Transcript

00:00:00 - Speaker 1: And I don’t hold this against Apple or anybody else. There are billions of people on Earth who need great computers, and most of them are not scientists or authors or policymakers, but we believe that this has left a gap where there just simply isn’t a major computing group that’s actually focused on how to use computers in this intelligence amplifying way.

00:00:25 - Speaker 2: Hello and welcome to Meta Muse. Muse is a tool for thought on iPad and Mac. This podcast isn’t about Muse the product, it’s about Muse the company and the small team behind it. I’m Adam Wiggins here today with my colleague Mark McCrannigan. Hey, Adam, and my frequent collaborator, Peter Van Hardenberg from I and Switch. Hi, it’s good to be here. And Peter, one of the things that I find fascinating to talk to you about is your varied hobbies outside of the world of computing. What’s your latest interest?

00:00:56 - Speaker 1: Oh yeah, I mean, when I’m not, you know, brewing beer or getting into some strange creative hobby of one form or another, well, I guess I’m always picking up another one.

So lately I’ve been doing a lot of hydroponic gardening, and I’m not growing anything of legal dubiousness, but with the pandemic, first I was focused on trying to basically have some green space in my home, because I was living in San Francisco and we didn’t have a yard.

And then more recently, I have moved back to my ancestral homeland of Canada, and it gets real dark and cold here in the winter, and I wanted to make sure I could secure a steady supply of Mexican ingredients, particularly herbs and chili peppers and things that you just can’t find. Sort of north of the wall in the offseason.

So that’s been a lot of fun.

I’ve been learning a ton about electrical conductivity and how to take pH measurements and having a lot of fun with automating all the components of this system over time. So these are going to be the most expensive cherry tomatoes in the world by the time I’m done, but it’s just a blast learning about all this stuff and, you know, what kind of absorption wavelengths are better for different kinds of plants and so on and so forth.

00:02:09 - Speaker 2: And correct me if I’m wrong, but I think the term hydroponics typically refers to growing plants without soil or with minimal soil. That’s right, I assume here you also have the artificial light element as well with the northernly climate.

00:02:22 - Speaker 1: Yeah, and basically what I have is sort of, if you imagine like a 2 m by 2 m bed or maybe a 1.5 garden bed, that’s sort of cut into 3 pieces and then stacked into a bookshelf. And then nutrient and oxygen-rich water circulates sort of around these U-shaped trays and then cascades back down to a reservoir in the bottom, where it gets pumped back up to the top in a loop. And so I routinely come in and I have to top up the water, there’s a lot of water loss due to transpiration, which is where basically the water goes out to the leaves and then evaporates, and then you have to sort of monitor how much water there is, but also regularly top up the nutrients to make sure that all of the vital macro and micro nutrients that the plants need to live are present in the solution. It’s a fun little chemistry project.

00:03:10 - Speaker 2: And I would be disappointed if there wasn’t, I don’t know, a raspberry pie or something connected in this mix somewhere.

00:03:17 - Speaker 1: Oh yeah, there are 2 computers now involved. There’s one that doses out the nutrients using peristaltic pumps.

And there’s a separate one that came with the unit, which handles notifying me when the water levels are too low, there’s a little ultrasonic sensor that can tell when the reservoir tank gets low on water, and also handles scheduling the lights off and on during the day.

I’m still hopeful about automating the nutrient sensors, but it turns out that monitoring pH over time is actually a surprisingly difficult problem, and that a good hardware solution actually involves a fair amount of like upkeep and maintenance and expense just in sensor probes alone. So so far I haven’t quite taken the plunge there, but I think it’s just a matter of time. We’ll get there, don’t worry. The end state, of course, is to build a robot arm that will like use computer vision to spot when a cherry tomato is ripe, and then pluck it for me and place it on a conveyor belt. But uh, you know, I think we’re still a few years out from that, both in terms of like technological feasibility, but also just in terms of like, I got a lot of other projects on the go these days.

00:04:23 - Speaker 2: And surely plucking the ripe literal fruits of your labor is the funnest part of the whole thing, so save that for last to automate.

00:04:33 - Speaker 1: Yeah, my kids really enjoy going into the garden. They get the little step ladders out and climb up and peer into the different tiers and pluck leaves to sample. Honestly, this thing is the only way I’ve ever convinced them to eat green leaves, because they’ve learned now that there’s lots of interesting green leaves they’re allowed to eat in the garden. But if you put a salad on their plate, they won’t eat it, but they’ll go and like forage some mint leaves or basil or things like that. They’re into that.

00:04:57 - Speaker 2: And I mentioned we’re frequent collaborators, so the three of us worked together all the way back in the Hiroki days. Nowadays you are leading, administrating, directing the I and Switch Research Lab, which was a position I had previously held before entering the new spin out, which we can talk about the relationship there, but I’m sure the audience would love to hear about your background. How did you end up connected to this unusual band of misfits and unusual space in computing?

00:05:26 - Speaker 1: I always tell people that I like to move like a knight across the chessboard of the industry. I don’t like to go in a straight line and I don’t like to follow an obvious path, so I move a little bit to one side and a little bit across.

So before Ink and Switch, I had been working at Hiroku with the two of you. But my background is super varied. I have been an Arctic oceanographer and spent time collecting data at sea.

I’ve had waves roll over the side of a flat bottomed research vessel and swamp my boots while I was working on the aft deck. I’ve worked in game development and written physics engines for the Game Boy DS.

You know, it’s really hard to ride a physics engine using only integer math. You really sort of take the ability to like divide numbers for granted in day to day computing, and so that was a really cool challenge.

Yeah, I’ve done computational Shakespeare, which was cool. There’s the Internet Shakespeare editions are like a scholarly Shakespeare, so if you want to be able to ask questions about, like, can we tell who typeset this page of an original edition by analyzing their use of combined letter ligatures. You know, because they think they can tell different typesetters had different habits for where they would draw type out of the frames. But all those kinds of questions about provenance and genuineness of Shakespeare, we were building technology to enable that kind of computationally. So, yeah, I’ve done a lot of weird and wonderful things before I wound up at I and Switch, and if I’m lucky, I hope to do many more in my career, so, so far so good.

