Peter Wang: These people are artists and craftsmen building incredibly thoughtful work, getting paid nothing. Open source has got to actually figure out a way to make money sustainably and honestly, to put regenerative economics back into the ecosystem. Eric Anderson: This is Contributor, a podcast telling the stories behind the best open source projects and the communities that make them. I'm Eric Anderson. Eric Anderson: We're very lucky to have Peter Wang here with us today. I could go on and on about his bio and we'll get into all of it, but we'll center some of our discussion around the Conda open source project of which Peter was one of the creators. Peter, thanks for coming on the show. Peter Wang: Thanks for having me, Eric. Excited to be here Eric Anderson: As is customary, tell us what Conda is, although we won't limit ourselves just to Conda today. Peter Wang: Yeah, so Anaconda is a software distribution that we built many years ago to make the open source data science tools around Python much easier for people to install and update. And Conda is the package manager tool inside Anaconda. And it's a package management format that you think about sort of like if you use RPM on Linux, right? It's something like that. And the wonderful thing about Conda is that it is cross-platform so you can use it on Linux, Mac, or Windows. And it actually supports more than just Python, it supports R, it supports C++, any number of different languages. Peter Wang: So we built the system, it's one of these stories where we built it as sort of an accidental side project in our quest to make Python more powerful and better and easier to use for data analysis, we sort of had to make a way so people could actually get all the tools. And so we built this as sort of like, Hey, let's do this thing to make it easier for people to install everything, and then it's the thing that became really, really popular. Now a lot of the tools we built also became very popular to be very clear, but certainly Anaconda and Conda are now a staple in the data science toolkit for many millions of people around the world. Eric Anderson: I got my start in open source in data science, and I got my start in data science on Anaconda and Conda. I think everybody has the trouble setting up their Python environment the first time, that's like the getting started, that's how they screen people I think for the profession is like... And it's not about intelligence, it's patients. If you can't get through this and feel comfortable and Conda solves a lot of those problems for me, or Anaconda did. Now you already kind of alluded to why you needed this, I should mention to our listeners we had Travis Oliphant on earlier and he kind of teased at the need for package managing in Python, that it was hard to kind of distribute libraries. And eventually he realized that that was kind of one of the most important aspects of the ecosystem. You mentioned this as well, tell us how you got your start and what led to the development of Anaconda. Peter Wang: Yeah, so I fell in love with Python as a programming language starting around 1999. Over the course of several years I've seen it come up on Slashdot, and then finally in '99 version 1.5.2 was released and I said, okay, fine. I'm going to go and actually play with this language a little bit. And I played with it and I fell in love with it. I was like, this is amazing. So I kind of used it as a side tool for a number of years, and then in 2004 I got a job really that was at the intersection of many things that were interesting to me. So I got a job doing scientific programming using Python as a consultant. And so I used Python and NumPy and SciPy and all those tools at the time, actually 2004 predates NumPy, it was still called Numeric and Numarray at the time, there were two different competing libraries. Peter Wang: So I did that for a number of years, and that's actually where I met Travis initially was at that consulting company named Enthought. And we had a wonderful time there, a lot of good people. But towards the end of the oughts both of us came to realize that Python has a much bigger, brighter future than merely sort of a consultants tool, and at the same time it had some challenges, some missing pieces that were sorely needed if we really want to take it to the next level. So I started the company with Travis at the beginning of 2012, with this mission of rounding out some of the things that were needed in the Python ecosystem, and then really just showing up as a vendor to say, look, Python's great. You don't have to use Hadoop for big data. You can use Python for really scale a big data. And you don't have to just use R for doing statistical processing, you can do Python statistical processing. And so we sort of showed up in a time, and now it's quite ubiquitous as a tool. It's the most popular language for data science, certainly. Peter Wang: And it's, depending who's counting, it's possibly one of the most popular languages in the world, but back then it wasn't obvious at all. I got a lot of flack from people who are like, "Why don't you just use R? Or what's wrong with you, java has clearly won. I mean, you look at all these Apache projects for big data, Java is the thing for big data." And I would say, "Well I don't know, I think Python's pretty good." So we showed up sort of as kind of the weird kids on the block, pushing all the, a bunch of science people, right? And at the time Jupiter was called IPython at the time, the IPython web notebook had just come out. Peter Wang: And Travis and I, we were friends with