This is the full, unedited transcript of our interview with Conor White-Sullivan, founder of Roam Research. Learn more about Roam.
Anne-Laure: I’ve been reading about your journey and I’ve been stalking you a little bit on the internet and you’ve been working on Roam, or at least the concept of Roam, for quite a long time now, right?
Conor: Yeah.
Anne-Laure: I’d love to hear a bit more about what inspired you to start building this in the first place. What sparked the idea?
Conor: So in 2008 or so, I got really interested in collective intelligence. I mean, it’s pretty obvious now with coronavirus that the systems that we have, the institutions that we have, aren’t able to make fast decisions—fast, informed decisions in uncertain times. And so I was thinking about how you could build more resiliency in the society, and how you can use things like Wikipedia, and shift the paradigm of comment-based content. This is why I wanted to have a chat because it’s gonna take me some time to articulate some of these ideas, I guess, but also I’m still waking up over in San Francisco.
Anne-Laure: It’s alright. [laugh]
Conor: Yeah, the back story is I was originally interested in figuring out how you could figure out what’s actually true online. Because Wikipedia is still bottoms up with institutional news. It’s sort of like it’s adding a layer on top of it. Their trust still comes from legacy. A thing that gets vetted on Wikipedia has to have first been published in the New York Times or something like that, or from some other third party. And I didn’t think that that was going to work going into the future. So, the first problem that I worked on was trying to do this for hyper-local politics.
So the idea was that you had the people who actually lived in a town able to sort of crowdsource the pros and cons of any policy change that was gonna happen in that town and you could… You could bubble up the best ideas for an idea, or the best ideas against an idea, or for a local candidate or against a local candidate, and those kinds of things. But the more we did it, the more I saw that storing ideas as just strings of text doesn’t actually allow you to do any of the more challenging work. Especially if you have a large group of people and no one person’s gonna be able to read everything that every single person says, you need a better way to actually see the structure of arguments, or the structure of trade-offs in a decision, or all the competing factors in the decision.
So that first company… I eventually figured out that to solve the problem, a simpler problem to solve, was how does one person take 20 different people that they’re reading and start to create some sort of synthesis or math-like map of the idea space, across a bunch of different books that they’re reading, or a bunch of different observations or conversations that they’re having, and gradually be able to index into the individual things. So somebody else who’s reading their synthesis can check and say, “Is this summary actually reflective of the underlying idea?” Or, “Does point A actually imply point B?” Or those kinds of things. And then the specific thing that I found really helpful was Mortimer Adler’s “How to Read a Book”, where he talks about this idea of syntopical reading.
So there was a project in the 50’s. It was called the Great Books Series, where this guy, Mortimer Adler, and I think it was Encyclopedia Britannica, they got all the books of western canon and they indexed all the places in Hagel and Kant and John Locke and all of those classic books of western thought, and they indexed all of the different ideas that were being brought up in there. So, like from Angel to War, right?
And they summarised the core ideas and the core differences of opinion across all these different people and then indexed back to where in each of these books it’s said. So they put, almost like the Bible where they had a chapter and verse numbers, but for all these books in the western canon. And I thought that that was a really cool project that would be a lot better if it was digital. And it would be a lot better if it wasn’t just these 50 books that people had decided—this editorial decision had been made that those were the 50 books that really mattered. I was like, “I want the syntopicon or synopticon for physics and for ontology and for the Vedas and Upanishads and things that aren’t just in the western canon.” So those were where some of the original idea came from, was that we’ve always wanted to build a layer on top of the web where every person can have their mental model of how the whole world works, and they can start to share ideas across everything.
Anne-Laure: Yeah.
Conor: And build off of other people’s work and reuse it. But…
Anne-Laure: It’s interesting because you’re mentioning two different planes almost. You first mentioned speed of thinking, and then when you think about taking notes and creating an archive of your thoughts, which is more like depth of thinking. How do you see Roam helping with connecting the two, thinking fast and thinking better?