00:06:57 - Speaker 2: And then another important piece here is, what actually is Ik and Switch? What does it do?

00:07:01 - Speaker 1: Right, yeah, I and Switch is an independent industrial research group.

We’ll break that down a little bit more, but basically, the way I describe what we do to people who are sort of outside the field is that we think that computers have much more potential to Make us smarter and more capable than we’re really seeing happen with the kind of technology platforms we have today, and we know that that’s true, but we don’t really know how best to pursue it, and we sort of feel like a lot of the fundamental requirements to be able to do that work aren’t in place yet.

And so, I can switch research. It is about trying to light the way to make possible these kinds of breakthroughs in the way people build software and the kinds of software people can build by unblocking and discovering paths forward for people.

00:07:57 - Speaker 2: And if one were to flip through the publications on the I can switch website, I’ll link in the show notes, you’d quickly get a feel for, I think a lot of research areas that are things that Mark and I mention and with our guests here all the time on this podcast. So end user programming, for example.

Yeah, what a coincidence, increasing the accessibility of programming, obviously kind of tablet touch interfaces or next generation interfaces generally.

Local first software is a huge one. We had Mark Klepman on just. Recently to talk about that, and there’s many kind of sub branches of that sort of you think of it as like oh what comes after the cloud or how do we give data agency back to creative people.

So those topics, many, many of the things that we talk about here really do come from or have their roots in the research there, and the research is ongoing, so those ideas are continuing to be developed and pushed forward by you and your team.

00:08:48 - Speaker 1: Yeah, there’s an endless amount of work to do, and I think the hard part is really just sort of continuing to find the efficient frontier, right, which is where are the most important unanswered questions, who are the people that can do that work, and how can we convince them to come and spend a bit of time with our ragtag band of adventurers?

00:09:09 - Speaker 2: Could you give us an example of something you’ve published recently on, just to give a taste for the kind of research work that’s undergoing there?

00:09:18 - Speaker 1: I’ll give a little preview of two unreleased papers that are coming up. Maybe that’ll be even more exciting for folk if anyone out there follows us. We have two pieces that are currently in editing and sort of preparation for publication.

One of them is called Inkbase and the other one is Paratext.

Now, Ibase was led by James Lindenbaum with Shimon Kliski, and it was an exploration of how you could program ink. Now, One way of looking at this is to say, well, what should the programming language for Ik be? But in fact, we sort of deliberately cut that out of scope. What we wanted to know was what would an ink programming environment feel like? What would it be capable of? You know, how could you use it if you had it, even though we didn’t have it, right? As a research project, what we wanted to know was kind of like, well, what would happen if you did? This was also in partnership with Josh Horowitz, who came to us from Dynamic land on his way to grad school.

00:10:22 - Speaker 2: So we built an environment that allowed you to program ink drawings to make them interactive, and ink here you mean the digital ink, the same exact kind that is in use, except there you scribble on your board or whatever, and that ink can be moved, it can be deleted, it can respond to To your interactions and create interactions of its own.

00:10:31 - Speaker 1: And so we saw really cool ideas explored there like Shimon built a music sequencer that produced MIDI out of the system and let him drive sound synthesis engines using drawn lines.

James demonstrated how to build like a fitness tracker, you know, sort of like the Seinfeld streak that people often use. What if you could make that a programmable thing or like a to do list? You know, like a Zettel casting or a personal GTD kind of system, but one that you could mold and actually make smarter over time by adding interactions as you develop them the same way that sort of like an Excel spreadsheet grows out of a simple table of numbers into some fully featured and powerful computational environment.

We want to know what happens if a drawing evolves over time to become a computational environment that responds to your needs. And your discovered requirements. So that’s the ink-based paper. It was presented at live very briefly this year, but in classic Ink & Switch style, we have a massive paper going into all kinds of detail and with links to some code coming soon.

And that kind of represents one angle of our work.

Another piece that we’ve been working on, we’ve talked about local first, you mentioned earlier, we’ve been exploring something called CRDTs, that’s a conflict-free replicated data types. A really gratuitously confusing name for what is basically maybe more usefully to think about like Git for data structures, mergeable data maybe mergeable data, yeah, if you have like a JSON file in Git and then you check in changes and then somebody else checks in changes and then you go to merge and you just have like some kind of nightmare of text editing and comma placement and figuring out what really happened. The idea behind automerge is like, well, what if the data structures were smart enough to synchronize themselves and what kind of new capabilities would you get that? And that’s really been sort of foundational research that’s enabled a lot of local first work over the last few years and we’re by no means the only game in town, but one of the problems we keep running into with that is we write a lot of these big essays and we want to have local first tools for it. But today the kind of state of the art is like, OK, well, you can do plain text editing. Or you’ve got nothing, and our essays are these like complex multimedia documents with videos and asides, and we won’t be able to make suggestions and comments all throughout them as we edit. And so, what we needed was a rich text CRDT and that is to say a CRDT that was aware of formatting and spans. And we talked a lot in the Paraex paper, which is coming out soon, I hope, is in concert with Jeffrey Litt, who led that project with Slim.

00:13:12 - Speaker 2: Notion, former guest of the podcast, in fact. Yeah.

00:13:15 - Speaker 1: So we’re looking forward to publishing that because it turns out to be a surprisingly subtle problem. If you’ve ever dealt with non-hierarchical data, like extent data overlapping extents, it just leads to a lot of like really interesting and subtle, quite literal edge cases. So, we’re eager to present that to everybody as soon as we stop finding bugs in the implementation. I think we’re close, we’re close.

00:13:43 - Speaker 2: I think we said before we talked with Lis Lee in a previous podcast about whether software is ever done and certainly research is never done. You have to just kind of draw a line in the sand somewhere and say we’ve advanced the state of the art enough that we think others can build on this. Let’s publish.

00:13:59 - Speaker 1: Oh, I have a great Dune quote for this, it’s right in the zeitgeist now.

I’ve been using this line for years, but the way I think about this is in the words of Frank Herbert, Araki teaches the way of the knife, it is finished because we stopped here. Yeah.