Brian, Fernando, Min, these guys who've been making the IPython project, and we've been pushing on them to make sort of a web experience, right? Because they had a rich client one and we wanted to make a web one. They made this web one and we said, "Look, this is the bee's knees. This is going to be awesome." We didn't quite say bee's knees, but we thought it was awesome. We're not that old school it wasn't that long ago. But we were like, this is going to revolutionize everything. This is amazing. And so it was a bunch of ex-physicists, ex-engineers putting together these computational tools, none of us had a CS background formally. And then we so show up at the big data party, we show up in various places where people are trying to use things like SAS and R and we say, "Hey, try Python." And now 10 years later it's, I think those efforts have really paid off. Eric Anderson: Yeah, I can identify some with the kind of Java world and the Python. I was at the Strada conferences and the Hadoop conferences. And I worked on the MapReduce team at Google and so I often was in the Java camp, and you're right. There was this bifurcation, it was like we were talking past each other. And you probably feel like you were vindicated eventually, Peter, the world's kind of tipped towards Python it feels like. Peter Wang: Yeah, I have a sticky note here on my wall and it says, the only fairness in life is the justice we create. Or I'll state it differently, be the change you want to see. I mean I'd done enough programming in C++ and Java to know that I enjoy doing Python a lot more when it came to data processing and scientific computing. And also to make it more serious, in those years that I was doing consulting we were brought into the heart of fortune 500 fortune 50 companies, with all the budget in the world on IT and compute. And we would find scientists, teams of analysts, you go into a JP Morgan or a Bank of America and you have quants, and they loved using Python. There's something about an interactive read eval print loop, and this was even prior to the notebook, right? This is just people having an editor. Sometimes they were just using Idle, and they would be able to think quantitatively, build little apps, run massive scale computations, and they love that they could do all these things themselves without having to do a big project with IT and with software developers. Peter Wang: And that concept of empowering the domain scientist or the domain expert was really fundamental to Travis and my mission, and why we were promoting the use of Python. Because we said, "Look, this is a tool that clearly fits in people's heads." And when you see the spark of joy on people's, in their face, in their eyes when they're able to wire something together and make it really work, that was something that we thought the world needed more broadly. So for all of the awkwardness and pain of Python packaging, I mean there's an XKCD comic about how terrible Python packaging is. That's kind of a level of infamy, there's infamy and there's XKCD makes fun of you infamy. Peter Wang: But for all of that pain at the end of the day when it works, we have to keep in mind what it's doing is it's empowering people who are not software developers to go and do some amazing things that without it, what would they be stuck doing? They would be stuck playing around with Excel or trying to beat a bunch of MATLAB, cobble a bunch of MATLAB into something usable as a user interface. I do think that ultimately we've been very successful in moving the world forward to more usable computing for a more broader set of people. Eric Anderson: Yeah, yeah. You're a champion for the every man, but specifically the domain expert, right? The people who can affect change if they could just put their ideas into action. And that the barrier not only was the language, Python gave them a path, but then the barrier within Python seemed to be packaging or part of it. My naive view of the world is you first start on PIP and then you'd get frustrated, at least when you're doing data science, and you end up in Anaconda. How did that evolve? How did that play out? Peter Wang: Well what's interesting about all this is the long sorted tale of how all this stuff kind of went down, if you will. So packaging in Python has been, historically it'd been a bit of an afterthought. And so there's a system called distutils, which is how Python loads modules and packages, then set up tools came around in the mid 2000s. And then it sort of got abandoned a little bit, and then people tried to fix some other things around it. PIP was written as a wrapper around set up tools to make it easier to use kind of, and then it also, I don't think he got abandoned but different people picked up development. And so it's just a classic case of band-aids on band-aids. Peter Wang: And one of the things that in talking to Guido van Rossum, the creator of Python, he just is very transparent that he never really was that interested in solving the packaging problem. So he was just like, always left it as someone else to do. He has this amazing quote, I think it was PyCon in 2017 or something, and he was at the Anaconda booth, or it was Continuum at the time so he was at the Continuum booth and we were just chatting. And he said, "Yeah, you know I really don't understand the whole packaging problem, because for me whenever I needed something I was just pulling into the standard library." It's like, that's a solution that only he gets to do. Peter Wang: For the rest of us the packaging system has been, well I guess on a sort of less