinds of human routing or information routing problems that he lays out pretty well in there. He calls it efficient allocation of expert attention. And so one of the reasons Roam is block-based even is just trying to work with that. So not just thinking about thinking in terms of blending programming and writing. So you're not just writing paragraphs. You're actually trying to think about a kind of data structure of a pattern of thought.

And that's a lot of what I've been trying to create as a medium is, you know, if you think about block references, which is something that none of these so-called Roam clones do at all, I don't know any of them that are actually multiplayer, right? The reason I referenced your offline talk is like, we've been multiplayer from day one, even though we've been a single player tool.

Right.

That was actually architecturally some of the hard stuff to figure out was like, how do I make this thing work as sort of a collaborative real time thing with a graph database and start thinking about the interpersonal dynamics of referencing somebody else's thought or like what are the different ways that you write when you're trying to write almost as statements that you're expecting to be reused by other people? How do you think about version control of a statement or like the way someone might transform a statement or rephrase a statement?

Like these are the kinds of, you know, it's thinking about language in a different way than paragraphs or pages, because we're trying to think about how to create an object where you're not going to have to read the entire history of a Slack channel when you go into it to get up to speed on what the group knows. Actually, if you've got a new piece of knowledge that really would have unlocked something that the group is talking about six months ago, you know, and the group kind of shelved that whole discussion because they didn't have that knowledge.

How does your knowledge immediately fit in and unlock that?

Right.

So it's thinking about a different kind of collaborative thought data structure. And so things like block references and the ability to build a statement up out of other statements, having unique ideas for those. Yeah, that's the kind of work that I think is important about Roam and something that never gets talked about on any of the YouTube videos of users.

I'm not complaining about it because if people weren't happy with the things that I thought of as basic, which is the stuff that everyone imitated, right, I needed to get those basic things to even get to the place where I could think about the block references and all these other things, which are still rough, therefore a problem people don't know. They have like nobody's trying to like reinvent pros. I'm trying to reinvent pros. I don't think pros works for large scale, collaborative problem solving, like essays do not work. I mean, they can work.

They're the best that we have right now. I saw you shaking your head, Mark. Right. Like and in fact, there's a whole other thing, which is like being able to go from convincing rhetoric that is storytelling, where the author is taking you on a journey with them into the structure, you kind of may want to have both, you might want to have a sequentially ordered narrative that is being presented.

But then if you're trying to analyze the logic of it or like debug the program, you might want to have a more sort of structured graphical representation of it for the analysis of like where the weak points in the narrative.

I find it very interesting, your vision for where you want to be longer term, which is really about collective intelligence more than individual intelligence, but you can't bootstrap into getting everyone to use something at the same time.

Also, your past and future selves are totally different people. Right. The past is a foreign country. So the other reason that I started with a single player tool has been I didn't have to convince anybody else to use it to be able to iterate on it. Other authors were other people and myself across time was other people. And so it is an easier and still extremely difficult problem to solve the problem of organizing your own thoughts over time as your thoughts change, being able to go back and actually reexamine them.

These are actually more related than people think. Like people don't realize how many selves they have. I actually think the idea that you are just one self is kind of so many different sub agents running around. Like one day you think this, if you're hungry, you think that. If you're tired, you think that. Like, how do you actually bring all your internal family systems? I don't know if you guys have ever gotten into that kind of stuff. But like, you're many.

You are many.

Yeah.

Contain multitudes.

Absolutely.

Certainly writing is a technology for not only communicating with others, but also communicating with your past and future self is a powerful piece of it.

And present self. I don't know what I think until I write it sometime.

Absolutely. Yeah. The externalizing the thought, the conversation with the page. You see what's there and that becomes a loop that's different from the kind of thinking inside the brain.

Or to tie it back to Mark saying, you know, when you were talking about you got into programming so you could build those multi-agent models for doing economic simulations, like that's the kind of stuff I want people to be able to do. It in Roam. It's like Roam is the database of all their notes, all their thinking. And so if they want to just start playing with stuff, they shouldn't have to worry about setting up a web server or web page or whatever. It's like, OK, they write some JavaScript and suddenly they're embedding a little.

And that's one of the cool things with Clojure is like Clojure interpreter inside Rome and a JavaScript interpreter inside Rome. So hopefully someday the future Mark is thinking through his economics thoughts with little simulations inside the notes and like they're part of his scaffolding of his own thinking. And he's going to be able to go back and not just read his old thoughts, but like play with the simulations that he was writing.

That's super interesting. And definitely the programmability built into the tool that again programmers, editors have had since forever, but bringing that to something that's more for other kinds of knowledge workers or other kinds of obviously power users, but people who are more working in the realm of ideas, not necessarily code. Putting those two things together, I think was a surprising but important innovation.

I'll say one word here too, which is that someone asked, what are we working on? What have we been working on? One thing that is still an open research problem that I've seen no one else even thinking about is the idea of that there are higher order functions for regular thinking.

Right.

If you do weird San Francisco hippie, like intentional relation stuff with other groups of people, like you get used to these kind of patterns of questions.

Right.

For instance, I ran a learning cult because I was trying to do it with workflow in Excel. Like I was trying to build a sort of peer to peer research group with just friends and folks that I met when I was in the Bay Area. I was trying to figure out the minimum thing I would need to build for Rome to build a decentralized research group.

Right.

And so I was in order to stress test, I was like, how close could I get to the ideal social and like information structure without building anything with like off the shelf tools? And I used workflow, which was in tree based outliner, and I used Google Sheets and I would do stuff like I would ask people, what are the like seven best books you've read in your life? OK, for each book, like what were the three big ideas for each of those big ideas? How did that impact you for each of those ideas? Can you find two quotes from the book?

Can you go back and do these things? And even just a simple thing like a for each function, right? Even just like being able to separate out, I want to ask questions and then I want to take those answers and I want to map new questions onto the answers, onto these things, and I want to create a data structure out of this map, filter and reduce. Like we don't have higher order functions for these basic qualitative kinds of interpersonal interactions.

And so one of the main things that I've been working on with Rome is trying to build a programming system for maybe for like teachers or like group facilitators or something like that. For me, this was a very important practice for being an autodidact was I had a methodology for like deconstructing books, for like managing my attention. And it was always really inconvenient to have to go back and forth between like what step am I on right now?

And then I couldn't just like set up this is the sequence of events that I want to do and just like look at each thing one at a time and separate out the sort of cognitive scaffolding from the actual thinking.

And so I'm trying to build a sort of higher order programming language for creating cognitive scaffolds, for guiding your own thinking, for like intentional reflection or investigation and like that kind of stuff like that, because I think there's a whole domain of programming that is about programming your mind, you being the programmer of your mind and being able to like let me think about the thinking I want to be doing and let me create prompts for myself where like I'm the evaluator. So that is what Roam really is.

It's about building a programming language for human cognition, which could be individual or multiplayer.

Well, let's wrap it there. Thanks, everyone, for listening. Join us on Discord to discuss this episode with me, Mark, and our community. Links in the show notes. And you can follow us on Twitter at MuseAppHQ. Connor, I'm so glad that you're helping us think big thoughts about how we can just be better at collective intelligence because it's pretty clear that that's a place that humanity can get a lot better. And thanks for coming on the show.

Thanks so much.