The a16z Show - a16z Podcast: AMPLab, the Power of Open Source, and the Future of Systems Software

Episode Date: October 14, 2015

The place where Apache Spark was born, UC Berkeley’s AMPLab has not just created a major open source software platform, it’s spun out more than its share of ground-breaking companies (full disclosure: a16z has invested in three of them). So how did they get there? How has open source and the AMPLab approach reduced the friction between student and faculty ideas and launching them into the real world? Co-founder and director of the AMPLab, Michael Franklin, joins a16z’s Peter Levine to discuss the AMPLab model, and their own relationship as an academic and an investor. Haoyuan Li also joins the discussion -- which was part of our 2015 Academic Roundtable -- to offer another perspective. Li’s company, Tachyon Nexus, came out of work he did as a student in the AMPLab and the resulting open source project. Li describes his struggles and victories in making the transition from student to founder and leader of a company. The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments and certain publicly traded cryptocurrencies/ digital assets for which the issuer has not provided permission for a16z to disclose publicly) is available at https://a16z.com/investments/. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information. Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

The content here is for informational purposes only, should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security and is not directed at any investors or potential investors in any A16Z fund. For more details, please see A16Z.com slash disclosures. Welcome to the A16Z podcast. I'm Michael Copeland. The place where Apache Spark was born, UC Berkeley's AMP Lab, has not just created a major open-source software platform, platform, it spun out more than its share of groundbreaking companies. Full disclosure, A16Z has invested in three of them. So how did they get there? How is open source and the AMP Lab approach

reduced the friction between student and faculty ideas and launching them into the real world? Co-founder and director of the Amplab, Michael Franklin, joins A16Z's Peter Levine to discuss the AMP lab model and their own relationship as an academic and an investor. How Yuan Li also joins the discussion, which was part of our 2015 academic roundtable, to offer another perspective. Lee's company, Takyan Nexus, came out of work he did as a student in the AMP Lab and the resulting open source project. Lee describes his struggles and victories in making the transition from student to founder and leader of a company. The AMP Lab example and what every academic and entrepreneur can learn from it on this segment of the A16Z podcast.

So let's start with you, Michael. You were an entrepreneur in between being an academic for a long time, but Amplab came out of the desire to do what? And how did you sort of make that happen? Yeah. So, boy, where do you start? So, yeah, so it was interesting because I was off, I took two years off. from teaching at Berkeley to get a company started, and then kind of came back after that.

Jan Stoica, who's also one of the founders of the Amplab, at the same exact time, went off and was away for two years doing a company. And when we came back, I think there were a couple things going on. One is, after you've done a startup, academia can seem a little slow, a little quiet. And so, you know, you get used to being in this environment where kind of every day is kind of a life or death situation. Right. And you walk in, and, you know, mostly in academia there's definitely life and death situations, but they're sort of not every day.

Death is slower. Death is slower. Right. And so I think, you know, I won't speak for Jan, but I know from my perspective I was kind of craving a little bit more excitement, a little bit more interaction. You know, a lot of traditional academic environments are a little solitary. And so there was a project going on at Berkeley called the Rad Lab, which was a project of systems people and machine learning people who had gotten together to do what at the time was called autonomic computing, where you use machine learning algorithms to help manage large computer systems. And so I ended up joining that thing, not because I was so interested in autonomic computing, but just because I like the environment of having all those people around. The problem was that Radlab was getting ready to wind down.

It had been started with a five-year timeline. And so they were thinking of splitting up. They were talking about how to split up the lab. You know, we had this big open, collaborative space. And, you know, when I had been out in my company, I saw what now is what's called big data revolution or whatever. But every company I talked to to try to sell them stuff, it was just clear they were getting more and more data.

And just that seemed to me. to be kind of the perfect problem for this group of people who on one hand did large-scale computing. On the other hand did machine learning. And started talking to Jan, who was out seeing similar things and we realized that was the next thing to do. So it was a technology trend that you really, I mean, it's sort of bigger than that in a lot of ways, but it goes out from there. But big data was it. Yeah. Yeah.

