Read more about Stephen Wolfram in “The Machine that Turns Ideas into Real Things.”
Stephen Wolfram started out on an academic career path, but eventually realized that founding a company would allow him to pursue his scientific work more efficiently. He’s served as a remote CEO of Wolfram Research for the last 28 years. In this episode, Stephen shares with host Matt Mullenweg — another remote CEO — his perspective on the value of geo-distribution, and the processes his partially-distributed company uses to make world-changing software.
The full episode transcript is below.
MATT MULLENWEG: Howdy, howdy. Welcome to the Distributed podcast. I’m your host, Matt Mullenweg.
I got the chance to catch up with today’s guest, Stephen Wolfram, because my company Automattic invited him to give a talk at our annual company meetup in Orlando, Florida. This is a magical occasion where all 950+ Automatticians (which is what we call ourselves) get together to meet up face-to-face. This gives us an opportunity to hang out, break bread, and collaborate over the course of a few days. We also invite a number of speakers, smart people like Stephen.
Stephen’s been leading Wolfram Research for 32 years, which is a really long time for a tech CEO. The company has pioneered a lot of different technologies in computation and education. Wolfram Research has about 850 employees, many of whom are scattered across 29 countries, so it’s pretty close in size to Automattic.
But Stephen’s been doing the remote CEO thing for way longer than I have — about 28 years! So naturally I wanted to pick his brain, which is an amazing brain, on why he chose to go the partially-distributed route, and learn about how he’s led his company remotely for longer than just about anybody.
Alright. Let’s get started.
MATT: Welcome. This is Distributed with Matt Mullenweg, and today we have Stephen Wolfram, who is the Founder and CEO of Wolfram Research, known for things such as Mathematica, an amazing tome called A New Kind of Science — which I guess you worked on for about ten and a half years — created Wolfram Alpha, which is one of the intelligence engines behind Alexa and Siri…
STEPHEN WOLFRAM: Siri and other things, yeah.
MATT: Yes, and of many other things. And just an incredible amount. I encourage you to google him. Go down the rabbit hole of all the amazing stuff and people he’s worked with and everything. Thank you so much for joining today.
STEPHEN: Nice to be here.
MATT: Part of the reason, in addition to all those fun things, that I wanted you to be here is your company is a similar size to Automattic, around 800-900, and is geo-distributed as well.
MATT: You have been a remote CEO since 1991.
STEPHEN: That is correct.
MATT: And your company goes back to..?
STEPHEN: Late ’86.
MATT: Possibly one of the older geo-distributed companies I’m familiar with.
STEPHEN: I think so. I mean, I know —
MATT: It’s got to be one of the longest, if it’s not—
STEPHEN: Yes, you’re right.
MATT: Maybe there’s something out there but that’s got to be one of the longest.
MATT: That’s pretty cool. What caused you to be geo-distributed at the beginning?
STEPHEN: So OK, the first thing was, I started the company. This was the second company I started. The first company I started, I started when I was 21 years old and that was in Los Angeles, and that company was a pretty traditional company. It was venture-capital funded, I brought in a CEO, I didn’t CEO it myself, I was the technical person —
MATT: Did they call it adult supervision back then?
STEPHEN: No, no they didn’t. I mean that company went through various mergers and things and eventually went public some time in the 1990s in a very undistinguished IPO. So anyway, so my first company was not distributed at all, it sat in this building near the Los Angeles International Airport. [laughter]
I was involved in basic science and I had developed this area of science that I guess now gets called complexity theory, and so I tried to figure out what university would give me the best deal to start a research center in this kind of science. So I went around to lots of universities, and the one that won that was the University of Illinois in Champaign, Illinois. But I said, “OK, I’m going to start a company that builds the tools that I want to have for myself and that I perfectly well know are going to be useful to lots of other people in the world.”
So we started the company in Champaign, Illinois, which is not probably most people’s first choice for where to start a tech company, even back in 1986, but it was great. We got terrific people from the area — it’s a good university, producing a lot of interesting graduates. We were the only game in town, so to speak, and I think we’re still the largest tech operation there. So we started in a slightly outlying place. Then [we] got off to a very quick start, up to a couple hundred employees, maybe 150 or something. I was injecting ideas into this thing at a very high rate.
MATT: So when you inject ideas, you’re coming and saying, “Hey we should do this thing.”
STEPHEN: Yes. And getting more and more frustrated that they weren’t getting done. I have to say that just a few years ago we finally finished my 1991 to-do list.
MATT: Ha! [laughter]
STEPHEN: So it’s done, that one is — it finally, finally happened.
MATT: That one’s… Yeah, the thing that got you to leave is now all the way complete.