So by that, I mean, you know, you kind of go into a space and you work for a while, and you can spend your whole career bashing away at one problem, but one of the things I think we do really well and actually that we’ve inherited from your time as lab director is like this ruthless discipline about sort of pencils down on a given date and moving on. If you don’t do that, it’s really easy to get caught up in what you’re doing and to continue to pull that thread, you could be at it for the rest of your career. And maybe that’s good, but by stepping back and sort of re-evaluating, it forces you to kind of Both take the time to publish, but also to sort of take the time to think about what’s next and what’s really important now.

00:14:49 - Speaker 2: I think that maybe this is different for other researchers or labs, but Frankenswitch, I think we want to take kind of a breath first search of the adjacent possible for how we can improve creative computing rather than sort of the depth first, follow one fork in the trail and go as deep as we can.

00:15:07 - Speaker 1: Yeah, the way I talk about this is I often describe us as a lighthouse, you know, we can’t necessarily walk the path for everybody, but we can light the way and hopefully show people how to get to where they need to be and show them opportunities that may not have existed before.

And related to that, one of our core values is to produce actual demonstrations of our ideas, right? Real tangible objects, computational objects that a human can use.

You know, one is this just kind of comes from our background as like hackers and entrepreneurs, you know, you have to go and build it, but it also kind of comes from a DARPA sort of motto. DARPA, of course, is the American Department of Defense Research Group. And they talk about how you really have to show somebody a robot climbing a wall before they’ll believe you. It’s one thing to say, oh yes, robots can climb walls now, and to write papers showing sort of thrust vectors and forces and friction and so on. But if you walk into a room and a robot climbs a wall, then you believe. So we like to show people the robot climbing the wall, that’s what we do. That’s a long way from having robots in every home, but, you know, once you’ve seen it happen once, then you believe.

00:16:18 - Speaker 2: Well, I think we sort of naturally slid into the topic here, which is independent research and actually you previously described I can switch as being an independent industrial research lab, so it seems like it’d be worth breaking those down a little bit. I guess I’d like to start actually really fundamentally what is research and how does it differ from other kinds of work you could do in the computing field, and then what do those modifiers mean for you?

00:16:43 - Speaker 1: Yeah, so, OK, what is research? In the words of Richard Hamming, who I think is quite an influential figure in this space, his art of doing science and engineering, essay, lectures and book. There’s a really great book from Richard Hamming, published by Stripe Press. If you haven’t got a copy of that, you should definitely pick it up if this stuff interests you.

But he said, if you know what you’re doing, and it’s science, you should not be doing it. And if it’s engineering and you don’t know what you’re doing, you should not be doing it. So science is oriented fundamentally towards the unknown. That’s what research is. If you know how to do it, you shouldn’t be doing it. You should be oriented towards the unknown.

So, as a research group, we are oriented towards how to solve questions, how to answer questions that nobody knows the answer to, or at least that we don’t know the answer to and can’t find, you know.

There’s an old joke that 2 years in the lab could save you 2 hours in the library.

So we do try and find prior art and do some survey of the space before we do research, but, you know, sometimes you miss things and then you find out when you go to publish that somebody else had already written about that.

But that’s OK. We’re learning as much for ourselves as for anything else.

So that’s research, but of course, research comes in a lot of flavors, right? There’s sort of pure research, you know, if you think about the National Science Foundation will fund you. To do things potentially if you have the right connections and proposals that may not bear fruit or may not have any direct industrial application ever, or certainly not for hundreds of years.

And a great example of this would be prime number theory. Prime number theory is at the very heart, the very heart of cryptography.

Now it is extremely important mathematical research.

But most of the core of that research happened like 100 years before anybody found any use for it. So it’s undeniably good for society that this research was done. I mean, depends how you feel about cryptography, but let’s say that it’s good for society that this research was done, but the people who actually did the research, they were dead for centuries or decades at least before any real fruit of this sort of transformational work was known. And that’s not to say that there isn’t value in it, it’s just industrial research. What we’re trying to do is be much closer to what’s happening in the world today, and to sort of connect problems that we see in the real world of humans and software authors and businesses back through this sort of lens of what we care about to doing research. So we want to identify problems or we don’t see solutions. And then we want to develop candidate solutions that could be applied in at least the span of our lives or hopefully over the next few years.

00:19:27 - Speaker 3: Yeah, and I think importantly, it’s a two-way relationship with the industrial ecosystem. You’re solving problems that have some potential industrial application, but you’re also drawing from the experience of industry, you’re talking to the customers, you’re understanding the current technology and what’s possible, you’re aware of the trends and things that are happening and it’s that sort of alchemy of not only solutions to problems, but potential that you’re identifying from the industry.

00:19:51 - Speaker 1: That’s right. And so I think you often find industrial research groups inside big companies like Microsoft Research or Xerox PARC, or Bell Labs.

So there have been a number of these very impactful industrial research groups in recent decades, but of course, all of those examples I just listed. are anything but independent.

And this brings us to sort of what is perhaps most strange and most unique about ink and Switch, which is that we are not part of an academic body and we are not part of another corporation.

We are autocephalus, we are our own head, there’s nothing above us, much like the strange Greek monasteries of the Orthodox church that date back their independence thousands of years.

Well, OK, we don’t have thousands of years, but, you know, the independence is a blessing, right? We’re not sort of tied to the Quarterly returns cycle of some business, but it also creates all of these kinds of pressure and constraints. We don’t have this like cozy relationship with the money fountain that will keep trickling us budget year to year. We have to kind of find ways to carve out our existence as we go.

00:21:03 - Speaker 2: And if you’re interested in kind of all the ways that research and labs work in the world, I point the audience to Ben Reinhardt’s work. He’s got a pretty Extensive set of writings on this. He writes quite a bit about DARPA, which you previously mentioned there, Peter, just because they’re one of the more interesting long running government sort of research funding institutions. And he speaks about ink and which specifically as sort of inverting this normal relationship between an innovation organization and it’s money machine, right, which is that the corporate research labs, yeah, Bell Labs certainly is has been a creator of so many incredible things from lasers to GPS. To the transistor, for example, and Xerox PARC, of course, for its short run was legendary. There’s many others like that, say the skunkworks, but typically you have a corporate parent whose job is they’re an industrial company, their job is to commercialize to sell things to make money, and then the money they make from that, they want to put into this research arm. But it does mean that always that that organization, of course, is subservient to that.