And I think just about everyone here is a professor. I mean, the message I gave, so when we have a new change. Chancellor at Berkeley, and when he showed up, he had heard about the AMP Lab, and he kind of wanted to understand more about it. So he invited me to tell him about it. And the thing I said to him was, you know, if you remember one thing from this conversation, here's what I'd like it to be, is that AMP Lab happened because there were two faculty members who got to go off away from the university for, you know, a couple of years, and engage with what was going on out in the world. And that's where we saw it coming. We saw it before a lot of other people in academia because of that. So AMP Lab, you know, it's working. I mean, it's from our perspective. And Peter, I want you to jump in here. It's like it's working incredibly well, both from an academic standpoint, you know,

to get this kind of critical mass of people and ideas together. And then boom, out the other end comes, you know, all these great companies and ideas and projects. But so, and I want you both to answer this. Peter, you first, what are the ingredients sort of the AMP Lab has mixed up or that you can identify, that makes Amplab work so well from your perspective? From my perspective, I think the most interesting aspect of the Amplab is sort of this macro trend in system software now being done by a whole new generation of folks. Enterprise and infrastructure software, system software was traditionally done by folks, let's say,

leaving Cisco or leaving Oracle. And it always was a derivative, like you work at Cisco and then you go do a startup or you work at Oracle and go do a startup. And it was always kind of, to me, there was a bit of incrementalism. And it was sort of an older generation of people actually doing the work on some of those new companies. And what's happened in many universities, but Amplab in specific, is we find that there's now the new generation working on the most. interesting computer science problems applied to system software. And yes, big data, but it's databases, operating system software, kind of all the infrastructure pieces. And I got to tell you, that has fundamentally moved the needle in the industry to where, you know, the companies coming

out of there are way, way different than what I found, you know, from people coming out of, let's say, Oracle or Cisco. It's just a revolutionary new thinking. And kind of it just opens your eyes to really new ways of doing computing. And that's been, to me, that's been amazing. If you were to distill that to an ingredient, for lack of a better word, I mean, is it focused then? Is it like that you guys don't try and boil the ocean and solve every problem, but you're focused on, look, if I want to do research

and do my graduate work in infrastructure and, you know, big data, I go to Amplat. No, we were trying to boil the ocean. Are you telling us we didn't know the ocean? I don't know. I mean, I think from my perspective, I think another way to say what Peter was saying was, first of all, open source has made a lot of this possible. And so, you know, I always tell people when I started my career as a database systems researcher, you know, you'd come up with an idea. If you thought it was a pretty good idea, you'd go and you'd tell you'd give a talk at IBM and you'd give a talk at Oracle. you give a talk at Microsoft.

You know, and, you know, either they would take your idea and steal it. Steal it. Well, they'd hire you. They would put it in the system. And then, you know, maybe you'd hear about it. Maybe you wouldn't. Or they'd say, oh, no, we don't need that. Whereas now, you know, the way our students do work is, you know, they work on a piece of

software. If it looks useful, they blog about it or somebody else does, or they put it out in the Amplab repo and flip a switch and all of a sudden people start trying it. So the friction from, you know, student idea to actually people using it. I think that's great. It was a great friction has been way reduced. And so then that friction and that relationship between the corporate world then, how do they arrive at the Amplab and then how do you leverage them? So, yeah, so, I mean, I don't know if everyone here knows a lot about the Amplab.

But one of the things that's interesting from an academic perspective in the Amplab is our funding comes about half from, you know, the government from NSF, and DARPA and places like that, and half from companies. So we currently have 30 companies that help sponsor the work. And we engaged with them very early, telling them not only what we've done, but what we're thinking of doing. And that has a couple nice features. One is we get kind of instant feedback on what sounds right to them and what doesn't ring true. And we also get feedback from them on problems that they're having. that we're not really looking at.