STEPHEN: Yes, right. My original plan had been to build these tools that will, among other things, allow me to do the basic science that I wanted to do. I said, “Let me step back a bit and let me go off and spend some chunk of my time doing basic science, and let the company grow up and we’ll see what happens next. Maybe there will be a coup and somebody else will say ‘I can run this better than you can’.”
MATT: So everyone was in the office except for you?
STEPHEN: Not quite. What happened was — I’m trying to remember how many people weren’t in the office at that point, but I was definitely the main go-offsite person.
MATT: What were the tools at that time? Was that because you had been in a university setting and were familiar with email or internet?
STEPHEN: Yes, well I started using email in 1976. So that was —
MATT: That was pretty familiar to you.
STEPHEN: Yes, right. It was modems dialing up. My email would arrive in bunches every 15 minutes. And I was on the phone a bunch. I had hired a chief operating officer.
MATT: I was going to ask what the exec team looked like. You had a COO who could be the on-the-ground lead?
STEPHEN: Right, who was with us for six or seven years. And he did a good job of maturing the company from being high-growth to something which had good systems in place and could roll forward.
MATT: And this second iteration of the company, did you also raise venture capital, or did you decide to take a different approach to have the longevity?
STEPHEN: No, no, not at all. Version two — I was the CEO from the beginning and forever type [of] thing, no outside money ever. It’s been a shame for me that I’ve never really had a quote “business partner.” I think of myself as pretty average at business kinds of things but I’m not totally incompetent.
Personally I view it as an optimization. You can do things that are very commercial, but a little bit intellectually boring. And it tends to be the case that you’re doing a lot of rinse-and-repeat stuff if you want to grow purely commercially, so to speak, or you can do things that are wonderful intellectually, but the world doesn’t happen to value them and you can’t make commercial sense that way. And I’ve tried to navigate something in between those two where I’m really intellectually interested and where it’s commercially successful enough to sustain the process for a long time.
MATT: I love finding that intersection.
STEPHEN: Oh yeah.
MATT: And it feels like now that the world has caught up a little bit to voice assistance and the natural language processing stuff you’ve been working on forever, that you found — is that the main business model for Wolfram Research now?
STEPHEN: No, no.
MATT: Oh, I would’ve thought that licensing drove it because there must be billions of users of these voice assistants.
STEPHEN: There are. Yes. And it’s a good source of revenue. At this point it’s fairly diverse. There’s a big chunk of licensing software but a lot of that — there’s, for example, the academic sector, there’s site licenses to universities where basically the goal is to make the software free for people to use at universities, and we’re complete now in the U.S. in the sense that all the major universities have site licenses. So if you’re any kind of major university in the U.S., you will be able to use our software for free. And there’s a lot of commercial use of software where we’re basically selling pre-packaged software. It’s been nice that it has been gradually quite diversified between different kinds of channels.
Back to the whole no venture capital do. It’s been great. I don’t have a boss. I recommend it. [laughter] And also I can do things —
MATT: I feel like you always have a boss. It might be a customer or all of the customers in the world.
STEPHEN: Yeah, at some level —
MATT: It could be the employees for whom you’ll respond to if they reach out.
STEPHEN: That is correct. Yes, that is really what happens. But in the first approximation you think, “Oh, I can do anything,” and then the second approximation is, “Oh, we have all these customers. Oh, we have all these employees.” I take those responsibilities really seriously. But I’ve been doing them so long that they seem natural, so to speak. It’s a thing where there is a certain kind of intellectual freedom, that I at least believe that I have by virtue of the fact that I’m just responsible to myself.
MATT: I will say that I think, as a result of so much capital flooding into the market, that there now are very long-term investors who, if you choose them correctly, can be aligned on longer time spans than the standard five to seven [year] fund life of some of the historical venture capital.
MATT: And founders now are able to retain a lot of control through voting mechanisms. They might sell economic interest but [you’re] no longer selling control interest.
STEPHEN: I always wonder how that’s going to come out in the end. I thought about taking my company public back in the early 1990s. We were on a great trajectory and etcetera, etcetera, etcetera. But then I was like, “Hmm, I don’t know.” Actually the final thing was my employees saying, “Why are we doing this? This doesn’t seem like a way to have a good time,” type thing.
MATT: That’s great that you had the kind of culture where people felt comfortable challenging that as well.
STEPHEN: Yes, yes. I think it’s fair to say that we have a culture — We have a lot of very bright people who have a lot of opinions. And I like to believe that we have a company where what we are mostly interested in is finding people who can be productive in our environment, and they have very different personalities. Some of them, I think I at least, have the view that they pretty much couldn’t work anywhere else because they’re —
MATT: [laughs] Unemployable.
STEPHEN: They have very unusual personality traits and poor internal politics skills and so on.