So with it could switch the idea here a little bit is to kind of switch that around a bit, or at a minimum, make it kind of standalone.

00:22:15 - Speaker 1: Right, and so for us, you know, there’s sort of two puzzles, one which is, you know, how to get our work out there, how to have an impact, because You know, as an industrial research group doing work but not actually impacting the industry, you know, what was the point, you know, you could have just stayed home and watched Netflix all day, would have had the same end result. So it’s important to us to actually sort of change how things happen, but we also need to find ways to do that sustainably over time.

So we have a number of hypotheses, you know, as with all kinds of like, long term projects, you tend to have long cycles before you see results.

But we believe that there’ll be a combination of strategies that ultimately make this thing self-sustaining, and one of those is to commercialize our technology through producing spin-out companies and then maintaining sort of a share in those companies.

And of course, the flagship example of this is Muse itself, right?

00:23:13 - Speaker 2: I mean, that’s kind of the videos.

00:23:14 - Speaker 1: Yeah, spoiler alert. Here’s the big reveal, you know, Muse came out of work we were doing at I can Switch. And the original version of Muse was a prototype built at the lab, and of course, the 3 founding partners, 3 of the 4 founding partners, were there 4 of you in the beginning?

00:23:32 - Speaker 2: We were 3 actually, so they were all 3 lab participants.

00:23:35 - Speaker 1: Yeah, so, all 3 of the founding partners of Muse were ink and Switch staff, and of course, that was the 2 of you, and then Julia as well.

So the lab helped develop the concepts that led to Muse. It brought the team together and it built the initial prototypes, and then as a spin out from the lab, the lab retains some stake in the company.

So we’re not only delighted to see our ideas put into practice, we’re incredibly excited to see the work that you’re doing and the testing of our values in the world, but we’re also sort of directly incentivized to see you succeed.

And so I really love this kind of like Symbiotic relationship where we have both proof in the market of our ideas being feasible, but we also have this incentive to follow closely and make sure that, you know, we’re doing research that can help and that we’re communicating with you and vice versa. So I think it’s a really great relationship and I’m looking forward to many more of those as the years go on.

00:24:36 - Speaker 2: Yeah, I hope at least the flywheel that we’re trying to go for here, but as you said, it’s a very long cycle to prove that out is do research that is driven by real world problems. And of course, basic science and just the pursuit of truth for its own sake is absolutely incredibly valuable, the prime number theory you mentioned or maybe Gregor Mendel sitting there breeding his pea plants just to learn about how the world works or to learn about mathematics or computing’s most basic principles is worthwhile, but in it which really goes after things that are related to real world problems, all the ones that are maybe too far out, for example. Startup or a commercial entity to tackle.

00:25:16 - Speaker 1: What a great segue. Let’s talk about why Ink and Switch exists a little more. What we’ve talked about is the notion of like independent industrial research in the abstract. You could do this kind of independent industrial research in any field, material science, you could build spaceships if you had enough capital, which apparently some people do, you know, you could study journalism, as long as you have this sort of like connective loop.

But in the words of Thoreau, it’s not enough to be busy, so are the ants. The question is, what are we busy about? And what we’re busy about is sort of this despair that computers have increasingly become mechanisms entirely for consumption, and that so many of the groups and bodies that were building the bicycles for the mine that were pursuing this sort of vision of computers as these intelligence amplifying devices have kind of retreated from that vision towards a more Natural consumer demand. And I don’t hold this against Apple or anybody else, you know, there are billions of people on Earth who need great computers, and most of them are not scientists or authors or policymakers, and so it’s natural that they would sort of follow the pull of their users and go to where the market is. But we believe that this has left a gap where there just simply isn’t a major computing group that’s actually focused on solving problems.

Around how to use computers in this sort of intelligence amplifying way.

And specifically, the reason why a research group exists to solve this problem is that people are, oh, well, startups are these great innovators. That’s only true in a very narrow way.

All three of us on this call here have plenty of background in startups, both working in them, and founding them, advising them, etc. And startups can take certain kinds of risks, but they can’t take every kind of risk, and it’s sort of like the old advice about only break one law at a time, you know, it’s the same thing with startups, which is you should only take one risk at a time, and generally for the startup, there’s like a core hypothesis to the startup that they are trying to test. It’s Some new kind of product or in some new kind of market.

And so the advice people give startups is, well, you should be conservative in all your other choices because you’re already taking a massive risk on this one axis. And so the way we see that play out is lots of people are building startups that would be much more effective from, we think a software and user perspective as local first software, but they build them as cloud sass because that’s what the market expects and that’s what’s easy, that’s what’s known how to do. It’s also what venture capitalists expect.

Right, venture capitalism is, let’s be honest, mostly about pattern matching and comparing with other past successes and trying to into it based on this sort of both guiding and being guided by the structure of what the market is doing right now. And that’s fine too. It’s just, this doesn’t leave much space for the kinds of major transformative direction. That a research group can pursue. We can think a decade out. A startup can only think 2 or 3 years out at most. And so that means that we have the opportunity to do work that simply isn’t possible in the context of a startup. And in so doing, we can bring those time horizons in closer and bring things into sort of striking range for a startup, and then those startups can go and pursue these ideas. And in fact, we’re seeing that all the time these days, not just with Muse, but You know, we’re seeing lots of startups these days who reach out and contact us and say, hey, thanks so much for your essays, they were really inspiring, they’ve been really influential on how we’re thinking about this, and we’re hearing from venture capitalists when there are local first startups in some new domain, we’re seeing tangible results of this work on the industry today.

00:29:10 - Speaker 2: Your point about markets and the fact that sort of even startups or commercial ventures, I think just generally speaking, need to follow market trends.

It would be dumb not to. In fact, that’s almost the definition of a good opportunity is capitalizing on an immediate trend and In fact, what we see that for example, ARPA slash DARPA, what they do is try to change the whole industry. So one famous example of theirs is they decided in the early 2000s, we think self-driving cars should be a thing, and they put their money to work with prizes and other kinds of funding grants to Get a bunch of both researchers but also companies and investors interested in that, and they basically kicked off that whole revolution that has yet to perhaps fully yield fruit, but at least you see they got a bunch of people caring about it, they got a bunch of money into it, they got a bunch of smart people who want to make that future come true.