And then the other advantage it has is that by getting them involved early, they're more inclined to try to use what we do and give us feedback. Right. So who gets IP rights among all those contributing companies? It's a brilliant question. And this was done way before Amplab. You know, Berkeley has a long history of open source, and so there's been a bunch of projects that work this way.

And the way it works is we don't give the companies any IP rights. And we... It works for you, at least. It seems to work for the companies, too. We can talk about who it might not for, maybe. But, you know, the idea is we do everything in open source. And so then the next question is, well, then given that the companies could get at the stuff anyway, why do they give us money?

And that's a more complicated question. Different companies have different reasons. Well, Peter, you have a good answer for us. You invest in multiple open source companies. So open source projects that become companies in a different sort. We'll talk, H.W.A. with you about it more in a minute. But how does it work for you? And when you're approaching an academic institution, where do you start?

And let's take Amplab. How did that relationship even blossom? Well, we had a number of connections in the, to Amplab. Mike and I actually work together in a past generation, so we knew each other. Ben, of course, knew folks there. And so we kind of, look, it's always been a matter of finding a couple of people inside a university and then

building on that relationship. I think what's been unique about Amplab, it's actually one relationship with multiple projects. Typically, it's one relationship with one project. So it has become a very effective sort of a mechanism for us to go in and actually look at all the projects going on. It's just a centralized place at Berkeley, which has been fantastic. And then, look, I mean, once we see the projects and we look at them and evaluate them as potential businesses, we invest as we would any other company,

right? And so then it starts as, you know, commercial enterprise and we go from there. Maybe you can respond to this question of IP and the corporate world, because, you know, again, In the open source project that become companies, you have a very specific kind of view of the world and why open source works as a business. And that might sort of answer to why the corporate world is fine with investing the way they do. Well, one of my beliefs about open source, which gets to sort of this point, the IP ownership, first of all, with open source, it's out there. IP ownership is, to me, less important than who actually wrote the code. And so what I look at fundamentally for an open source project is I'd like to see that the inventors of the code are part of the company. And whenever you have a fork, which is, you know, let's say you have two projects that use the same code base after all it's open source, the fork typically is not as thoughtful as the folks who have invented it.

and all the projects coming out of Amplab. In fact, one of the rules, I do most of our open source investing, and that's one of my guidepost is to say, look, it's got to be a real exception if we're going to invest in something that's a fork of the real folks doing it. And so what we have found out of Amplab, it is the inventors of the particular project who are starting the company,

and that gets you great competitive differentiation. Because if anyone forks it and tries to come up with another team, It's like, look, we're the inventors of the project. We run the roadmap. If you want support or add-on products or whatever it might be, the fork company is never going to be as strong as the one who has invented it. So the projects coming out of Amplab certainly fill that first requirement for me. You know, it's an interesting question that we get periodically whenever somebody new shows up

in the administration of Berkeley of, you know, why are we doing this open-source stuff? Why don't we patent these things? And then there's this conversation that you have to go through about, well, okay, you know, how many software patents, you know, we brought in, you know, eight figures worth of, you know, industrial donations for AMP Lab. How many software patents has the University of California ever had that brought in that much money? And the answer is maybe one. They had a patent on the browser, I believe, right?

The first browser, where's Mark? Yeah, Mark's out here. One of the very early browsers was done at Berkeley. Let's put it that way. And they had some patents on that. And I think that was the one that they made money on, but I'm not even 100% sure. And the good news is it'd be hard to find an administrator at Berkeley who could give you the answer either. And so that's the thing.

Once you have to make this argument about the importance of building the brand, building the level of activity. And then, you know, it's a little indirect. But, yeah, these, you know, the opportunities for philanthropy start coming in. And, in fact, Berkeley now started this thing they call the Founders Pledge, where they ask people who are doing things in the university, you know, to pledge, you know, some portion of their future revenues or profits or whatever. And it's non-binding, but whatever.