MATT: What makes them successful at Wolfram?
STEPHEN: That they do good work. That’s the thing that I —
MATT: Part of good work is working with colleagues and the teamwork aspect.
STEPHEN: Yes, yes. But I think that the role of management is given who these people are, how do you fit them together with what we are trying to do in such a way that you take advantage of their good traits.
MATT: So to end the business thread, do you grow the employee base of the company along with the revenue or with your scope of what you’re trying to accomplish?
STEPHEN: Good question. We have been lucky enough, touch wood, that we have been profitable every year for 31 years now. That is achieved by a very simple process, which is, spend less than you make. Thankfully we have never had to have a big “Oh my gosh, the revenue is less than we expected, let’s let people—” No. We’ve fortunately never had that. Our general principle is we’ll pay Champaign, Illinois rates for a certain set of people. Other people, it will be based on where they are, but we’re never going to go above that.
So that means when people wind up in San Francisco or New York, we’ll often lose them because we’re not going to pay the rates that people expect and need in those places. On the other hand, if they find some obscure place somewhere, they’re going to be doing really well because —
MATT: I recommend Houston, where I’m based. Not too obscure and a very reasonable place to leave.
STEPHEN: Yes, we have quite a number of people in Texas. I think we may even have a couple people in Houston. I’d have to look at the map.
MATT: I read some really interesting research hypothesis around why geographic mobility has gone down in the United States. You would think people would move more for economic opportunity but it has actually gone down. And the conclusion of this was the blocker was two-income households.
MATT: Where if one person were moving, and the sole breadwinner, you only have to find one job. And now you have to find two jobs in the new city, which is at least twice as hard.
STEPHEN: That’s interesting, yes.
MATT: But when one of the partners has no geographic attachments, you end up following. I just talked to a colleague who ended up in Copenhagen and he’s from — I forget where — some place very far from Copenhagen. He said, “Yeah, my wife got a research job and now we’re in Copenhagen for a few years.”
STEPHEN: Yes, we find that all the time. There were some obscure ones. Like there’s one guy in our user interface team who’s been at the company a long time, but he said, “Well I’m going to go remote and I’m going to this island off the coast of Nicaragua because my wife is a primatologist and she is studying… What were they? Some kind of monkey-like creatures on this island. And he got a microwave link set up, and there he was, being perfectly productive.
STEPHEN: To me the focus is on — Can you be productive? Are you doing good stuff? And then where you live is your independent business, so to speak. There is a certain amount of complexity in scheduling things and meetings and so on when people are in all different time zones, but somehow that doesn’t end up being that horrifying.
MATT: Is there some sort of predetermined taxonomy or is it more of a free-form?
STEPHEN: It’s free-form. One of the things that is fun about my company is I think of us as a microcosm of what goes on, because we have people from the history PhDs who are actually doing history stuff for us, to the people who are doing — whether it’s graphic design or whether it’s some software engineering, server infrastructure thing.
One of the things we’ve done — I don’t know how you’ve done this — but the company is pretty vertically integrated in the sense that we don’t outsource anything really. It’s all in-house graphic design, legal, this, that, and the other. Partly because we’ve got enough stuff going on that there’s not really a group that’s going to have nothing to do. And it also really, really helps to have, let’s say, designers who really know and understand our technology, and they’ve met customers of ours and they know the story. And plus some of the people in our design group become really good Wolfram Language programmers.
MATT: Cool. Actually some of our best developers and architects started off as designers.
STEPHEN: Yes, I think it’s a really good field to start in. It’s funny, because I track where people come from and so on, and physics, for example, is good. For the techies, it’s a really good export field, so to speak, and I think graphic design is another one of these good export fields, where you learn a certain discipline of thinking to do good graphic design.
For a while it seemed like UX was — we were pretty early in the hiring of official UX people and it seemed like that might be taking off more in that direction, but I think it’s a little bit merged with the graphic design. And I feel like sometimes the designers — they always have something to show for what they’ve done, in a sense, whereas UX is a little bit less clear. Somebody made this flow diagram or something but that turned into something different from what finally came out.
MATT: The visual communication helps a ton, right, for getting people on the same page?
MATT: It’s one thing we’re trying to do more, is just draw something, even if you’re not an artist or a designer or anything, so you really are on the same page, literally, with what you’re trying to communicate.
STEPHEN: Yes. I don’t know in your geo-distributed setup, but we never use video conferencing, I mean really never, to the point where —
MATT: It’s always screen sharing and voice?
STEPHEN: It’s just screen sharing and voice.