00:30:05 - Speaker 1: Yeah, and DARPA refers to this as the challenge-based model.

So they use this trick everywhere. DARPA doesn’t actually do any research in-house. They are fundamentally a funding agency.

And so, with self-driving cars, they said, hey, we’re gonna give you this challenge, which is you have to drive this car across the desert. And the first year, I think none of the contestants made it to the finish line at all, but each year it got closer and closer until they were eventually, there were entrants that were able to traverse this whole landscape and they won some prize money, but they talk about at DARPA how The way they evaluate projects is that they want things that are really ambitious in their outcome, but they also want things that are tangible in their direction. And so they need to be both very hard, they call it DARPA hard as sort of like a discriminator of ambition. They don’t want something that’s just difficult, they want something that’s really difficult, but they also want something that can be articulated in like a real world results that can be measured or evaluated.

00:31:06 - Speaker 2: Yeah, so obviously I can switch doesn’t have that same structure. We do the research in-house. We don’t have government money to spend, etc. but I think just broadly this idea of moving the market rather than going with the trends that exist, we’re saying we actually don’t like some of the trends where the massive consumerization of computers, which has many nice benefits like most people use the computers in their pockets as Messaging devices, right, and stay in touch with friends and family. Great, that’s good, but then who is working on the creative tools? Creative tools are just not sexy, not interesting, not where it’s seen to be the place that smart people or smart investors put their effort.

00:31:45 - Speaker 1: Well, the smart investors is, I think the secret here. This is the core, right, which is that if you’re an investor, you need a Big market and it turns out that a bunch of novelists and like researchers, there aren’t a billion novelists and researchers to buy your product and if there were, they probably don’t have the budget themselves to go after it. So I think it’s perceived as a relatively small market, though we think there’s a lot of opportunity there if you build tools focused for them.

00:32:10 - Speaker 2: I don’t know, Microsoft Office did OK back in the day. Adobe seems to be doing all right.

00:32:14 - Speaker 3: That’s true. Yeah, so here’s my variant on this. I think there are actually 3 big markets that are already very well funded and being effectively pursued.

One is what I would call consumer software, this is ad funded, engagement based software, Facebook, and so forth.

There’s enterprise software, which is perhaps actually the most profitable domain right now. This is now Microsoft. AWS stripe, things like that.

A third one actually that we often don’t mention is like surveillance and weapons related stuff, like kind of government stuff. That’s all very well funded.

So this is an area with individuals that’s perceived to be less economically lucrative. So the entirely predictable and predicted outcome is that there’s just much less investment and therefore, much less progress in it. And we’re trying to jumpstart that flywheel a little bit and get some more progress in this area.

00:33:00 - Speaker 2: But I do think it’s not just on the investor side, we talked to Jason Yuan about this, which is, for example, doing design work for designers, do they want to go to work on the next Excel or do they want to go to work on the next, I don’t know, Snapchat and.

At least in the recent past, I feel like these consumer facing companies because they are so big because they have such big outsized brands, but you’ve heard of them, but also they have so much money to put into it, they can do a lot to build those brands and be perceived as a place that the best talent comes to.

So I might think of one of the goals of use and I can switch and all the other folks who are in our sphere of creative tools and tools for thought and so forth is just making that a cool area that people are drawn to work in.

00:33:47 - Speaker 1: Obviously you got to pay your rent and Nothing says cool like 6000 word essays on esoteric technical subjects, and then we all got our own definition of cool.

00:33:55 - Speaker 1: Oh yeah.

00:33:56 - Speaker 3: I think that’s true. I think that gets to this idea of institutional alchemy, which is it’s not a matter of just going to a space and working on it or even of just providing funding for it. You gotta bring together all these different factors. There’s funding, there’s vision, there’s people. There’s memes, you know, there’s lines of research, you need it all to be there. Communities, yeah, exactly. And one of the things I’m really proud of with ink and Switches I think we’ve managed to do that somewhat for this space. There’s this nexus, this scene that is formed around the work.

00:34:29 - Speaker 1: Yeah, I think that’s true. It’s very exciting to see a growing space, and of course, we’re talking about Ink & Switch and so I’m sort of focused on that, but I want to note that like we’re by no means operating in a vacuum, and there are plenty of other people like Gordon Brander out there who are writing about this, and there’s lots of startups like Rome Research, who are out there building products in addition to Muse, and I like to think that we are. Certainly a significant player in that ecosystem, but I don’t mean to sort of imply ownership over it or take credit for what is by every indication. A rapidly growing and very dynamic space. I think that’s true.

One thing I would like to talk about is why this is important, and why we’re doing this. Now, I think all of us, sort of longtime members of the lab. Believe that the world is kind of in a rough spot right now, like in a big picture sense, right, we have major climate change problems ahead of us, we’ve got this sort of breakdown of the old social order, which is largely driven by new dynamics and communication tools through social media, without getting too much into that, there’s these attentional economy problems, right? You know, you go to unlock your phone and there’s 75 push notifications waiting for you from a mixture of actually useful things and marketers that have somehow managed to slide into your mentions. So you’ve got like all of these kinds of, like, real major issues, and here we are monkeying around with CRDTs and programmable ink. You know, I’ve had friends who are like, if you think these problems that are facing the world are so important, why are you playing with computers? Why aren’t you out there working on the front lines of climate change? I think it’s a good critique and it’s a fair question, but I think, specifically our work. And our motivation behind our work is that we want to empower the people who have those first order skills to be able to attack these problems. We want to support physicists, we want to support journalists, we wanna support writers and authors, and there’s this whole pipeline of knowledge. Pipeline is Gordon Brander, if you’re listening. There’s a systems theory explanation here. It’s not a pipeline, there’s lots of feedback loops at every step along the way, but there’s this system of cultural change that occurs where real knowledge about the universe produces change in the way that we live. And at one end of the pipeline are scientists and social theorists and other people, activists, and then, you know, it goes through a long process that involves a lot of different parties and systems like journalists and writers who bring these ideas to the public and talk about them and communicate them and eventually winding up with policymakers and entrepreneurs and government officials who either change public policies, introduce new laws, or start new businesses and pursue opportunities revealed by this. And so all of that kind of structure. You know, we believe that there’s a lot of unnecessary friction and loss of productivity caused by these bad tools. Every time someone who could be working on climate change loses their data because whatever, the computer reboots to run an update or, you know, they have some great idea and they go to take a note, but they get distracted before they take that note, because their phone is telling them to play whatever Flappy Bird or something. Like anytime those kinds of things happen. That is slowing down and taxing our ability to respond to all of these problems. And so we as a group, the people in the lab, don’t have physics degrees. We’re not social theorists, but we hope that by working with journalists and talking to them, and working with scientists and talking to them, and working with writers and talking to them, that we can learn what the challenges these people have are. And develop a set of values and theories and practices that we can then build infrastructure around that will enable these people to be more effective. But why that, right? The answer is just, this is what we’re good at, this is what we know how to do, right? We’re taking our skills, our unique advantage, and we’re trying to apply them indirectly, perhaps, but sincerely and over a prolonged time scale and intentionally towards enabling those kinds of changes.