So they're working on it. Michael, how do you view projects that come out of Amplab and research? Is there sort of a sense that you have like, oh my God, here's another one. Like this one's going to be off to the races and let's go. Or how does, I just wonder like the sort of when research becomes commercialized or commercializable, what are the clues there? I don't know about commercial or commercializable. Geez, that's a tougher one. I mean, just I thought where you were going with that was sort of what's a good idea and what's not a good idea.

Well, we can go there. Let's start there. Whether it's from a research perspective. Or do you guys even care whether it's commercializable or not? Yeah, I think that's why I hesitated because that's not really high on our list when we start thinking about what to do. Yeah, that's what I would imagine. So what's a good idea? And then I'll ask you the commercializable part because you probably do care. Well, wait a minute.

By the way, you know, I don't want to say we don't. care, I'm just saying that's not the first thing we worry about. But I think, you know, a lot of it just has to do with somebody in the lab being, you know, being, I mean, it's going to sound cliche, I guess, but being passionate or being excited about working on something. Because, you know, that's what happens. You know, a lot of our most successful projects now are things where, you know, somebody goes off and works on some piece of the system, you know, puts it out on GitHub and somebody finds. and somebody finds it, and all of a sudden there's a community starting to form around it. And so, you know, we're in this environment now where ideas go out there quickly, and then very quickly you sort of see what gets traction and what doesn't.

So it's traction, it's sort of passion, it's, you know, and it's got to be a good idea. I mean, that's what a good idea is, traction and passion? Absolutely. And, well, and also, I mean, we are a university. We're a research lab. where our main output is great students. And, you know, one thing I want to mention to all the professors here, you know, you might think, well, okay, but, you know, I have this problem.

I have to decide, do I want to do, you know, good research, or do I want to do things that are, you know, maybe useful? Right. And, you know, if, I think most of the people, since you're sitting here, I'm probably preaching to the choir, but that's a false dichotomy. I mean, we were very fortunate in this current year. You all know, you all know the ACM dissertation awards. They give an award to sort of the best PhD work in computer science around the world.

So every school gets to nominate two people. I'm chair of the department. It just so happened last year. The two people got nominated were from MAPLAF. Just a coincidence. And, you know, once. Mattes, Aria won the award. He's a guy who did Spark.

John Ducci, who's the other guy who nominated, got one of the two honorable mentions. So of the three awards, ACM gave worldwide, last year, two of them came from Amplett. What does that say about technology and kind of its place in just broader society then and culture and research, et cetera, in academia? I think there's probably a better appreciation now for impact.

Right. And maybe there was an academia, you know, even five years ago. I think that's right. If you look at, you know, a lot of innovation up till, well, to me, up until recently, it came from like Bill Labs and IBM and Microsoft Research. And that's where a lot of the new, I would say the new commercial ideas came from folks leaving those or pushing the envelope. Now I see a lot of it coming out of universities, which is, that's fantastic.

I have to tell this great story, though, if you don't mind. Please. So we had a delegation from China visiting this weekend, and we showed them, you know, the AMP Lab and all the stuff we were doing. We showed them a bunch of other data science things going on around Berkeley. And at the end, the president of this university, who's very prestigious university in China, said to us, well, whose plan was this? You know, who came up?

This is, you know, the work is wonderful. Who came up with this plan? And, you know, we all sort of just looked each other and laughed because there really wasn't a plan. It was just, it was really just a bunch of, you know, different faculty who were doing what they were interested in and decided to collaborate. Let's shift gears a little bit to advances in computer science. What are the most important things you're starting to see and what are getting your students excited? And Peter, you know, answer this question as well. I think so the whole idea of big data has marked.

I think it's a generational kind of, there's multiple aspects of big data. Now that we're able to collect data and now we're able to do real-time big data a la spark, I think machine learning and deep learning are probably the next. I mean, to me, I see a lot of those projects, you know, people coming either out of university, out of Google, out of a variety of places to think about how we process and become much more predictive about the information that we have. And I'm actually pretty excited about that. I mean, there's a lot of noise in the space right now.