MATT: It’s mostly video actually, with us. And what was interesting to me observing it from the outside, is before Zoom, we had very few meetings because I felt like the process was so frustrating to everyone, we just didn’t do it. And Zoom spread like wildfire in our organization because there was just something to it that made it — It worked just a bit better, it got over that threshold, whatever that uncanny valley was that kept us from using it before. And now I think we have too many meetings, perhaps. But video is a thing, and people think about their background and try to have good lighting. Good audio is really important to me personally.
STEPHEN: Yes. Good audio is very important to me. I am always complaining about that.
MATT: I will send you a link to the Sennheiser headset we like. It’s USB and it’s only about $34. It’s become very standard.
STEPHEN: OK, I’d like that. Yes, we’ve spent lots of effort on headsets for people and complaining about people not having the right audio setup and so on. And I am glad to know that I’m not the only person who really cares about that.
MATT: It makes a huge difference. I’ll use this opportunity to tell people to check out your blog post. Was it “Seeking the Productive Life?”
MATT: On your WordPress-powered blog.
MATT: Which is a very comprehensive view of your entire personal operating system and productivity, which I found — well, it’s impossible to summarize. And then two, we used this to point to [the fact] that you have started livestreaming many of your design meetings.
MATT: And so I could probably google “Wolfram livestream” or something?
STEPHEN: Probably, yes, right.
MATT: And so these actual internal meetings are now broadcasting publicly, which is pretty unique and fascinating.
STEPHEN: That is correct. This livestream meetings thing, I started doing it about a year and a half ago or something because I thought, these meetings are fascinating. It’s a shame for these things to just go off into the ether and nobody — so I thought, let’s do this.
MATT: It’s mostly screens and then audio?
STEPHEN: Screens and audio, no people. We’ll get people who are both expert users of our product and often world experts in the particular thing that we are talking about. And they show up and they contribute useful things.
MATT: It’s interesting. You have a lot about your company which is very much like how open source works, but a lot of what you do is not open source, correct?
STEPHEN: That is correct.
MATT: There is a post I read about this while doing some research, 12 reasons — I think one of your colleagues wrote it — that Wolfram Language wasn’t open source. And one of them was, things like language design aren’t benefited from being more open, is how I interpreted one of them.
STEPHEN: Yes, probably, yes. I think — look, it’s a complicated thing because my goal is to be able to produce this intellectually valuable, long-term thing. And the question is, given the world as it is, what is the best way to do that. And we have built, back 33 years ago, when I started the company — it was like, OK, we’re going to sell pre-packaged software. And I didn’t know whether that industry would survive. I mean back at that time —
MATT: It wasn’t clear, right? Piracy was huge.
STEPHEN: Yes, piracy was a big problem. Another big problem was, what was the price point going to be for software? Because at that time there was Borland versus Microsoft. Borland was at $100, Microsoft was at $500 for typical software. And it’s like, who’s going to win in that space?
MATT: How much is Mathematica?
STEPHEN: Well it’s very complicated. [laughter] It ranges from —
MATT: How much was it at the beginning?
STEPHEN: It was $495. But no, I think what we’ve done — it’s a thing I’ve thought about quite a bit — what is the right way to slice — We are trying to do a long-term, intellectually valuable thing, having something — maintaining that kind of coherence over a long period of time, I don’t know what the best way to do it is. I think we have found a pretty good scheme. Wolfram Alpha has been free from the beginning to use.
We recently introduced this thing, a free Wolfram engine for developers, which people seem to like. And the deal is, using it for development, it’s free, you can do whatever you want with it. But if your work product is what it is producing, if it is actually in production, that’s when you have to start paying something for it.
MATT: I think it’s interesting, you’re adopting some of the elements of what helps open source become ubiquitous, and some of the open things that, by the way, open source didn’t invent — they came from academia and many things before, in terms of the collaboration — and I think it’s very cool.
STEPHEN: An important thing for us is the alignment of where we make money versus who is actually getting value, who our real customers are. And I think some of the cracks that are happening in some of the areas of the technology industry — I think come from a lack of alignment between — who are the actual customers? Well the actual customers are advertisers, they are not the people who are — whatever else. And I think we have been both lucky in that our customer base is wonderful people we really like.
The thing that tended to happen though is these projects that — it starts open source and then there is some kind of bait-and-switch somewhere. And it just drives me crazy. We keep on —
MATT: I totally agree. And that was the open-source model for a long time. We’re going to be open source but then hey, that open-source license is really scary, don’t use it, buy our proprietary license, or things like that.
MATT: Or we’re going to put all the good, new stuff in this proprietary version of the license.
STEPHEN: Right, right.
MATT: That’s why I want to talk more about open source at some point because, if you look at — They’re not always the best products, Wikipedia was not the best, WordPress in some ways you could say is not the best. Chromium engine — I don’t know if you saw the engine behind Chrome was just actually adopted by Internet Explorer.