Conor: I think that you need to be able to get compound interest on your thoughts. Good ideas come from when ideas have sex basically: the intersection of different things that you’ve been reading or different things you’ve been seeing and you’re like, “Oh, this thing actually connects to this other thing.” And so you can have better ideas faster if you’re actually reviewing the old things and your building up. You’re not throwing away work. And a lot of times people forget everything that they read in books and they’re not often able to be like, “Oh yeah, this thing that I just read reminds me of this book that I read six months ago, or two years ago, or five years ago.” So I think solid foundations do make you faster in the long run. Right? If you have a solid knowledge management system, you’re able to… And that’s the core idea in the Zettelkasten. You move a little bit slower as you go through a book. But you get some nuggets of thought articulated that then end up allowing you to produce, and when it comes time for you to write something it’s much faster because you’ve already done a bunch of that ahead of time. And so you’re not duplicating it over and over again.
Anne-Laure: Yeah, that makes sense. I saw a tweet from Tiago. You did a chat with him, right?
Conor: Yup.
Anne-Laure: I read a tweet from him recently where he was saying that after a while, when you’ve been taking notes for a long time, the extra benefits of it starts to wane a little bit. You don’t need to take as many notes anymore because you keep on reusing what you’ve written about before and just connecting the dots instead of taking more notes. What do you think about that? Do you think that’s the case?
Conor: I don’t actually buy that thesis. I guess it depends on how far into new territory you’re trying to push yourself, right? So, if you learn the pentatonic scale and you learn how to play blues guitar, you can just be like, “Ah well, now I know those fundamentals and so I’m just gonna play blues all the time. And I don’t need to learn new stuff because I’ve got a steady repertoire and I can bang out great blues riffs really easily.” But you can also keep doing the same deliberate practice and keep the same level of intensity if you’re pushing out into new territory and then you’re incorporating those new notes. I would actually say, as you get a more developed Zettelkasten, you get more evergreen notes. Actually one new thought can generate 70 new thoughts, because there’s more stuff to connect it to inside. So I think the connections can increase. You might take longer when you have a new idea to connect it up to all the old things you’re thinking about, or it can generate more new insight. I read one new observation, and then I can generate more new insight because I’ve got a larger set of things to compare it against.
Anne-Laure: And what about connecting… Because I think what Tiago was talking about was finding links and connections inside of your own notes. But Roam allows you to also collaborate with other people and connect the dots across a team. So how does that work?
Conor: So right now, Roam’s a real-time collaborative, so you can have multiple people in a single workspace. I think the real opportunity for us is when you can share notes across databases and across workspaces. So I would say that today, the level of collaboration isn’t quite… It’s much more limited. It’s somebody you start a project with and you start collaborating with them from the beginning, and it’s a little bit less of discovering other people who’ve written about similar things. So I would say that, yeah… Today it’s still… And for the next year or so, it’s gonna be primarily still a single player tool. But in the long term…
Anne-Laure: How do you see that working in the future? If we project ourselves in a year, or two years or whatever, and technology is not an issue, how conceptually, how do you imagine this working? Because, for example, if we’re taking the super slow grandfather version of Roam. Like Wikipedia, for example, where it’s lots of people collaborating on content and connecting the dots. They do have a moderation system and it’s slow, and there’s politics involved because it’s like what edit is going to be more important than another. How in the future do you see Roam work?
Conor: I see a pretty strong difference from Wikipedia, right? Wikipedia has this idea that it is possible to get to a single source of truth. Through that slow deliberation, moderation process, there’s a neutral point of view and there’s one thing which just reflects reality, and I think that’s completely bullshit. I think that’s just not how… Roam’s a bit more like, I guess, post-modern in its perspective on how truth works and those kinds of things. So I don’t actually think you’re ever gonna get a situation where you have… Maybe some one person’s Roam, who’s curating a community, like they can have a slow moderation process.
It’s more similar to a federated Wiki. So the guy who invented Wiki, Ward Cunningham, realized that, I think even before Wikipedia came out, that Wikis don’t allow you to deal with original research. So, what we’re more interested in is, if I like some of your thoughts on meta thinking, if I like some of your thoughts on like how to use Roam efficiently, and you’ve made those thoughts public and you’ve made this reusable, I can just grab them and put them into my workspace. And if somebody trusts me, they can be listening to my workspace, and they can then grab your notes from me.
