Subscribe
Share
Share
Embed
Hosted by Matt Klein, creator of Envoy and co-founder of bitdrift, Beyond the Noise goes inside the minds of the engineers, founders, and technical leaders defining the next era of app-based computing. Forget the buzzwords — this is where the people shaping modern systems share how they really build, debug, and scale.
[00:00:00]
Matt Klein: All right, folks. Welcome to another episode of Beyond The Noise: Signals, Stories, and Spicy Takes, the show where we dig into the stories of the people shaping the future of app-based computing. I'm your host, Matt Klein, co-founder and CTO of Bitdrift, as well as the founder of Envoy Proxy. Each episode, we'll talk with engineers, founders, and technical leaders who've transformed the way their companies build and understand what's happening inside their systems.
We'll dig into the challenges, the breakthroughs, the lessons learned, and wrap it all up with their hottest takes. So let's dive in. Today, I am thrilled to have with us Randy Shoup, who is a 30-year Silicon Valley veteran, most recently SVP of engineering at Thrive Market. He has also been VP of Engineering and Chief Architect at eBay, VP of Engineering at Stitch Fix, and a Director of Engineering at [00:01:00] Google Cloud.
He talks a lot, sometimes at conferences and sometimes in software podcasts. Welcome, Randy. Thank you for joining us.
Randy Shoup: Matt, it's so much fun. I'm so excited for our conversation.
Matt Klein: Yeah, great. Um, you, you've obviously had a, a long and storied career, so I'm, I'm excited to dig in. How we usually start is, especially for someone who's been in the, in the industry as long as you have and, uh, worked at some really fantastic companies-
Randy Shoup: Yeah
Matt Klein: would love to just hear how you got into tech. I mean, like, tell us your, tell us your origin story.
Randy Shoup: Oh, man. Uh, this is gonna be fun, and you told me I had a little time for this, so I'm actually gonna spread out and tell you the whole story.
Matt Klein: Pl-pl-pl-please do.
Randy Shoup: Okay. So the original, the earliest part of the story, it, it starts way back when I was born.
I'm not even joking. So my parents met at what's now CMU. My dad got his double E degree, and when-- then I was born, and he got his PhD in 1970, uh, in computer science.
Matt Klein: Oh, wow.
Randy Shoup: Moved out to the-
Matt Klein: Wow, okay, yeah.
Randy Shoup: Yeah, 1970 computer science.
Matt Klein: Wow.
Randy Shoup: One of the very first [00:02:00] computer science PhD programs in the country.
Matt Klein: Yeah.
Randy Shoup: Uh, his PhD thesis, by the way, proposed FPGAs
Matt Klein: Really?
Randy Shoup: Okay.
Matt Klein: Wow. All right. Yeah.
Randy Shoup: Uh, no pressure, no pressure-
Matt Klein: Yeah
Randy Shoup: ... next generation.
Matt Klein: Yeah.
Randy Shoup: Um, uh, so yeah, he, he proposed, uh, he called it, uh, reconfigurable cellular automata, I believe. Uh, reconfigurable computing, but essentially, instead of building chips in the von Neumann architecture that's, like, very structured, what if you built a more generic chip?
And the, the initial proposal in the, uh, PhD thesis, which I have in paper behind me typed by my mom, uh, uh, is about, uh, the, the proposal is essentially, or, you know, the, his research was... what if we used reconfigurable silicon to design new silicon, right? So, like, FPGA is essentially, like, y- the programmable gate arrays, field programmable.
So, like, you can ins- uh, without [00:03:00] stamping the chip out again-
Matt Klein: Yeah, sure
Randy Shoup: ... you can reconfigure the, the circuits essentially, and you use that to design new circuits. So anyway, so he, he did that. Um, and, uh, we moved out to the Bay Area right after he got his PhD, freshly minted PhD. He joined a company called Berkeley Computer Corporation, which is only known because it collapsed at exactly the perfect time.