MATT: So we’re getting a ubiquity and a de-facto standardization of a code base that becomes so useful.
MATT: You could never force Microsoft and Google to say, “We’re going to do the same thing.” But they have, through their own personal utility maximization, have chosen to collaborate in this one area, which is really beautiful.
STEPHEN: That’s interesting. There are these different things, like, let people see the source, right? I don’t care if people see the source, that’s not the issue. I mean that’s not the point. Although we did have an interesting experience.
So one thing happened because our products get used for lots of fancy, tech-y things. At the very beginning we would have the following sort of thing that would happen. Some mathematician would come up at a trade show and say, “How can I trust your stuff, I can’t see inside how everything works?” And I’d say, “Well how can we trust these papers you’re writing?” It’s kind of like, their mechanism for validity is peer review. Our mechanism for validity is software quality assurance, which is a lot better.
MATT: I take it you have some opinions on peer review?
STEPHEN: Oh, it’s terrible. I think peer review is — I always thought it was broken, even when I was in the business, so to speak. My point of view was, if I have an original paper or idea, it’s going to be really tough to get it through a peer review process. 1986 was the last time I published a paper in an academic journal. I decided they were a bad idea and I wasn’t going to do it.
MATT: And now you have a blog, which is even better.
STEPHEN: Yes, the blog is — some of the things I write on the blog are quite academic. And it’s really interesting, I remember I was writing a piece about Ada Lovelace, actually, at the time of her 200th birthday celebration. And I happened to, at the last minute, I decided to go to England and go to some fancy celebration that was happening about it.
And I was visiting some museum that has a bunch of her papers and so on, which I’d already got copies of most of. But the curator there was saying, who was used to academic stuff, was saying, “When do you think you’ll be publishing this?”, thinking the answer would be a year from now or something. And it’s like, well no, I’m going to publish it tomorrow morning before I get on a plane. Because you can write anything there.
MATT: And you do.
STEPHEN: And I write some stuff that’s quite technical and some stuff that is quite product-oriented.
MATT: Back to that live-streaming of meetings. I think it’s great that you say the meetings are kind of like they were before you live-streamed. They’re not too performative or anything.
MATT: I’ll add something else to that, which is you said you’re a CEO for life, more or less, there is a lot of authority vested in you and this company, but it also seems simultaneously that people are comfortable with challenging you and presumably the other executives. What do you feel like you’ve done to foster that culture where there is a comfort, it sounds like, at many levels to challenge you very directly?
STEPHEN: Yes, it’s an interesting question. I suppose that I’m actually prepared to listen at some point. Some of these things get quite heated. It’s hardly as if — I don’t necessarily — I kind of know myself well enough to know by the time I’m getting heated, I don’t know what I’m talking about. That is the typical — and I will even say that because if I can explain myself, I’m just going to explain myself. And it’s always frustrating when people don’t speak up and don’t say —
MATT: Do you do anything to encourage people who you might feel are hanging back a little in the meeting?
STEPHEN: Yes. I am enough of a people person. The fact is that having been running a company for 33 years, if I wasn’t a people person it would have driven me nuts, OK?
MATT: Or them away and there wouldn’t be a company. [laughs]
STEPHEN: Yes, yeah, right. I like people, I find people interesting. Sometimes I would say, more cynically I would say, after I’ve been managing people for 40-something years, and you might have thought you’ve seen every pathology that could possibly happen. But no, there is always a new one. At this point I just find it faintly amusing that — “Oh my gosh, another bizarre thing I’ve never seen before.”
And I see a large part of my role being that I’ve got all these talented people, I’ve got all of these projects we’d like to do, how do we match talented people with projects we want to do?
MATT: Do people feel like they’re moving around or is there some stability once they get onto a project?
STEPHEN: There is as much stability as they want, basically. There are people who have worked at the company for 20 years, and have never done management even though I can tell they have a good personality to do management. And finally we persuade them, you should do this, and they say, “Oh this is actually quite interesting.”
MATT: Is that more in your head or do you do any sort of testing or other objective ways to determine people’s strength or potential abilities?
STEPHEN: No, nothing, other than the Who Knows What database of just factual information.
MATT: And does this all come from you or is this also the rest of your executive team?
STEPHEN: No, there is a bunch of people who have gotten experienced at doing this, I would say. When I look at org charts of companies I always —
MATT: I love org charts.