00:38:58 - Speaker 2: 80,000 hours has a kind of whole set of essays and theory about how to apply your career in the most effective way as possible to do good in the world, and they reference exactly that point you just mentioned, which is you have to find the overlap between a meaningful problem in the world. What you specifically are great at, drawn to or passionate about, or you have skills or knowledge or experience that very few others have. And the third one actually is an area that not a lot of people are working on. And so I think that’s part of what drew us to this space for for and switch.

00:39:31 - Speaker 1: Not a lot of people were working on this when we started here. Certainly there have been times when people have kind of looked at me and been like, why aren’t you doing AI or ML or blockchain or cloud software or like any of the things that anybody ever raises any money to work on. It’s like, well, doing what everybody else does takes you where everybody else is going. If you want to solve interesting problems, you’ve got to work on things that other people aren’t already doing. That’s part of it.

00:39:57 - Speaker 3: Yeah. Another angle on the why that I would give in addition to this like kind of productivity and effectiveness motivation is one of just aesthetics and beauty and sentiment.

I talk about how we are using these tools 468 hours a day, and it’s incredible. demoralizing if they’re not beautiful, if they don’t feel good.

And I think that is the state of a lot of our tools. And you would feel terrible and in fact, refuse to go to work on a really creative project, 8 hours every day in like some terrible concrete block building, right? It would be demoralizing.

That’s a situation that I feel like we are in with a lot of our tools. And just making people feel better about what they’re working with well I think also enable more good work.

00:40:37 - Speaker 1: Yes, we’re asking all of the artists and scientists of the world to do their best work inside a Walmart full of billboards, while the internet is screaming at them. Yeah. It’s not ideal.

00:40:49 - Speaker 2: Yeah. Yeah, to sort of summarize what both of you said or perhaps my take on it would be that computers and software and the internet have become a really fundamental infrastructure in the world. They’re like roads and bridges, they’re like our agricultural systems and our economy, and those things need to work well and for us to do anything else, solve any other worthwhile problems or do. Things like make great art, create beauty in the world, and computers have gone from being something that is maybe more of a toy or a novelty to being something that is fundamental and everywhere. And so them supporting the humanity’s best aims and noblest pursuits versus being swept up in that Walmart of billboards demanding your attention is something we think will produce returns for all of society.

00:41:41 - Speaker 1: Talking about all of society, I think one of the really valid criticisms of both our work and a lot of the other work in this space is that it is rooted in this kind of technocratic elitist view that there are experts and brilliant people out there in the world, and what we really need to do is enable them to do better work.

You know, I have young kids and I was at sort of some 3 year old’s birthday party in a playground.

And I got to talking to a golf course equipment manager who lives here in this town, super nice guy, sort of opened with, oh, I don’t know too much about those computer things. But he’s interested in what we did, so I was trying to sort of translate my work for him because, you know, it’s nice to hear about what other people do and share what you do.

00:42:23 - Speaker 2: Just as a note, the cocktail party conversation, which is exactly what I described this, trying to bridge, you know, describe your nichey work to someone that’s completely outside the field, was one of the biggest challenges for me working on Incode switches very hard to summarize that. It’s gotten a little easier since I’ve been onus cause I can just say, yeah, we’re making an iPad app.

00:42:43 - Speaker 1: Well, so, funny enough, this gentleman who claimed not to know much about computers was actually more plugged into the problems of our industry than anybody I know in Silicon Valley that you just pick off the street. He starts telling me about how he doesn’t trust the tractor companies. Because they’re taking all their data and putting it in cloud services and it’s not necessarily to the benefit of him or the equipment managers at the golf course, and how he’s been building all his own software and FileMaker Pro to do all of this stuff in-house, because he doesn’t like that all this data is going to these.

Enterprises who are deciding what they pay and how much maintenance is going to cost and all these things, and he’s trying to maintain his independence and he doesn’t trust them not to change the data. I’m like, oh man, you get this work better than almost anybody I talked to. You understand the problems cause you’re out there on the edge. You’re the guy who is suffering because of all these venture capital backed startups and this big push to cloud sass, like, yeah, there’s a lot of benefit, but there’s a lot of cost. And this guy sees it in a way that most Silicon Valley software people don’t, because he’s feeling the pain. And so one of the things that I think we need all to do as a group is do a better job of like.

Realizing that climate change is not something for scientists to solve, it’s too late for that, right? We need resilience at every level of our communities. Everybody is going to be finding new ways to live and working on new solutions to these problems, right? Social media driven sort of like cultural collapse. It is not something that only media theorists are going to solve. It’s all of us on our phones, on our computers, in our lives are going to find new ways to communicate. We’re going to need new tools to find that quiet space for ourselves, whether you’re, you know, designing knitting patterns or whether you’re Organizing neighborhood barbecues and potlucks, right? Like this is not solely a problem that affects sort of elite thinkers in society, and we should be building tools for everybody and thinking really about deep resilience through all of society. And so I think that’s something that I’ve been challenged on lately and I’ve been thinking a lot about and so I want to encourage everybody listening to remember that like this is an all of society problem and so we should be building all of society responses.