What I mean by that is that it collapsed in 1971. Oh, they were building a, um, multi-user, uh, time-sharing system, I believe. Uh, they collapsed at exactly the same time as Xerox was opening up the Palo Alto Research Center- Yeah, right ... or PARC. Um, and, uh, my dad was in the first six or eight researchers that they, um, that Bob Taylor brought into the lab.
Um, and so other people,
Matt Klein: um- This is amazing. I, I did not know any of this. Wow, you, you have a, you have a pretty amazing story. This is great.
Randy Shoup: That's right.
Matt Klein: Keep going.
Randy Shoup: Yeah. So, uh, you can Google Richard G. Shoup, uh, computer scientist.
Matt Klein: Right.
Randy Shoup: Uh, he's got his, he's [00:04:00] got his little, um, uh, Wikipedia page. Uh, unfortunately, he's not with us.
He passed away in 2015. Um, but there are still a bunch of videos where he talked about his orig- his early stuff, which I'll tell you in a moment. Uh, okay, so he's an early guy at, uh, Xerox PARC. He joined just before Alan Kay. Chuck Thacker, Butler Lampson, these are names that I think the current generations don't know, but, but like those guys invented Alto, which I can tell you about, microcode.
Matt Klein: Yeah, for sure.
Randy Shoup: Um, and so just taking a step, a, a broader step back, what did Xerox PARC invent? And then I'll tell you what my dad did. Overall, they invented the graphical user interface. They invented the laser printer. They invented object-oriented programming with Smalltalk from Alan Kay and his research group.
They invented-
Matt Klein: Mouse
also? Is that right? Did,
didn't-
Randy Shoup: Pardon me? No, not the mouse.
Matt Klein: No,
not the mouse. Okay.
Randy Shoup: No, they used the mouse. It's, it's, uh, uh, the mouse was invented at Stanford, uh, SRI, Stanford Research Institute-
Matt Klein: Okay ...
Randy Shoup: a nearby, like very nearby-
Matt Klein: [00:05:00] Interesting ...
Randy Shoup: thing. Uh- Yeah ... my grandmother was actually a secretary there, but that's a small thing.
So Unconnected to PARC, but a deep inspiration to PARC was the 1968 mother of all demos by Douglas Engelbart. I, I think I have his name, his last name correctly. It was in San Francisco, and in 1968, he demoed like hypertext and all sorts of like amazing stuff, and he showed off the mouse.
Okay, so, uh, a bunch of people used the mouse at the time, um, but yeah, but PARC is known for having driven most if not all of... Well, most, not all of their interfaces off of the mouse. Everything was with a pointing device. Uh, okay, so let me finish. Graphical user interface, uh, laser printer, object-oriented programming, Ethernet.
Okay. Also the touch screen, uh, also VLSI logic for chips. Um, and I feel like I might be forgetting one. My dad did none of those things. Um, [00:06:00] he did he did color graphics. So when all the researchers, uh, rocked up, Bob Taylor had already had a storied history at ARPA. He built the, he built the internet.
Um, and he, not himself, as he would say, and others would say, he was not technical himself, but he was like a absolute world-class bringer together of hotshot technical people and organizer and chal- channeler of them into like doing great stuff. And he did that at ARPA, and then it brought him into, um, run part of lab, at Xerox PARC.
And he knew all these hotshot, young, freshly minted PhDs, from the ARPA work and including my dad. So what he, his MO was he said, "All right, you're cool," people who bro- who, who, uh, arrived. Um, "You have a year to figure out what you wanna do." And my dad was like, "Okay, cool. What I'm gonna do is color graphics."
And, uh, the Alto, which is the predecessor of the [00:07:00] Mac or, you know, the Apple Lisa and the Xerox Star, Apple Lisa, Mac, that whole thread, um, was everybody else in the lab did that. Um, but my dad's pretty stubborn. And so when Bob Taylor told him he could figure out, take a year to do whatever he wanted, my dad said, "Okay, I'm gonna take you at your word."