STEPHEN: Yes, right. That was probably a decade ago, I was like “Oh, our org chart is such a mess.” For a long time, our org chart was classified because there were people, it was like, particularly one person who was with the company for a long time, and has now spun off doing related things, but who was an executive at the company and was always like, “We shouldn’t really tell people what all the weird reporting arrangements are.” I wasn’t a big believer in this but it wasn’t…
But then, look at our org chart, and it seems messy in many ways. And then I was working — actually that was with Microsoft — and somebody who is actually now an even much more promoted executive at Microsoft, said to me, “We were trying to understand how it worked.” He said, “There’s this company on the outside that has reverse-engineered the Microsoft org chart, we just buy their stuff to understand what’s going on.”
MATT: That’s funny.
STEPHEN: Yes, right. So I get this thing and it’s a big foldout thing and it’s fascinating because it’s a huge mess. I mean, I don’t remember whether it’s still a mess but at that time it was a huge mess and you could see lots of historical stuff. And so after that I felt so much better about our org chart. Although, our org chart is now much cleaner, I would say.
MATT: It’s the map and the territory, right? Our org chart is currently clean but it’s not comprehensive, so there are some things that aren’t represented on the org chart, which sometimes a person is in two different places, that might be represented, but then they have informal authority in an area that we don’t try to demarcate necessarily.
MATT: We have had a lot of success, actually, removing everything compensation-related from the management structure. So there’s only a few, only the centralized HR deals with all that and everyone else just focuses on getting the job done and peer feedback and everything.
STEPHEN: One thing at our company, which was a long-time joke, was the claim there are more projects than people at the company.
MATT: Yes, we have the same thing, wow.
STEPHEN: It has been a running joke — can we actually inventory all the projects? Well now we’ve, thankfully — I’ve got this team together to just go and do that and it has been a good process actually.
MATT: Are you familiar with Marie Kondo?
MATT: You know her thing is you put all the stuff that is the same, like all the clothing, in one pile. It sounds like you’re doing that from an organizational point of view.
STEPHEN: Yes, yeah…
MATT: You’re trying to get all that stuff together so you can sort it out.
STEPHEN: Right. I’d certainly like to apply software design methodologies to the way that we’re doing things. Look, the thing that we have done that I think has worked pretty well is we have pretty good project management, tradition, and infrastructure. And part of that is, I’ll come up with some crazy idea, actualizing that has to be flowed through the organization.
Now to be fair, I think I am probably — one of my less-bad traits is that I’ll come up with these ideas and then people will say, “How on earth are we going to actually do that?” And I will actually know, that is, if they don’t. It will be like, because I know enough about engineering and the whole technology stack, that I’ll be like, “Look, we can — some thing or other thing has to be authenticated here, and do this and that and the other, and I can actually dive down to a pretty high level of detail about how it should be done.
And I think part of my role is to explain why things aren’t impossible. And I see increasingly with a lot of projects we have done, the first response is, “That’s just impossible.” I’ll have some idea and it’s — and actually I am happy when people say that. When I’m not happy is when people say, “Oh sure, we’ll do it,” when I plainly know there is no way they can do it, it’s too hard. And so then I’m trying to figure out, “OK, so let’s see whether we can figure out how to do it.”
I think the other thing that I find I do a lot of now is, we have all these projects connecting — I’ll be in some meeting and somebody will be talking about, “Oh, I’m doing this, this, and this,” and I’ll say, “Oh, you should talk to so-and-so because they’re doing something similar.” And we were doing stuff to do with external storage systems. We have this group that’s been doing database integration and a blockchain group, and they all have things to do with this external storage story, and that becomes part of the role — explaining, just knowing enough about what’s going on around the company.
MATT: And how in this — I’m curious from a software point of view — how does maintenance work in this structure?
STEPHEN: Well that’s a good question. First of all, a lot of things we do in the language are quite modular. There’s a function, it fits into the language, but it doesn’t have a lot of nasty, dangling interdependency.
STEPHEN: We’ve gotten a lot of experience and good systems for doing software quality assurance, and I think we are pretty good at doing that. Some of the QA we have to do is kind of hair-raising, like a lot of real-time things, which are always difficult, but even for Wolfram Alpha, there is — the natural language is a fundamentally non-modular thing.
STEPHEN: So the classic example is if you type in “50 cents,” you have money, if you type in “50 Cent,” you get the rapper as the default thing. What happens when there is another rapper who is called “30 Cent” or something? That’s a very non-modular thing, and you have to be able to deal with all those kinds of situations. But I think we have done pretty well at that.
What we tend to do are these big subsystems. They get swept through every some number of years, and then a lot of stuff gets — a lot of things that we’ll fix —
MATT: Who decides to sweep through the system? Is it because a number of bugs accumulate?
STEPHEN: No, it’s not as organized as that. That’s a good question. It ends up being the people who run those engineering teams. I have to say that I am pleased with their level of responsibility in the sense that we are going to rewrite a bunch of the UI stuff because it’s out of date.