There's a lot of people sort of interested in looking at that. But I think there's a relatively new frontier that's occurring at a lot of universities around sort of machine learning and deep learning. Yeah, I mean, so machine learning, deep learning. I also want to know, like, with your colleagues... New databases would be another thing. Right. Anyway. With your colleagues outside of the Siesta Department,

Is that all part of their language now? Like, oh, I'm working on a big, you know, AMP lab. We're all about big data, machine learning, you know, deep learning. Do they get that? Yeah, I think a lot of people do. I mean, one of the big buzzwords around our campus these days, I think most campuses is data science. And that one tends to be a little more inclusive than big data.

Big data is looked a little bit as kind of a computer science thing. Whereas data science, really all around campus, I mean, even astronomers go around saying that they're data scientists, for example. So data science is a really hot topic and one about how do you solve real analytical problems in a bunch of domains. I've been getting interested in things around trying to, I've been doing databases my whole career. So I'm sort of interested in reaching out from the database to actually go out and affect the world. So, you know, if you look at things like cloud robotics or, you know, think about things you can do with drones or anything like that. I think a lot of people are getting excited about Internet of Things, but in the new version, you know, in the old version of Internet of Things, it was just you got sensors out there and you're collecting the data and you're doing something with it.

In the new version of Internet of things, you're actually going. out and interacting with the world. And that involves also the people that are in the world that have to coexist with all these machines. One of the areas that we're seeing from, you know, both from universities and new projects coming up, this whole idea of moving the compute out to the endpoint. So like, it's really interesting to see this, the ebb and flow of computing from sort of centralized to distributed. We've been in this centralized world, cloud computing, and then you have the endpoint which shows pixels. And now we're starting to see a number of really interesting projects where the endpoint,

our supercomputer in our hand, is actually being used as a computer, not a display device. And I think the Internet of things and drones and however you want to define these, this sort of range of computers that exist, not so much in our pocket, but out there in the world, I believe that there's going to be computing that will be fundamental to how we start to think about data and, you know, real-time analytics that occur not so much in the world. the cloud, but at the endpoint. And so we're starting to see that. In fact, out of universities and all of that. So there is a trend now back towards, I predicted distributed computing will be back. And it's going to be with all these endpoint devices and drones. Michael, I've been asking you one last question, and we'll bring H.Y up. Amp, the P is for people. You just mentioned people. And maybe you guys were a little bit ahead of yourselves five years ago,

about how do people fit into sort of the AMP Lab view of the world? And again, when you think about these sort of database-backed Internet of things and how you bring in all the people and all the data, how do people fit into it all? Yeah, it's interesting because the view of people in the lab is, or the role of people in the lab has evolved since we started. When we initially started, the idea that we had was that algorithms, in terms of machine learning, statistical processing and machines.

The idea was cloud computing and large clusters. And then people, what we were thinking about, well, these are the three resources that you have available, three types of resources you have available to make sense of data. Make sense of big data in particular. So our view of people, when we originally started, was to think about human computation and crowdsourcing and things like that. In fact, Tim's sitting back there.

Tim was one of the early people in the App Lab. and we worked on a project where it was called CrowdDB, where I like to call it the world stupidest database system, where if you asked it a question and you weren't attached to the network, it would just say, gee, I don't know the answer to that question. What do you think? But if you were attached to the Internet, then it would create a bunch of jobs on Mechanical Turk or something

and send them out and let the crowd answer. And then all of a sudden it wasn't such a dumb database. system anymore. And so, you know, we still have a big part of what we're doing in terms of bringing in human-in-the-loop analytics, looking at how do you get people either as, you know, individual experts or analysts or as crowds to do data cleaning. The idea was you wanted to be able to bring in people to solve those parts of these, you know, machine learning problems that the machine learning wasn't quite up to speed for. And how things have evolved kind of over the five years is now we're also being a little more concerned with the fact that, you know,

ultimately the results of the analysis is, in many cases, going to end up in front of a person. So how do you support that person in their job? But the original vision was really about human computation and crowdsourcing. Great. Well, Michael, we're going to kick you off now, and thank you so much. All right. Thanks, guys.