And so what he built was a system called Superpaint. You can Google that. In 1973, April 1973, before his 30th birthday, he produced the world's first I mean, people had done paint programs before a little bit, and people had built something called a frame buffer, which I'll tell you about in a moment. And he combined those two things, a paint program with a frame buffer, and the, and the first thing that could take digital video in at frame rate, 30 frames a second, uh, digitize it, uh, manipulate it digitally in a paint program essentially, and then send it out frame rate, 30 rate- frames a second out the other [00:08:00] end.
Uh, what's a frame buffer? It's you take a frame, like an image, and you put it in a buffer, uh, and you manipula- you, you do that to manipulate it digitally.
Matt Klein: Yeah. So, so this would be like the precursor for when you were doing live TV and you wanted to overlay some graphic on top of it? I mean, is that, is that basically what we're talking about?
Randy Shoup: That is a, a, a initial-- the, one of the initial use cases for this kind of technology was exactly that.
Matt Klein: Okay.
Randy Shoup: Yeah. Um, and I'm jumping around, but, like, I can tell you, I'll fill it in in the middle. But, like, after PARC, he, just like everybody, he left and founded his own company. Um, his company was called Aurora Imaging Systems, and that's exactly what they did.
They built graphics hardware mostly used in television-
Matt Klein: Right
Randy Shoup: ... to do exactly that, weather graphics, uh, overlays. When he was still at PARC, somewhere in the 1977, '78 frame, there was the Pioneer Venus mission, like a f- a probe that we sent to Venus.
And, uh, Ames Research Center, which is in [00:09:00] Silicon Valley nearby, uh, in, in, uh, Mountain View, was the place where they were controlling the probe.
Matt Klein: Right.
Randy Shoup: And they wanted some way to describe t- And it didn't have video. Like the, you know, the probe was, like, 1970s technology. Like, they didn't, it didn't have any cameras essentially, um, or cameras that they could send stuff back in real time.
And the, and NASA wanted some way to be able to describe to regular humans what in the world they were even doing. Uh, and so they, uh, put, they... The system that I'm halfway telling you about they, they, like, took it apart and moved it over to Ames Research, re-set it up, and, um, they did live television with all that stuff.
Matt Klein: Oh, wow.
Randy Shoup: So, uh, they did it for the Pioneer Venus mission. They did that for the Pioneer Saturn mission, which was the year after that. So again, it's like '77, '78, '78, '79, something like that. Um, anyway, and so that was a little bit of his calling card when he left and founded his own company to, like, commercialize this kind of technology and use for television weather [00:10:00] and stuff.
Matt Klein: Yeah, I, I mean, I think for many people who are probably listening, it, it's hard to even wrap their heads around how much technology has changed. I mean, I mean, we're, we're talking, like, 50 years ago.
Randy Shoup: Yeah.
Matt Klein: Um, and I don't know. I, I, I think it's fun to actually think about that occasionally, uh- Oh, totally.
Yeah ... in, in terms of how much has happened over the past 50 years. It's pretty incredible.
Randy Shoup: Oh my gosh. Yeah. I did another podcast that was two hours long, don't worry, we won't do that here, on all this, like, 50 years in Silicon Valley-
Matt Klein: Yeah
Randy Shoup: ... learning from what I'm talking about.
Matt Klein: Yeah.
Randy Shoup: And, you know- I'm, I'm in my 50s, so, like, there is the concept of, like, the younger generation doesn't really know where y'all came from.
Uh, I don't mean you, but like many of the listeners.
Matt Klein: Sure. Yeah.
Randy Shoup: And, like, that's not even wrong. Like, uh, same with me when I was the same age. But I don't think we fully appreciate how outrageously forward-looking all the stuff in 1973 was. I mean, not even just my dad's work, although it was, that was. But, like, the, all the things I mentioned, like no one had ever-
Matt Klein: Yeah
Randy Shoup: done [00:11:00] any of those things ever. And in retrospect, it's like, well, it's so obvious you should have pointing devices, and it's so obvious things should be bitmapped.
Matt Klein: Sure. Yeah.