And you don’t need to worry about some slow, deliberate, moderation process. If somebody trusts me, it’s much more like Twitter. So when I think about the future of collaborative thinking, I think more about how Visakanv uses Twitter of just threads on threads on threads. And when he’s making threads, he’s re-tweeting himself, he’s re-tweeting other people, he’s searching through Twitter to find the appropriate tweet for each thread.
So that’s more what I think about than a slow, deliberate system. That’s the state of the world on what coronavirus looks like. Those things are useful. There might be individual curators who have done a bunch of editorial work, and are making decisions about what gets incorporated into their workspace for that.
Anne-Laure: What I like about this approach is that I don’t need to trust you because I can go back and click and click and just go back to the source and decide if I feel like that’s a good source or not…
Conor: You might need to trust me if I’m the source. So if I’m making an original observation…
Anne-Laure: Oh, yeah.
Conor: If I am, if the source is me, then you need to decide, “Okay, do I want to listen to what Conor is saying about… “
Anne-Laure: But I get that choice because eventually if everything is linked together I will get back to the original source, so I can decide if I trust you because I will find you, which currently is very hard on the Internet, right? To find out who was the original person who had that thought.
Conor: Yup. And you can see if somebody has a critique. There’s a lot of complicated design problems for us to solve when you’re thinking about the multiplayer version of it. But they’re still very related to the problems that we have to solve right now. I’ve got a collection of 100 backlinks that relate to knowledge management, and I’m deciding which of those backlinks I’m pulling up into the body of the page itself. And so taking this big noisy buffer of potentially useful thoughts and then figuring out how I wanna structure it and figuring out which one of those basically should be elevated to the body of the article.
Anne-Laure: Yeah, that makes sense.
Conor: They’re similar problems.
Anne-Laure: And so, either as a team working together or an individual, what are some of the most creative use cases that you’ve seen so far using Roam?
Conor: When I think about creative use cases, one of the things that pops to mind is somebody who created a Roam page for every ingredient in their house and then they had another page where they were… And each of those ingredients had a check box next to them. They would reference the ingredients in the individual recipes they were using. They were able to see just by opening up their recipe page what they already had all the ingredients for. That was a pretty creative use case that sort of was practical.
Anne-Laure: That’s cool.
Conor: Yeah, I mean I really like Nat’s way of drafting articles up and that’s given us a bunch of ideas. There’s a cool Roam right now where… I’m actually collaborating with one of our users on it, where we’re mapping out all the transcripts of some Eric Weinstein’s podcasts. In one of his podcasts where he’s talking about the Jeffrey Epstein conspiracy, he makes about 100 claims inside that article and he’s just sort of… He’s just talking right, but saying like, “Oh yeah, then the CIA character assassinated this person.” and like, “This is where the intelligence communities killed this black panther. And based on those things, we can argue from analogy, that this thing is possible.” Yeah, so those kinds of maps of a complicated domain.
When I think about creative use cases, the people who are using it for space repetition is pretty interesting, where they’re taking advantage of the fact that you can expand and clap bullets and like set a date for something so that it appears in the back lines of a future day and they’re able to set up basically flash card systems so that they can get information from their second brain into their first brain. That’s been another pretty creative use case that I’ve gotten a big kick out of. And I can send you some links to some of those.
Anne-Laure: Yeah, for sure. I’d love to see those. What about discoverability? Because right now the way I’m personally using it… I know that Tiago created this as a brand, but Roam does feel like a second brain. But it’s still tethered to my brain, and so usually when I re-read old notes, it’s either because I’m clicking on the link, either one I proactively created or a bi-directional one that was automatically created, or because I know exactly what I’m looking for. But once we think about it as a multi-player thing, how do you think about discovering stuff that other people created?