00:44:57 - Speaker 2: Well, after that grand and sweeping discussion of the big picture stuff, I’d actually be curious to just get a little nuts and bolts. What does an Ink & Switch project look like? Mark and I have talked before here about the Hollywood model a little bit, and you talked in the very beginning about the pens down, the strict pens down at the end of the project, but for example, two of these two projects you mentioned earlier, what does it actually look like start to finish?

00:45:22 - Speaker 1: So, we have a number of sort of active areas of research, and there are other areas we need to expand into and haven’t, but that’s a separate problem.

So, to do a project, the first step is to find a problem, and so we have a process of sort of developing project briefs we call pre-infusion.

And so, you know, around the lab and in our community of past and future collaborators are a lot of people who are sort of tracking these kinds of interesting problems, but a good Ink & Switch problem is one that we think is really important. That we don’t know how to solve and that we don’t think somebody else will solve for us if we just wait a little

00:46:03 - Speaker 1: bit.

00:46:04 - Speaker 2: So maybe an example here might be the rich text CRDT you mentioned earlier. We’ve done lots of projects with CRDTs because that’s a great technology for collaboration and local first, but then anytime we go to add text, you go, can we make this rich text? Nope, because there’s no such thing as a rich text CRDT. OK, well, put that in the future research bin somewhere.

00:46:26 - Speaker 1: Right, and I do want to point out that Kevin Jan’s uh YJS project has a rich tech CRDT, but it had some properties that we felt didn’t make it suitable for the particular task we had at hand, and also, while he’s done a ton of great work there, There’s no sort of documentation about how to do this that other people could pick up and draw from because it’s, you know, an implementation he wrote to solve his problems.

So we wanted to write something about this, both so that we could implement it as well, but also so that other people could learn from this and build on this work in a more sort of open way. So that’s a great example of a problem.

So basically do some research upfront. We usually have one person doing this sort of pre-infusion thing, they’ll look at the state of the art, they’ll try and scope a problem, they’ll define what we need to do, and they’ll kind of write up like a 1, maybe 2. Page description of what’s the problem we’re trying to solve and how are we going to solve it and who do we need on the team to be able to solve that.

And that’s important because as you’ve talked about the Hollywood model before, but for folks who haven’t heard that episode, we sort of treat each research project kind of like a film production. We put together a team of experts or domain knowledge or generalists with the right kind of philosophy, depending on the project. To execute on some vision, and then when it’s over, we ship the paper and most people move on and go on to whatever they’re doing next, just like you would if you were building a film, rather than these sort of like permanent staff researchers. So we do have a couple of folks like that around as well. And so that’s kind of the pre-infusion.

00:47:57 - Speaker 2: One thing I’ll just note briefly on the staff side is that being able to hire people kind of as short term for these projects, which are usually a couple of months, means that you can get the exact combination of skills that you need.

When we’ve done more kind of innovation oriented R&D within an existing company that has a standing staff, you know, when it comes time to say, well, we need to implement this back end system, what are we going to use? Well, we use Erlang because we have an Erlang programmer on the team rather than coming at it from what is actually the best technology or the specific skills that we need and go.

Find those people.

I think one of the more dramatic examples to that for me, Peter, was when you came into the lab and started advocating for CRDTs as a solution to the kind of sinking problems we’ve been thinking about a little bit, and then we decided to do a project around that, you know, this was 2016 or something, because this was such an early time for that.

There was probably only 10 or 12 people in the world who you could consider experts or even reasonably knowledgeable about CRDTs.

We literally made a list, you made a spreadsheet and contacted every one of them and saw who might like to work with us or who did we vibe with, and that’s how we met Martin Kleppman.

00:49:08 - Speaker 1: Yeah, and we’ve been working with him ever since.

00:49:11 - Speaker 2: Yeah, it turned out to be a great collaboration, but it’s a good example of where You don’t start from, you know, you say, well, we’ve got some post graphs experts on the team, let’s do post graphs. No, you start from what is actually needed if you look at the whole world of skills and technology and design needs and what have you, and then go find the specific person or people who have that exact skill.

00:49:33 - Speaker 1: Yeah, and one of the other sort of killer advantages of the lab is that there’s Code for America, and Code for America is this great organization that basically gets engineers to spend some time doing basically civic service, building software for some kind of governmental body or agency and. You know, we could debate sort of what percentage of that work ends up being sort of abandonwa or how effective that is, but I think the model is great, which is it sort of recognizes that in your career, you may have these gaps between things, you’re finished one thing, you’re not quite ready to start something new. And so Code for America, it’s sort of like an escape hatch that people use when they’re tired of their job or just need a break or want to do something different.

Similarly, because we run these sort of short term projects, We’ve had this amazing ability to bring in incredibly talented people who will be in eventually are snapped up by the top companies in our industry. Moments later, but we get to work with them for a few months on some really interesting project and not only do they influence our thinking, but we influence their thinking, and so we’ve really been able to scale our influence on the world. And to bring in a ton more insight and knowledge and sort of inspiration by working short term with people than we would have by working long term. And so I think that’s one of our like secret weapons for sure at the lab.

So, once we scope a project and kind of say, OK, well, we want to do this thing, we’re going to need these kinds of people, we go into recruiting mode, as we said, and we bring in a team. And then at that moment, our goal is like on kickoff day for a project, not always, but this is sort of the archetype. We want to have like a really clear sense of, you know, what’s the tech stack, what are we going to build, what are some vague milestones along the way. At this point, we’ve done prep, and you know, no plan survives contact with the enemy. Once the project gets started, things tend to change, but we want to come in with like a really clear direction and hit execution mode, and the team just focuses, they go head down. And they just engineer the crap out of it for like 2 or 3 months. We always pick sort of a fairly strict end date at the start of the project. You know, the fear of like impending deadlines helps everybody stay focused. And the other thing we do throughout the project is we do sort of a show and tell every other week where teams come and say like, hey, this is what we’ve been up to, you know, share with the rest of the lab, get feedback, but that sort of regular cadence of showing your work both helps people to stay focused and on task, but it also gives you kind of like regular mile. Stone so you don’t end up rabbit holing too bad. You know, it’s not like there’s like a law, but you have to show up on this day and have something new to demo. But if you’re going to turn up on Friday, on Monday, you know, this week we’re going to be showing our work, so you sort of make that extra little bit of effort to like make sure things work a little better or to polish something up a little bit or just to make sure you don’t break the build on Thursday kind of thing. So I really like that cadence. And then at the end of that time, it’s sort of like pencils down, project is done. And in the beginning, we really just sort of like at the end of the project, someone on the project, the project lead, would usually write a little internal Google doc talking about what we did, and in the really early days, everybody at the lab was on the same project, so everybody knew what the project did because you did it together. But at the end of the project, these days we’re spending more and more time writing, editing and publishing. And today that process generally takes 2 months, 2 to 3 months of work, and it’s certainly not full time work and often overlaps with pre-infusion for the next project, but it’s very much this kind of like process of not just publishing to share the knowledge with everybody else, but also mining your memories of that experience for insight and figuring out what happened. So that you can tell that story, and the process of writing, I think, really turns the work into knowledge in a really powerful way. And so as valuable as it is to publish this work to everybody outside of the lab, I think it’s at least as valuable to those of us inside the lab because it adds this extra level of rigor and understanding that we all benefit from. And then at the end of the publication, we put up these massive essays with lots of cool interactive elements and links on the sides. And then you dust your hands and take a quick breather and head back into pre-infusion and do another.