Randy Shoup: And it's so obvious that you should have... Like, no, none of that was obvious. These were all, like, absolute cutting-edge research, like, related things, like analogous to, you know, what the frontier AI labs are doing.
Like, it's a pushing, pushing, pushing the boundaries-
Matt Klein: Yeah
Randy Shoup: ... of what humans had ever even done. Uh, anyway, to finish the thought about Superpaint, so again, combining a frame buffer, which is a place where you could store an image digitally, uh, with a thing that could manipulate it digitally. And so the way that the system worked...
And his idea was artists and regular people could do this. And so the metaphor, the paint program metaphor is super familiar to everybody, but again, this was invented in 1973-
Matt Klein: Yeah
Randy Shoup: ... by him and his collaborators. So there's a pa- a palette on one side, like color, choose your color, choose your brush. Like think Mac Paint or, uh, MS Paint[00:12:00]
or like-
Matt Klein: Yep
Randy Shoup: ... any paint program you've ever used. Draw the lines of, uh, origin lines back, and it was this thing that they in- they invented. Um, and then the other, uh, thing is the canvas, and you choose your brush and your color, and you go and you draw stuff on the canvas. And how does that even work? Uh, the magic is they have this frame buffer and, um, I can describe the hardware of it if it's at all interesting to people, but we don't have to go there.
It's not about me. Um, but anyway, conceptually, there was this big, uh, thing that could store an entire image, which was a lot of money back... of memory back in, uh, back in 1973, and manipulate it. Anyway, you know, I was inspired by that kind of stuff. Uh, but, like, that's kinda fancy.
I mean, like, he got an Emmy Award and an Academy Award and, you know, uh, he's kind of a rock star. Uh-
Matt Klein: But when you were a kid, I mean, uh, you, you must have been exposed to this. I mean, you probably went-
Randy Shoup: Yeah. So-
Matt Klein: ... went, went to his lab and, I mean, I would imagine that, that [00:13:00] you were exposed to a lot of very early technology.
Randy Shoup: Yeah. Yeah, totally. And, and, and as with, um, as with most people, you don't fully realize how different or the same your childhood experience... Like, your childhood experience is normal. Doesn't everybody, doesn't everybody have a dad where, like, you beg to go to work with him on the weekends? Like, apparently not.
Um, but my brother and I absolutely begged, begged, begged my dad to go with him on the weekends-
Matt Klein: Yeah
Randy Shoup: ... and he would be very happy to go and work. Um, and in retrospect, it was honestly the, his work which didn't figure in the main line of PARC. Again, they were building the, the, this, personal computer essentially for business.
Um, so his was a sidelight from PARC's perspective. But it was literally the only thing that they developed that would be interesting to, like, a six or eight-year-old Like, I mean, it was neat to be able to ch-change bold, make fonts bold and like-
Matt Klein: Yeah, sure
Randy Shoup: ... italic and stuff. I mean, that was like, that was cool.
But like, that's interesting for a couple of minutes, like if you don't have [00:14:00] something actual to accomplish. But like, hey man, drawing spaceships in color and then printing them out on a color laser printer, holy cow, that was pretty cool. Uh, yeah. So, uh, so that, that gave me the bug to do... to love math and computers.
All right, so let's fast-forward. We grew up in the Silicon Valley, but it wasn't even called Silicon Valley then. I went to Stanford, so local college. Uh, maybe you've heard of it. And, uh, but my mainline thing that I wanted to do with my life was be an international lawyer. So this is the end. I graduated from high school in '86.
My college experience was '86 to '90, and so height of the Cold War, US and the Soviet Union have tens of thousands of nuclear weapons pointed at each other. And like that was a thing that actually mattered to me to try to like get rid of. Uh, thankfully it solved itself without my help, but, uh, at least for a while.
We'll see how it goes. Uh, we'll, we'll, we'll see how well that, that statement ages. But, that was the thing I wanted to do. And like, look, [00:15:00] Dad's a freaking rock star, and like, as much as I was loved and was inspired by what he did, he never made me feel like I couldn't do it in any way.