glib note, the reason why packaging in Python is so much harder and more terrible historically than in lots of other languages is actually because Python is more powerful in this regard. Precisely because the Python VM, the virtual machine that runs the Python interpreter, a lot of users of Python don't really go below the surface of the ocean but there's miles of ocean down there. And if you look at the structure of the Python VM, one of the things about it is that it's got a low level interface that is not what your typical Python user would interface with. But if you write extension libraries and modules you can interface with the VM at a much lower level. And it's very powerful, you can extend the types available inside the programming language runtime. Peter Wang: So as an extensible VM, it was really powerful. And it's precisely because of this that we were able to wrap all of the C and C++ of Fortran libraries. The reason why the scientific community adopted the Python language so glommed onto it was because we could extend it like this. And the reason why Java has always had this, it's always treated the scientific community as a second class citizen, is precisely because the JVM does not have this level of a thing. The memory management, the jitting, all these amazingly powerful things inside the JVM make it really hard to interface with native code. And so Python doesn't, because it has this connection it makes it really simple. Now the downside of that is now you've got these extension libraries that drag in a pile of C++ and Fortran, and now the build system is horrible. Peter Wang: But to be clear if you look at something like Node.js NPM, it's kind of like a god-awful mess of package management, and they don't even have to compile a bunch of C++ and Fortran. So I would say the Python community as a whole has done kind of okay in having shipped all of this really complex, gnarly stuff, kind of made it basically work. And what happened is actually I have this wonderful set of pictures from, actually we have it on video, we have it on video at the first PyData workshop at 2012 at the Google headquarters. And Guido was working at Google at the time, so he stopped by. We're about 50, 60 people just in a conference room. And we're sort of like, "Hey, he's here." We didn't know he was going to show up. [inaudible 00:11:49] let's pull him on stage and do a panel impromptu. Peter Wang: And some of us were asking him about, there were questions about like, number one, "Will you approve a matrix multiply operator?" And eventually we got one, so that was good. The second thing was they asked them, "Can you help us fix Python packaging?" And we have him on camera saying, "Look, the needs of the scientific and numerical community around packages may be so complex and esoteric that you're probably better off building your own thing than trying to sit here and not fight, but spend a lot of time spinning wheels with the kind of packaging efforts happening within core Python." And so we said, okay, we take that as permission, I guess. And so that's one of the things that made us feel like we were okay in building kind of an alternative system for doing this. Peter Wang: Now it's a very large community of people who use Conda, who built the Conda recipes, a package of thousands and thousands of different kinds of libraries. Anaconda is actually not even involved in that stuff. We help with the recipes on some of the things, but then we take those recipes, we build them on our secure hardware, and we put them behind our repository which we curate. And so there's no Typosquatting attacks, there's no whatever. We make sure that all these different libraries, when you install them they actually work well together, we test them all together. So that's kind of the value of what you get when you use the Anaconda official repository. But we've always kept it a very open system and we've tried to play well with the broader Python packaging community, but everyone's resource constraint is the part of the problem. It's all volunteer effort. So anyway, that's kind of how we got to where we are now. Eric Anderson: And fascinating. I mean I don't think I understood that Python's, its rise was coupled to this fact that it could mesh so well with legacy code, and the Fortran and the C++ libraries, and that super power that made it so popular also made it kind of tricky to do the packaging, but good work. Peter Wang: It's a bunch of scientists and engineers went and modded this cruiser into an aircraft carrier. Yeah, and now it's all these aircraft carriers and all these planes landing on them. But the other part of it is actually, I think another sort of subtle part of this is that the fact that we did have a packaging system in Python meant that you could democratize innovation. And this is a really important thing for the ecosystem. It wasn't just one group of people saying, we're going to build the de facto compute system. It's lots of people innovating in parallel agreeing about certain kinds of interfaces for interop, and then that just allowed to kind of instantly this, well instantly in the space of a decade, completely take over numerical computation everywhere. You couldn't have done that if it was just one small group of nerds in some office somewhere saying, we're going to build the world's greatest computing system. You wouldn't get it done. So that's another thing that sort of pulls, or it's another driving factor for why there's a packaging problem is because there's so many people doing all these things in parallel. Eric Anderson: Yeah, the decentralized organic evolution as opposed to a central planning exercise top down. Peter Wang: Exactly. Eric Anderson: Yeah, fascinating. And so you set out years ago to put Conda and Anaconda in the works, and then you've been at Anaconda since. Meanwhile you've seen the rise of this Python community go from kind of the thing that's not R or Java that domain experts like to use to being the tool that the industry all over... I mean when TensorFlow was all written in C++ and when it came out, it came out as a Python. Peter Wang: It had Python on top, right, right. Eric Anderson: Where does that leave us today? I mean where's Python headed now? Peter Wang: Yeah, that's a good question. There's still a lot of people onboarding and learning. I think it's funny when I talk to some folks, even smart [inaudible 00:15:24] folks let's say in the venture community. And they look at a lot of technology, they've made bets in the past on languages, maybe some worked out some haven't, many of them still look at Python as sort of a language bet. And they miss the fact that it's actually, it's a kind of a language bet, but it's kind of the language of a new professional class. And so it's wrong to sort of think about Python in the frame of, is it Python versus Ruby, or Python versus R, or Java script? It's like, no, forget the computer programming CS nerd kind of people building apps, kind of as software developers, instead think about quantitative computing. Peter Wang: What do people use for quantitative computing? They use Excel. And so Python for a new class of people is becoming a tool to augment Excel, to augment Tableau, to augment the traditional business data analysis tools. So when you look at it from that perspective, you really have to understand the revolution behind Python as a people revolution, not a technical revolution. So the people revolution part of it is I think a really important part of my reasoning for my answer, which is I think Python is here to stay for a long time. Because the people who are doing this, they are not the kinds of people that are going to go and learn Haskel tomorrow because you can prove something about types. They're not going to go and learn Rust because it compiles. If they can bang out a 10 line data script that kind of still works, they're good. They've got lots of other stuff to keep themselves busy. And this is sort of, the point is a little bit proven by when you look at the other data languages, how long they've been around. Peter Wang: What is the longevity of MATLAB, of SAS, of SQL? These are other data languages for humans if you will, as supposed to CS nerds, and those things stick around for decades. So now that we've got a language that's basically good enough, they're going to be variants of it. So you look at Micro Python for doing embedded. You'll probably see some additional variants to run it in the web browser as WebAssembly becomes more mature. We ourselves have some thoughts about how to build maybe a slightly tweaked variant of Python that we can jit compile faster, or do autograd kind of things better on. Peter Wang: But all these tools, at this point it's just the shelling point that's so darn ubiquitous. It's really hard to imagine what could be so compelling that you will cause tens of millions of people to instantly overnight change to something else. And if you can't convince all of them to change very quickly, the sheer gravity of all those users is going to pull those things back into that language, that syntax. And I think that overall it's going to evolve relatively slowly to encompass and grow a lot of these additional features. Eric Anderson: No, you're absolutely right. They're developers who know five languages, they feel comfortable in Ruby and they'll bounce over to, as you said Haskell or Go. But the folks who know Python, that's almost a career effort and kind of skill that persists with them over the life of their career. And it persists within their industry, I mean their industry is built around these libraries and frameworks in Python, and I don't see those going anywhere. Peter Wang: Yeah, especially if we're around to continue being a steward and shepherding a lot more great open innovation in this area. We're really coming to the end I think, of the idea of shrink wrap software being the primary value driver in the value chain. A lot of this capability in MLAI, which is clearly the cutting edge of where everything is going, all the software capabilities are being released in the open source. So people are going to basically pick up all of this stuff, you have this massive source commoditization wave that is going to keep adding value to the skill for people. So it's really, I call it literacy. Python is the language for data literacy at this point. Obviously SQL is another important one and R has its adherence as well and I don't want to poopoo R, it'll be around for a long time as well. But ultimately between Python and R and SQL, these are the components of data literacy. That's going to be a lingua franca that every white collar professional is going to have to know probably five, 10 years from now. Eric Anderson: Now you mentioned how we think about commercializing software, the business of software, and the role of open source. You had to make some decisions as you turned Anaconda into a business, how has that framed, maybe you could share some of your kind of decisions around commercializing Anaconda and how you're seeing those types of decisions play out for the industry. Peter Wang: Yeah, it has been a journey and a process, right? So