MATT: On the converse side, sometimes I believe it’s easier to write code than read code.
MATT: And so you get sometimes new engineers on an area who want to rewrite it partially to understand it.
STEPHEN: That’s true.
MATT: So how do you keep the over-read factor?
STEPHEN: Yes, right. Well I push back on that and say, “Why are we doing this?” In fact, I was just doing that yesterday, a particular thing. It’s like, we can do this wonderful thing — why are we doing this? This is not an important feature, this is just not the time to do it.
MATT: Are there programs around learning and developments or ongoing training?
STEPHEN: The one constant is a lot of our development, most of our development is in our own language. But at this point, most of the people we hire know it.
MATT: Before they get there?
STEPHEN: Yes, it’s kind of scary that the system is older than many of the employees. And they’re like, “Oh yeah, I started learning this when I was ten years old” type of thing. So that’s —
MATT: I’ve started to have conversations like that in WordPress and it is very humbling.
STEPHEN: Yes, yes, right. For some areas I let different managers have different rules about how they set that up, but some people insist the person should come to an actual office for some number of months when they start to get —
MATT: This is onboarding.
STEPHEN: Yes, right. And that seems to work well in some cases. And I don’t know whether it’s really necessary, I haven’t really — we haven’t done a —
MATT: Would a prospective employee know what kind of team they’re joining?
MATT: Are you hiring for specific teams or are they generally open roles and you try to place people based on their —
STEPHEN: We are placing people. But by the time they are actually joining, it’s pretty clear what they’re going to be doing, for sure.
MATT: You’ll say, “Hey, you’ll need to come to Champaign for two months,” or something?
STEPHEN: Yes, right. sometimes we’re hiring people where it’s like, “When I finish my PhD I’ll come,” and they’ll start in some number of months and we’ll say, “Well we’ll have something interesting for you to do.”
MATT: By then.
STEPHEN: Yes right, yes.
MATT: Let’s say I’m a coder in Houston and I apply to Wolfram. What is my process like on the way to being hired, assuming I will get hired?
STEPHEN: That’s a good question, actually, I don’t — I have to say, I regret the fact that I don’t get to see most of the frontline applications, only some of the more senior ones. I used to look at a bunch of these things. Cover letters are, to me, a critical part of the story. I haven’t been reading them recently because I have —
MATT: That’s why I don’t like applicant tracking systems, or form applications. I want to see how they email, what client they used, what fonts there are, how they formatted it. I want to see everything about the application as free-form as possible.
STEPHEN: Yes, right. What I find is that when I used to read these things a lot, the ones that were just, “Oh my god, I can’t believe this.” “I just graduated in computer science from this place and therefore I am qualified for whatever.” It’s like, no. And then there are other ones which say —
MATT: There are some cover letters though that are such a good story, and you can get a sense for someone’s clarity of writing.
STEPHEN: Yes, yeah, right. Well, also some of the ones that are more fun are things like, “I’ve been reporting bugs in your software for six or seven years now, I want to come and help fix them,” type of thing. That’s kind of cool.
MATT: What are the characteristics that you feel particularly makes someone successful at Wolfram Research?
STEPHEN: A certain independence of mind, that they can think about — that they keep the thinking apparatus engaged at all times. Oh, yes, a very important one — no bullshit. That’s a —
MATT: How do you detect that though?
STEPHEN: It’s really easy, I think. You just listen in the interview how people will say, “Oh, yes I know all about that process.” Well, I know a tiny bit. And sometimes —
MATT: Some people can be more humble in their —
MATT: So interviews are very high-pressure. We do a lot of our interviewing on text, on a chat.
STEPHEN: Oh that’s interesting.
MATT: And that’s partially because we found that there is a class of people for whom a real-time conversation like this can be incredibly intimidating. There’s an asymmetry as well. They’re hoping for a new job and they just don’t perform well in that situation. That looks like almost nothing else we do in most people’s daily work. The interview, it’s very artificial. But the text chat looks a lot like what an average developer might do most of the day.
STEPHEN: That’s interesting.
MATT: So it can allow people that kind of space to — it’s like a take-home test a little bit. They can take five minutes to respond to something, they can Google something, if they get really freaked out by the question, they can take a walk around the table.
STEPHEN: I have to say, in all the interviewing that I do, people sometimes come into it tense but I don’t think — I think if you’re doing interviewing right, the people aren’t that tense.
MATT: It’s probably an inherent power dynamic that is impossible for you, in particular, to remove.
STEPHEN: Maybe, maybe.
MATT: Well text removes that even more.
STEPHEN: Yes. I would say that at some level, if the people are sufficiently intimidated, that’s probably not a great indicator, because the fact is, in our company, there are a bunch of strong opinions and people will express themselves. And if somebody is like, “Oh no, I can’t deal with that,” that probably isn’t a great indicator.