So H. Wally, he was the co-creator of Takyan, right? And so he is, as you say, Peter, the original OG open source guy on this project. And you can describe it. it's a memory centric distributed storage system, which you've now formed TachianNexus, which is the company that is built around that. Maybe it just gives a brief discussion of TACON and what it is, the problem that you've gone after. And actually, why you wanted to address that problem to begin with?

Oh, okay. So Tiken is a memory-centry distributed storage, and I think from the high level, the vision, is that memory-centric computing is the future, and you have a computer layer of a storage layer, and we want to be the storage layer. And why I wanted to do this problem, that's interesting.

So personally, I'm very interesting in storage, but when I joined like Berkeley Amplab, like four years ago, so at Amplab, we have other two projects, like Mesos, Apache Mesos, and Apache Spark, are very successful and actually the company behind those projects also like funded by Andreessen Harous and and and and but at the time the storage piece is still missing so so I was fortunate that my advisors and the lab give me the

opportunity to work on this piece and it's just a very fascinating to to work on something you think is a very important interesting yeah so Peter let's get back to this good idea bad idea question And clearly what H.I. was working on, you thought was a good idea. So what were the attributes of that idea that you thought then, and or evidence, that you thought this was something that you wanted to put your time and effort and the firm's money behind? When we as a firm look at what's a good idea versus maybe not such a good idea and time will tell, you never know, but we go into it. All our investments we obviously go into it think is a freaking great idea. They don't all turn out that way.

So that's just one caveat. The view here, I have been in the storage space for a long time. And when we met, one of the concepts that we've been talking about internally, imagine if memory in a computer flattened down to where there was only one memory hierarchy, or let's say two. I know you may not think so. But like, well, hang on, though. Spinning disks are going away.

Okay, so if you take that to the next lot, you know, we have SSDs and we have RAM and then we have, you know, CPU memory. And part of my belief was that if memory gets cheaper and cheaper, why don't we have a file system and why don't we have an architecture that actually supports this new type of computing? Because lots of computers and operating system code and the way we write files and the way we treat structures and all. all this stuff has all been written under the assumption that there's this memory hierarchy that goes from very fast and expensive to cheap and slow. And everything we do bounces through that hierarchy. And my belief, whether you believe, you know, look, I love that people disagree, you know, for the most part, when more people disagree, that's the project we want to go invest in.

Because like, you know, it's too crazy for, you know, hell, that'll never work. but when enough people disagree that there will never be flat memory structures, then it's a great idea, you know, at least to get started, right? It may never happen, but that was the thesis, that memory was going to get much. I mean, look, you look at cell phones, and you look at the memory in there, it may not be fault tolerant and may fall apart all the time, but part of the systems we want to go solve for that. But it's very inexpensive.

And so to the extent that the mobile supply chain starts to eat the back-end data center was kind of our thinking on how this memory, you know, in-core memory starts to flatten itself. So with that, if you make that assumption and if you believe it, which I, you know, believe to some degree, then this project exactly maps into that future. And even, I would say, even if it doesn't quite map exactly like there, it's still a great project. So like the grand view would be memory flattens, and this becomes basically the in-memory file system for all computing. And if we don't quite get there, it's still a great file system for big data and other applications. So that's sort of the, you know, we sort of look at the grand view and then say, okay, well, if that doesn't happen, because we were wrong about that, what else can this become?

So that, I mean, like when we look at ideas, I try to look at ideas on that. that dimension to say, you know, what transformations might occur in the future to where there really is a big need for a whole new design of something. And this was one of those ideas. Now, I do build taxonomies in my brain and probably the simplest thing that I do, and I'll leave you with this, like how do I actually, like when you see something, how do you know or how do you choose to invest in it? So what I have, I've been here now five years, and my whole philosophy around this is I take something that's very popular right now, cloud computing, and I say, what happens when cloud computing doesn't exist anymore? What fills that place? And all of a sudden, like, and I put in my head the most ludicrous things, because 15 years ago, if you assume Microsoft might not be at the top of the compute chain and, you know, everyone would look at you like you were crazy, or digital equanimous.