we started very honestly being like, Hey, let's build some additional advanced libraries and try to sell those. People are kvetching about this is slow and that is slow, so let's make some fast things, and then we'll just sell the fast things, right? We're going to charge like $50 for them or something. And surely some hedge fund quant who's getting paid half a million a year can afford $50 for a faster Pandas or faster NumPy. That theory did not play out very well. It turns out that the anchoring effect is real. Someone could get, if you really have them think about it, probably tens of thousands of dollars of free value from open source, but if you ask them for an incremental dollar it feels like you're poking them with a hot iron. They just somehow, the human psychology and the anchoring is not great there. Eric Anderson: Yeah. This is like the equivalent hour of your compensation and you'll fight it tooth and nail, exactly. Peter Wang: Yeah, in the time it takes you to tweet at me, how dare you charge to open source, you could've just paid us the money and then been good. And then we would have taken the money to build more great innovation in open source and in the... But no, so that didn't work out. That was circa 2012, 2013 timeframe. And then we built a GPU compiler, we were trying to sell GPU compiler for a couple hundred bucks and people were not having it. I am grateful to those people forever. But ultimately it wasn't going to scale. Peter Wang: And then of course the open-source community felt betrayed. They're like, you're taking all this open source stuff and then you guys, big corporate fat cats with your Delaware C-Corp, now you're making all this money. You're making like $200 off your source. And so it's like, that didn't play well with the community either. I mean most people understood, but there was still vocal people in the community who are very sort of absolutists about this sort of thing. And then as big data rolled on and the ML started coming around, we start seeing all these companies around us raising tens of millions of dollars building on open source, just taking Scikit-learn, taking NumPy, Pandas, Jupiter, and rolling them into I would say relatively shallow veneers, thin veneers over this bundle of things. And they would go and raise really a very large sum of money and sell their offerings at a very high margin, and then not give anything back to the open source community either. And so you're like, well that's not great. That can't be the way this goes, right? Peter Wang: Because I like to tell a story that I'm friends with many open source developers who write these foundational libraries, and I know that these libraries power literally billions of dollars of value at every single [inaudible 00:22:19] company. Every single one of these companies. And the problem is that these developers then, they're literally doing this as a side gig. And I don't want to call anyone out by name but I'll say that one of the most popular visualization libraries possibly in the world, the lead developer, he has to do it as a nights and weekends gig. He has a full-time job as a staff scientist at a national lab. And that's just a damn shame because I know that I won't even name it, but every single Silicon Valley startup that uses any kind of data science uses his library. And they should be paying him at least $100,000 a year for the value they get, but they don't. Eric Anderson: Or on the contrary, think of all the value they could benefit if he were given a little extra time to work on this thing. Peter Wang: Well if he and the people he identified who were aligned with the values of the project and who could do good work, the craftsmen ultimately that could do this. I mean maybe this is, maybe we should make open source, maybe we should put open source source code onto the blockchain as an NFT. Well I guess it's a git, so it's already in blockchain. But if he can wrap up each releases an NFT and you bid on it before it got cut into a release, I don't know. But the idea here is that these people are artists and craftsmen building incredibly thoughtful work, getting paid nothing. Meanwhile people who had nothing to do with it, take it, bundle it up, put a veneer on it, and then raise hundreds of millions dollars. And so I looked at this and I said, this isn't great. I don't begrudge the people who build good businesses raising money, that's fine. But it's like, this is not sustainable economics for this community that I care about. Peter Wang: And I've been an open-source nerd since like '95, so I started the Linux users group of my college. I was on the Slashdot bandwagon the whole time, like the XBill was my screensaver for a number of years on Linux. So by the 2010s I realized that open source has got to actually figure out a way to make money sustainably and honestly, to put regenerative economics back into the ecosystem. So that's really where Travis and I had a lot of discussions and thoughts about how to, and Travis to this day at [inaudible 00:24:18], what he's trying to do with open teams and many of these things, is to really think about what is a [inaudible 00:24:24] and open teams, all a lot of these things he's trying to think about how do we align the economic incentives of network collaboration with sort of the standard market capitalism sort of economics? Peter Wang: That's a hard question. It's one that actually we face as a species, because a lot of things that we do in the business world today are deeply extractive and not good. So you have sustainable investing as a real movement now, and