I have to say, it’s a different thing with more senior folk, like business people and things like that. That’s a horrifying world of interviewing, for me at least. Because people who have, who are used to a very polished presentation, it’s just so difficult to find out who is the actual person here.
MATT: How do you balance that pattern matching? Not leading the confirmation bias or where the company might need to evolve, which might be different from where it has been in the past?
STEPHEN: I’m a guy who likes new stuff, so that helps me be more change-oriented. I have to say that by the time it’s working smoothly, I’m not the person to be dealing with that. I have for myself and my personal life, I have endless systems that I have developed and they work, but also sometimes they stop working and then I’m like, “Oh, this doesn’t work anymore, I’m going to change it.”
For example, it could be the case that the things that I like and that resonate with me and that are the directions that I think of going are just not the ones the world happens to be keen on. A couple years ago I was like, “Let’s look at VR and AR.” And it’s like, well, I have a bunch of long-term people at the company, and they said, “You told us that in the early 1990s.”
MATT: That’s funny.
STEPHEN: What a bust that was at that time. You’re going to have to do better this time to convince us. And I have to say, I couldn’t. I couldn’t really convince them. Or IoT is another area where we’ve done a bunch of stuff, a bunch of interesting things, but it hasn’t quite taken off the way that I think some people thought it would.
This question of, when you’re the CEO, when do you fire yourself, type thing, and do you change, or is the thing going in a direction that you just don’t want to change it?
MATT: And your name is in the company, so how does that happen?
STEPHEN: Yes, right.
MATT: Is it just “research” at that point?
STEPHEN: Right. That’s an interesting issue. I was thinking —
MATT: We joke Automattic would be just “Auto-ic.”
STEPHEN: Right, right. I’ve been thinking of this. I was thinking at one time if I had a club of people who had named their companies after themselves. And you’ll be a partial member, right?
MATT: Thank you. Well we luckily have, I think, 17 other Matts at the company, so it could keep going without me.
STEPHEN: OK, OK. But I think —
MATT: Yes, I do joke — the egotistical founder, we always slip our name in. I want to ask two things as we wrap up. You have a lot of trust, it sounds like, which just allows this very candid communication, candor between folks. Do you have something like the meetups that we’re doing here?
STEPHEN: Yes, it’s not quite as formal. Once a year we have a technology conference. Users come, they have a good time, they get to meet a bunch of our employees. The employees come, they have a good time, they get to meet a bunch of our users.
MATT: Cool. Well after all this time, would you consider management or running a company computationally reducible?
STEPHEN: [laughs] That is a good question. This is one of the embarrassing things about people who like to think they’ve invented paradigms for thinking about stuff. The question is, “Can you live your own paradigm?” And it’s often the case in these things where I can see something developing at the company — I’ve been doing this a long time so I know how this story ends.
From the point of view of science and technology, there are things we’ve invented, I’ve invented. I could be very worried. Is the world ever going to pay attention to this, is the world ever going to care? Maybe I’m just too arrogant, but the fact is, I just know it’s going to go that way. This thing about computational language and the importance of having a rich language which can express computational thoughts, this is inexorable. This is going to be really important. I don’t know how long it will take the world to realize how important that is.
MATT: It’s not if, it’s when?
STEPHEN: Yes, yes.
MATT: Well that’s a good place to wrap up. Stephen, thank you so much. This has been an endlessly fascinating discussion and I’m looking forward to continuing it over dinner.
STEPHEN: Sounds good.
MATT: This has been Distributed with Matt Mullenweg and we’ll see you next time.
MATT: That was Stephen Wolfram. You can find him on Twitter at @stephen_wolfram. That’s Stephen with a PH, underscore, Wolfram — W-O-L-F-R-A-M. He also has an awesome blog, if you just google his name and “blog.” It’s powered by WordPress of course, and it’s really good.
Stephen says that he set out to build a company that would, quote, “allow him to build the tools to do what he wanted to do.” And by that he means working on projects that he calls “intellectually valuable.” It’s easy for CEOs to come up with lofty mission statements, but very few actually follow through. I’m inspired by Stephen’s steadfast commitment to intellectually valuable work and the organizational freedom that has allowed him to pursue it across three decades — and hopefully for many more to come.
Next time on the Distributed podcast, we’ll be talking to Taso Du Val, the CEO of Toptal, a freelancer network where employers can access a highly curated pool of remote contractors. Toptal itself is fully distributed, and they’re trying to convince blue-chip companies to integrate remote talent onto their teams, so I’m excited to talk about how this could help usher in a distributed future.
Thanks for joining us, and see you next time.