equipment or, you know, you say Google now, oh, Google may not be at the top of the food chain in the future and everyone would look at you like, well, you're crazy. They'll be there forever. So I think by take, for me, taking away something that's out there has been a really interesting exercise. And then you find projects that actually, you know, I don't go out and look to say what happens, let's say, when VMware goes away or whatever. You sort of weigh, I have these views in my mind and then all of a sudden you start to see things that fill. the spaces and so that's kind of my you know some of my models and a lot of actually a lot of our Amplab investments have been based on look phenomenal entrepreneurs great ideas in their own right but then me also thinking okay well like if something isn't there maybe this is the thing that fills it is that part of the sort of ecosystem of Amplab dough too because it's it kind of it's hitting all those well Amplab certainly if you look at Spark, Mesos, and Takion, I should say, Takion, Spark, Mesos in that sort of architectural order,

it does create a full stack of, you know, kind of the next generation of big data kind of infrastructure. And that's really interesting. I mean, it's in memory, it's got like, blah, blah, you know, scale out, all this stuff that currently doesn't really exist commercially. And so, yeah, we, you know, kind of, it's part of that. H.W.A., do you think AMP Lab is a place like that because there's this kind of, you know, ecosystem of smart people working on these different parts of it. So, look, somebody else is already here, so I'm going to go here.

I mean, and, I mean, does it, is it this sort of, you know, ecosystem that exists there and you need to figure out where you fit within that ecosystem, or is it just, look, here's my interest, and I'm going to go after it? But from my perspective, I think the lab gives us a very great environment. And from the individual perspective, we don't have that much pressure like to, you know, in terms of paper publishing, those type of thing.

Right. And in the meantime, in the meantime, we can focus a thing for a pretty long time. And, you know, different people, they have their own interest. We have a pretty large lab, and we have a lot of people to bounce the idea.

We can talk with professors and talk to the students. Actually, the professor don't have an office, like they just sit in the same cube as we students there. So it's a very good environment. I think we can leverage. In the meantime, the lab also talk with the industry very regularly. Like we have the meter retreat thing like twice a year,

and we have like maybe 20 industry sponsors that can talk, communicate. I think those are very helpful. Peter, in your experience, what are some of the traits that you see in H.Y. and other entrepreneurs that correlate with success. One is just, look, a deep understanding of the space that they're in, and the passion to really go after it. We always want to see that. I think also it's the willingness to learn all the things that the entrepreneur doesn't yet know.

And so, look, I mean, H.Y. is a great example and a very positive example. He's very passionate. and he understands his area better than anyone on the planet. But there's a lot of other things that you need to do to learn how to run a company. And all the little things that you all might think are easy and stupid actually trip you up. And so what we want to look for is somebody who is going to be coachable because there is sort of a pattern. There's a blueprint on how to go do this. I mean, now that I've done this multiple times and pretty much take somebody through, here's what,

you do this week, next week, next month, in terms of building out of sales organization, in terms of hiring a product manager, in terms of hiring a CFO. Like there's all marketing, how do you do that? I mean, there's all these other pieces to building a company besides the technology. And so one of the attributes that we want to try to assess up front is, is the entrepreneur and, you know, Ph.D. student grad capable of being coached into these new areas? let's face it. I mean, a lot of folks don't want to be coached. It's like, look, I'm just going to go do my shit and, like, leave me alone. I don't want to learn about this other stuff. And you probably can't build a company if the only thing you do is focus on your technology. And you can, I mean, we're in it all the time. So that's what I would say. That you thought you never would have to deal with and that you either love or wish somebody else would do it.

I mean, I just, I think I just want to, I have a. goal and I want to I just want to achieve the goal. And along the way, there are different things. For example, like Peter mentioned, in terms of hiring, like process, there's many, many things in the hiring. Like at the beginning, we don't have a process. And then you have a process. And you modify the process, improve the process.