I'd say that we're in our own way trying to figure this problem out for the open source community around software. So where we ultimately got to last year was I said, "I think what we could do is we can charge for value." We turned on a commercial repository for packages. We looked at successes like Red Hat, which was very successful in building a commercial package repository and charging people a fair price for timely updates to timely, secure updates, to sign packages with security guarantees around them and things like that. Businesses know what that's about. Free users can still have access to the free stuff, it's all great. Peter Wang: So we essentially went to that same model. And so with the package repository we have the community package repository that's free for everyone to use. We have our own commercial one as well that then the commercial users should be using. And if you're at a big business over like 200 people, you should be using the commercial repository. The price is not onerous, it's basically $15 a month, and in volume it's even lower than that. So we tend to do ELAs with big businesses and the value is clear. It's going to cost you like $100 a year to have a data scientist be productive and not waste a bunch of time trying to fight with packages. And meanwhile you have an actual vendor to call and when you get audited, you can say, well let's see, on our customer protected data we were running actual software from a vendor versus rando packages from the internet. Peter Wang: So it seems like kind of a no-brainer, so we have been actually selling pretty well with that. And then also in conjunction with that we announced our open source dividend program. So this year we're putting 1% of our revenue, we're going to just write that as a check essentially to the open source community through the NumFOCUS Foundation. And my goal, really sincerely my goal is to increase that percentage as we go forward in time. So 1% is a small number but it's a starting point. I mean Google and Apple don't do it, but I will. So that's kind of what we're doing now with that. Eric Anderson: That's fantastic. It's an exciting time. You mentioned, we chatted before the show just briefly and you mentioned that there's been some big companies who've had an open source association that have created, as you mentioned, this kind of wall of capital. I mean there's a lot of fundraising now around open source and it's kind of unproven as to whether this is all going to amount, we're going to see a whole wall of exits or or not. Peter Wang: Well there'll be exits one way or another. Eric Anderson: Yeah, yeah. But it's an opportune time for us to kind of figure out how the economics of open source can work in everyone's favor, and you seem to be kind of at the bleeding edge of figuring that out. Peter Wang: I really do believe we're kind of coming to the end of the era of software being that big margin driver. And so if people are able to use the open source software adoption as a way of pivoting themselves or bootstrapping themselves into a different kind of business model, the most obvious being cloud. So if you look at the database companies, Elastic, Amongo, these companies they use open source as a way of getting a lot of developer attention. So the scarce resource that they're actually stockpiling is developer retention and knowledge of APIs, and then of course production deployed services tied to those APIs. They're able to then pivot that into essentially a SAS business model. There are other business models that you can turn this kind of developer attention and adoption into. So we're doing something I would say related but different. I think that actually for us, we're closer to something like a Roblox's where it's a developer, it's a maker, a creator community, and it's a sharing network. It's an ecosystem and network a marketplace. Peter Wang: And so that, just to be completely candid in my conversations with potential investors over the course of the last year, has been embarking on this particular strategy. People seem to grok it 60% of the time. I mean they still want to apply SAS revenue metrics to it, but they kind of get like, Hey, that's kind of interesting but that's still kind of unproven. And so yeah, it's a little bit bleeding edge, but look, we have a massive user community. And so our job right now is to activate that community, get them sharing, connecting with each other, putting some more economics into it for everyone. And then once that stands up and really gets going, I think people will understand, people who are classic, down the fairway spreadsheet investors will be able to look at a spreadsheet and say, okay, I get how all these metrics tie out. Peter Wang: But I think that's really the problem that open source suffers is that for a long time people don't quite know what kind of platypus is it? Because you look at it in one way and you say, well, you're giving away all your IP. So there's no defensible IP, this isn't worth anything. Then smarter investors are like, wait, hold on. But you're getting a ton of users. If you turn into a SAS business your metrics are great. And so that's where your database companies exit. And then when you, I think the next generation is going to be looking at, not even the open source of it as much as just how much traction do you have, how much usage do you have, and then converting that usage into what kind of monetizable economic energy? Peter Wang: So if it's a marketplace, what's your VIG? If it's a straight up pipe, how much of a loss are you getting through that pipe