Conor: Yeah, right now we’re focusing on just helping you re-discover old notes that are relevant. So there’s a lot of workflows around that, around how do you take raw temporary notes, because we encourage people to use the daily notes and to do sort of brainstorm and brain dump, and just write all the things they’re thinking. And I think that the first thing that we’re interested in is, how do you build systems so that it’s easy for you to take those and basically gradually refine them. And there’s a combination of the tool, and there’s the habits people have around the tool. And so people are using it for actual knowledge gardening. Then you do kind of right now need to have a habit of tagging something basically as a To-Do to synthesize the idea further, and then periodically go back and review those and write them in a more crisp language, or basically build up your evergreen notes so that you have this library of thoughts that you’re able to get that compound interest on.
But yeah, there’s a bunch of design challenges that we’re working on. And there’s a lot of things that are going to make it easier. You can do crazy linear algebra stuff over text to find related notes, even if they don’t use any of the same words. And so I think there’s definitely a lot of opportunity there for suggesting possibly related notes, but the act of a person seeing the connection themselves as opposed to some algorithm seeing the two words are connected, I think is pretty important, because that’s where you get more insight, and you get more… The process of drawing the connections is pretty important. So you can’t really automate these things away. You can definitely suggest things and say, “Do you think these two things should be connected?” But I think the act of drawing the connection yourself is a valuable thing that we don’t wanna take away from people.
Anne-Laure: I agree. This is why it’s a proper second brain and not just a separate storage place. People need to do a bit of the thinking. And I’m loving the conversation, but I wanna keep it short so I have one last question for you and it’s a bit topical with what’s going on right now. What are some ways you think people can use Roam right now to tackle the pandemic? I feel like it’s such a powerful tool that there must be some ways it can help right now.
Conor: There are a bunch of scientists using the tool right now. That was actually one of our first… We started… We’d bootstrapped the company by working with a bunch of people in the field of AI research, particularly AI safety. So folks who were at the Center for Existential Risk, and Machine Intelligence Research Institute, and Center for Effective Altruism were some of the earliest people that we were building the tool for. And so I talked a lot when we were doing research studies with folks who are modeling pandemics. So hopefully they’re still using Roam and they’re still using it for mapping out their thoughts and figuring these things out. But I think for most people, now, if you’re not an epidemiologist and you’ve never done anything in biology, or figuring out supply chains or manufacturing, now probably isn’t the time to just start trying to teach yourself everything, to jump into the fray. I think… I tend to think more about… It’s a crazy opportunity for people to do deep work, because a lot of the other normal distractions of daily life have been removed. So, Isaac Newton basically invented calculus when he was in quarantine, Shakespeare wrote King Lear. There are lots of good precedents of people being able to do wild work because they had the time alone with their thoughts.
And so I don’t think right now is a time for people to just suddenly be like, “I’m going to address this plague.” But I do think it’s a time where people can think about what comes after this, the world’s gonna look really different. We’ve lost a bunch of trust in our institutions. And yeah, I’ve been pretty skeptical of media and governance institutions for a long time, and people are having to homeschool their kids, or are being forced to homeschool their kids. So maybe that’s more topical. It’s like if you have kids that you need to teach and you want to figure out how to create some sort of dependency graph of… Help them learn how to map out their questions and figure out, “Okay, what questions am I trying to answer? What are the resources that might help me do that?” I think maybe there’s plenty of opportunity for people to make… Yeah, guides for autodidacts, or here’s some… Now that you have some time to do research if you hadn’t had time to do research in college, since college, here’s some better ways to think about teaching yourself the things you wanna learn.
Anne-Laure: Yeah.
Conor: And I mostly think of it as a sort of a window of opportunity for a lot of people to get to spend some time on deep work or just play-work. A lot of times it’s like everything has to be some side project, or some side hustle or some business to get into, but this is a good time for you to think about religion, or mythology, or reading the books that have been on your shelf for forever that you actually wanna dive into, and get into some philosophy. I think it’s a great opportunity, or learn a skill that you’ve wanted to learn for a long time.
Anne-Laure: I agree. I’m teaching myself how to draw right now. And I have quite a few friends too who are doing lots of things that they’ve been wanting to learn for a long time but never had time to do so yeah, I totally agree with you.
Conor: Yeah, so that’s kind of how I feel like the virus should be impacting us.
Anne-Laure: Awesome. Well, thank you so much!