In the meantime, you have this and that condition. And we talk about this type of stuff. And some cases very, it's very new to me, right? But it's very interesting. It's also very interesting to me. Is there a class or something in school you wished you had learned that would have helped you? Do you wish you had taken more marketing classes or any for that matter? You know, like, I think school taught me really a lot, and that's the reason, you know, like, I can still do some stuff.

But in the meantime, like, I also think a lot of the things you should, like, you can learn a long of way as you go. if I have to put something, I'm an engineer background, I'm a CS background. So back to the school, maybe from the technical perspective, it's like, you know, be great if more students know how to build a production level like system, if they want to do a system level thing. Right. For other stuff, I think, I mean, you can learn along the way if you're willing to learn. Yeah. Right, I agree.

Yeah. Maybe you're not quite there yet, but your views on the other parts of the business, have they changed? So sales, marketing, you know, all that stuff. Are you just learning as you go? And are you a bottleneck still, or are you so so in tune to not being a bottleneck that that's not happening? I try not to be the bottleneck.

Yeah, but like the real, I mean, you got a lot of traffic. Right. So, I mean, Peter taught me a lot. Like, you have to prioritize. So in some sense, you're always about to knock since you have infinite things to do. But in the meantime, we got a great team. So I think our people are great and they can do things like move things forward fast

and without like my interaction there. So I'm pretty happy about that part. I think one of the most difficult transitions a person makes in their career, whether it's from academia or from just an individual. individual is from individual contributor to being a manager and leader. And when you're an individual contributor, you do everything, and you're the best at it, and you hack your code, and you do whatever. And when you become a manager, you have to do that through other people.

And that transition is non-intuitive and very, very difficult for, I mean, I had to go through it too. I'm not, I mean, yes, of course, I went to college and all that. I'm not at your level. But it didn't matter. I went from individual computer programmer to a manager, and it's incredibly difficult because most of your time, look, as an individual contributor, your time is spent writing code. And when you become a leader, which is the nicer way of saying manager, a leader,

most of your time really ought to be spent in hiring and coaching new people to come on. And that's completely counterintuitive. It's like, look, if I write more code this week, the project's going to move ahead. But if you don't hire anybody, you're going to continue to be the bottleneck. And I'm like that transition, and you know, you're still going through it. That transition is a real, it's difficult for everyone. And I think particularly difficult for people who have written the code and like, you know,

every line and every comment and all that stuff. It's particularly difficult to go through that. H.I. What hurdle did you recently put behind you? And what hurdle is next? I mean, so as an early startup, like one challenge is you always want to get the best people and takes time. And especially our project is growing pretty fast. So we all have a lot of inbound traffic. We always have trouble like in fully exploit that. So. And what's next? What's the hurdle that you're looking ahead? Is it still hiring? I mean, we just keep hiring.

Hiring. You know, I think, I mean, we spend a lot of time on this. I think the next hurdle for these guys is to figure out how do you commercialize the open source project that you have? It gets to what exactly is the business model? What's paid for? What's free? Is anything free? Like, coming up with that business model, such, because if you don't have that early, what starts, and you don't know what that is,

then potential customers and partners don't know what you're going to charge for in the future and all that. And it makes it, the longer you go on without having a model, the harder it is to roll back some of those things. Like in a year from now, so I'll go, well, we're going to charge for that. And everyone's like, well, we've been using it for free for a year. Like, you can't charge for that anymore. And so being able to lead with the crumbs along the way and saying, okay, here's what we're going to be doing, get ready for this, kind of gets the community you know, preconditioned to kind of

understand what the model is going to be. So I think right now that's what we're kind of talking about. H. My last question for you, you were back at that point where you were the bottleneck, you knew it, you had a choice to not go ahead and start Techian Nexus. Do you still think you made the right choice? Absolutely. Yeah. I think I made the rest of us.

Yeah. He'll let you know in five years. All right. Michael Reckland, H. W.A. Lee, Peter Levine. Thank you guys so much. Thanks. Thanks, everyone.