and what's your cut of it? So, I mean I think that we really are at the very beginning, I really liked James Currier at NFX, he writes a lot about enterprise gateway marketplaces. That's what we are right now with our commercial B2B side. But then when you look at something like Roblox, I really see that as being a touchstone of kind of what I think the AI data science machine learning creator community could be, and certainly Anaconda is the platform for that. Eric Anderson: And Roblox is not a bad comp. Peter Wang: It took a number of years, but hey man, not a bad comp. Eric Anderson: Yeah. I mean my understand is during the pandemic they like doubled usage. All the kids weren't in school and flocked to Roblox. Unfortunately I don't know that everyone all flocks to their Python notebooks, but maybe they did during the pandemic. Peter Wang: We have seen usage grow tremendously during the pandemic. A lot of people picking up data science, a lot of people taking advantage of all the downtime they have at home and the lack of travel time. They're like, well I'm going to go and do this data science thing. Eric Anderson: Yeah, time to up skill. Peter Wang: Time to up skill. So we're definitely seeing that growth last year. I mean it wasn't like Zoom-like growth, but it's been continued sustained energy and adoption and stuff. So I'm very pleased, and I think that we're really just the beginnings of a really interesting ecosystem and network activation. Eric Anderson: Well and you have a unique view on that, you're the [inaudible 00:30:53] is maybe not the right term, but you see people when they learn Python and you see the professionals that are hammering on it every day, and you can watch their progression, you can see where the community's headed. Peter Wang: Yeah. It's really kind of weird being able to, just being the package distributor. We sort of can see a lot of trends around different cloud services. We can see trends around what are people actually using versus what gets hyped. We see a lot of different kinds of trends, it's very interesting, Eric Anderson: Peter, as we wind down here I want to give you a chance to tell us where Anaconda's headed, if there's any details of share. And also how people can get involved in the work that you're doing if any of this has excited them. Peter Wang: Yeah. Well so Anaconda is I think really headed on a great trajectory, I'm very excited about what we'll be doing this year. I sort of already previewed for you kind of the strategy of what we're doing, we're revamping a lot of our offerings to really orient towards this creator community marketplace and helping creators and users connect better. Also having a role for vendors, for commercial vendors and businesses to plug in. So really excited about that. We do have a number of, I think we have 14 open positions right now, so would encourage people to look. We're hiring for everything from data science, to dev ops, across the board. And we are a very remote-friendly place and will continue to be moving forward. It's not just because of COVID, we were pretty remote friendly even before that. So that's where Anaconda is headed. Peter Wang: And also if you don't want to necessarily work for us you can work in the community. We have a lot of open source projects around visualization, scalable computing, data access, and so many other projects in the community. I just encourage people who are active users to think about how they can sort of be more makers in the community, start a blog, start a podcast, publish some of your notebooks, do some analysis and post on Twitter, just do more of that making. I think that makes the whole community feel more participatory and more accessible, and that's a really wonderful thing. So yeah, that's it Eric Anderson: Awesome. Just a plus one, what you said about contributing on multiple levels, I don't think people realize the power of evangelism. Projects will win and die not on technology but on the ability for others to communicate the value and the vision. And that can be somebody on YouTube and in a blog post as much as anything. Peter Wang: Yeah, and it's really, data science is a global phenomenon. I've spoken in India and Japan, I've spoken at PyCon Taiwan. And what's amazing is this really, it's kind of a beautiful almost enlightenment ideal sort of thing of quantified rationalism kind of empowering people across the world. And so in particular right now in the data sciences, I think a lot of the innovation in the open source stuff is very kind of Western Euro-centric, but the usage is global. And so it can actually have this weird effect where people who are not in the kind of Eurozone and kind of Western countries, they may feel like they're a little bit imposter syndrome. Like should I publish my blog or whatever? Peter Wang: And I just want to, if anyone is listening to this podcast who is in not the Eurozone and the Americas, if you are in one of these other countries I would strongly urge you to please actually tell your story. Please publish your work about doing data analysis relative to problems in your area or in your sphere. It is a global community and we want a global conversation across all the different economic areas, and language, and cultural zones. So just really want to emphasize the diversity there around that. Eric Anderson: Amen. What a great note to end on. Thank you, Peter for your time. Good luck. Peter Wang: Thank you, Eric. Eric Anderson: You can find today's show notes and past episodes Contributor.fyi. Until next time, I'm Eric Anderson and this has been Contributor.