Matt Klein: Yeah.
Randy Shoup: Like, he never put any, any expectations on me or my brother really in this way.
But like- It was-- He was so, such a l-luminary that like, "You know what? I'm gonna do something else." And like not scared away, but just like my talents and, and my, um, passions really at the time were in a, were in a different direction. Okay, so I go to Stanford, I start studying political science with a concentration in international relations, particularly US-Soviet East-West relations.
I go overseas. I lived in West Berlin. I say that because there was a wall around it. Um, I lived in Kraków, which was at that time on the Iron, behind the Iron Curtain in Poland. I actually had done, for reasons, an internship at Intel doing software after sophomore summer, junior summer, and then working through senior year.
So like I kind of had, had a foot in both camps. Oh, and then the other thing, uh-
Matt Klein: So when
you were in college though, I, I mean, h- h- how, [00:16:00] like were you minoring in computer science or like how,
how were you-
Randy Shoup: That, I think that's exactly where I was going.
Matt Klein: Okay.
Randy Shoup: Yeah, that's exactly where I was going. So thankfully, thankfully...
So Stanford does have minors now, but thankfully they didn't then. Uh, so the only way for me to study math and computers for serious was to double major. So I actually have two- Um,
Matt Klein: okay. Got it.
Randy Shoup: I have a, a political science degree from Stanford, and I have a mathematical and computational science degree.
So that's currently has evolved into the data science major. We didn't have that word then, but it's like applied math, statistics-
Matt Klein: Yep
Randy Shoup: ... what we used to call operations research, which is like optimization theory, queuing, uh, network optimization, Dijkstra's algorithm, uh, and then CS. And, uh, so I have those two degrees, and I was doing, uh, I was doing the summer internships at Intel and software, for fun.
Like it was fun and it paid well. Um, and so then when I graduated, I was planning on going to grad school in [00:17:00] international relations, which I did start. I'll g- tell you that in a sec. But I didn't wanna do it right away, so I worked for two years as Oracle, at Oracle as a software engineer. Loved it. And then much to the chagrin and surprise of all my buddies, I'm like, "Okay, now I'm gonna take the GRE and the LSAT and like go do my regular career."
And they're like, "What?" I'm like, "Guys, listen, man. When I tell you, I, I've been telling you that for years, like this is, this is a, a side gig for fun, and then I'm gonna get back to, you know, reducing nuclear weapons
in the world."
Matt Klein: And, and, a- and then, sorry, just from a timeframe- Oh, yeah ... this is in the late '80s or the-
Randy Shoup: Yeah,
yeah.
Sorry.
Right
Matt Klein: ... or, or the early '90s?
Randy Shoup: Great, great. Yeah.
Matt Klein: Just so that people can put it into context.
Randy Shoup: Yeah, yeah. Sorry. So I graduated from high school in '86. Stanford was '86 to '90. So I graduated in 1990, and I worked at Oracle in, from 1990 to '92, and in '92 I applied to, um, grad school.
Matt Klein: Right.
Randy Shoup: Um-
Matt Klein: I, I, I mean, but, but for context, I mean, this is the beginning of the boom [00:18:00] times for sure, right?
Randy Shoup: Yeah.
Matt Klein: Absolutely. So, so you were, you were making a conscious decision, I guess, to, to step away from that or, or I guess in that era it was a bit before the internet boom kicked off-
Randy Shoup: Yeah, yeah
Matt Klein: ... maybe a little after like Microsoft was ascendant. I don't know. I'm, I'm just trying to put it in context.
Randy Shoup: Yeah, yeah.
Well, thank, thank you for slowing me down. Yeah, 'cause again, I lived it so it doesn't seem interesting to me or like worthy of comment. But yes, great. So let me take a step back. Uh, the IBM PC is 1982 The Mac is 1984. Uh, so we actually used macs at school between '86 and '90.
Matt Klein: Yeah.
Randy Shoup: Type all our papers on there, whatever, printed things out on your laser writer.