00:54:01 - Speaker 2: I think the you can switch essay style is something that developed over time. I think you wrote one of the first pieces just kind of as a medium post, you said, hey, we should publish about this work, and then we’ve gotten more comprehensive over time and really putting a lot of effort, like you said, basically now, at least in terms of wall clock time spending essentially as much time writing.

And distilling the results to understand what’s important, what are the real findings, and then put it into this very nicely comprehensible thing that others can read and learn from.

And certainly I’ve experienced that thing you talk about, which is taking the time to do that.

I feel like I extract more insights from the project through the writing process.

And then in terms of remembering it both in terms of institutional memory, but also in terms of just as an individual, I remember the things better or I can go reference my own work. It happens with some frequency of muse where we say a design question comes up or a technical question. I say, you know, actually we researched that exact thing at I can switch, let me go pull up the reference and.

Sometimes we have some internal docs, but those tend to be pretty rough and they haven’t been through the same clarifying process of writing for an external audience. But when we’ve published about it publicly, then it’s much easier to both find what you’re looking for, but then also be able to draw from that and our team can use that to make decisions about something we’re working on the product.

00:55:24 - Speaker 1: Oh, it’s great to hear that that stuff is benefiting you and that you’re able to take advantage of that.

I mean, obviously you’re very close to the lab, but it’s nice to know that our target audience is able to get that kind of yield.

I’m always asking how we can reinvent the lab, you know, I’ve just talked about what the lab is, but I’m always thinking to myself about what’s next. I’m always thinking about how can we change.

We’ve gotten much better at writing, we’ve developed this, we call it the academic voice, it’s not academic, academia has too much jargon and It has a really powerful culture around like these latexch publications. We try to make our writing more accessible than academic papers generally are in terms of tone and language, but we try to take a lot of inspiration from academic writing in terms of maintaining a sedate tone, avoiding superlatives, supporting any statements that we make.

We do actually publish a lot of our work in various academic venues these days. But first and foremost, we published for the web and for this sort of industry audience and for ourselves, and the academic version sort of follows on from that.

00:56:31 - Speaker 2: Yeah, the academic format, you might think of as being an illustration of how the lab sits on the innovation spectrum between startups and classic academia, where, yeah, academia, you know, you typically have, yeah, these PDFs, they’re published in particular venues, you’ve got peer review, you’ve got a lot of conventions about citations and things designed to enforce rigor. And then in the startup world, people do publish, but it’s usually some engineer tried a framework one time and they wrote a blog post about their insights on that.

00:57:02 - Speaker 1: Right, or the point of it is like this marketing exercise for a product.

00:57:05 - Speaker 2: Yeah, exactly, marketing for startups, yeah.

00:57:08 - Speaker 1: We try to be very sober in our presentation and to be really explicit about trade-offs and shortcomings and open questions, and I think that serves us well and it serves our readers well because we try to make sure that we portray the work we do with the inspiration and the context that we hope will make people excited about it. We also try to be really transparent about what we have and haven’t done and what remains to be done as well.

00:57:35 - Speaker 2: Well, as a place to end, you mentioned what’s next for the lab or how might it expand or reinvent itself, what exactly is on your mind there?

00:57:45 - Speaker 1: Well, to me, the biggest areas where we can improve are both growing the scope of the kind of work that we do in terms of both getting broader and deeper, I want to see us expand into other adjacent areas, hardware, and I’d like to see us expanding into ever more fields of research, but some of that comes down to just sort of like funding bandwidth and people around.

But I’m also interested in Expanding the ways that we work and opening up the lab to more people. We have this vision of a new kind of computer, and you know, it’s that vision of a different kind of way of using computing, not just programming, but computing that is more empowering, that makes us better versions of ourselves, that lets us do more things to do better work, to work better together, to think deeper.

And I just think that our current working model is doing fine, but there has to be so much more opportunity for us to take what we’re doing to a wider audience and to get more people involved.

And I think some of that is writing more about the why behind the work, but some of that is also finding new ways to get more people involved.

So, I’m always eager to experiment with new things, not just in terms of doing the research, but also uh getting New kinds of projects started and approved and so on, and yeah, we’re just gonna keep pushing the envelope and keep doing the work out there. So, if you hear this and you’re interested and you think that’s something I really want to do, hey, send an email to me, hello at inkinswitch.com, and who knows, maybe there’s an opportunity there.

00:59:26 - Speaker 2: Well, let’s wrap it there. Thanks everyone for listening. You can write us on Twitter if you’ve got feedback at @museapphq, and we’re on email as hello at museapp.com. We always appreciate it if you leave a review on Apple Podcasts. And Peter, I’m so glad that you’re continuing to push both the vision for better computing and the vision for how we can do innovation in this independent way and really looking forward to seeing what comes out of ThinkSwitch next.

00:59:55 - Speaker 1: Well, thanks so much for having me on, and it’s just a joy listening to the podcast and seeing Muses out there, building great software and testing these ideas in the market that we’re working on at the lab. So let’s keep that alchemy going.