Um, so per- so it's the personal computing revolution is the time that we're talking about. The internet had existed since the '60s and '70s, but the web didn't, and the web was invented in '89, but like it actually didn't get commercialized-
Matt Klein: Yep
Randy Shoup: ... [00:19:00] or like widely known until '94, '95.
Matt Klein: Yeah. Right.
Randy Shoup: And that's-- And so again, we're, this is before that time.
Yes. So, uh, 1990, '91, '92 is personal computers are ubiqu- e- expensive but ubiquitous. Like more or less every family had a personal computer, and definitely everybody in business was using computers commonly to do their work.
Matt Klein: Yes.
Randy Shoup: Uh, but there was no internet. Uh, people had, there were, uh, there was email, there were Usenet groups, uh, there were-
Matt Klein: Dial-up modems.
Randy Shoup: RSB
Matt Klein: back then.
Yep.
Randy Shoup: Uh, there was Gopher.
Matt Klein: Yeah.
Randy Shoup: Uh, but no web. Uh, web did not, web did not exist. Um, so okay. So now, I am, now it's '92, and I start a joint program in law and international relations. Um, and for reasons, uh, I went to only one year of that program, first year of law school. Um, I had a summer [00:20:00] internship at a law firm, which should've been the perfect job 'cause it was like, uh, IP work on Sand Hill Road in, uh, Silicon Valley.
Uh, fantastic clients, Intel, Motorola, Qualcomm. I was the only s- intern they'd ever had. It was a very, very small office of a big New York law firm called Weil, Gotshal & Manges. Anyway, so like it should've been the perfect job, tick, tick, tick, tick, tick, and I hated it. And I hated it because getting people patents is painful if you wanna be an inventor 'cause the, the patent lawyer, like it's a real job that they have, but like they're a scribe.
They, like the inventor's up on the whiteboard describing her amazing invention, and the lawyer writes it down and does bureaucratic infighting with the patent office to get the patent.
Matt Klein: Yeah. I, I have personally filed over 50 patents, and it, it is sh- shocking to me how that process actually works. But that's probably a subject for a whole other podcast, so we can-
Randy Shoup: Totally
Matt Klein: we can gloss over that.
Randy Shoup: Yeah.
I-
Matt Klein: Anyway, k- keep going.
Randy Shoup: Yeah, yeah. I, I, I don't, I think I have, uh, probably have, [00:21:00] I have large number of teens, maybe low number 20s.
Matt Klein: Yeah.
Randy Shoup: Uh, yeah. And it's like- There's, there's absolutely nothing technical about that process. Like you do some invention, which is actually interesting potentially, um, but every place where that, where you get the patent, like there's nothing technical and interesting about that at all if you're-
Matt Klein: Well,
I, I mean, like we said that we weren't going to talk about this, but I guess my one comment is that it's also not necessarily invention.
I mean, a lot of patents, at, at least these days are-
Randy Shoup: Oh, yeah
Matt Klein: ... I mean, they're about protecting business interests and a bunch of other things.
Randy Shoup: Yeah, yeah, for sure.
Matt Klein: And it's, y- y- you know. A- anyway, sorry. Keep going. Yeah. The,
Randy Shoup: uh- Yeah ... yeah, I agree. In order to get a patent, it needs to be useful, novel, and non-obvious.
And the novel and non-obvious filters have a very low bar.
Matt Klein: Right.
Randy Shoup: A shockingly low bar.
Matt Klein: Agreed.
Randy Shoup: Uh-
Matt Klein: Okay. All
Randy Shoup: right ... wait, you mean I could make that red? Mm-hmm. And that's patentable?
Matt Klein: Yes.
Randy Shoup: Okay, sure.
Matt Klein: Yeah.
Randy Shoup: Uh, anyway, all right. So, uh, I'd been saying to [00:22:00] myself and everybody else I wanted to be an international lawyer.
When it came right down to brass tacks, combining the interests of what I thought would be interesting, law and technology, I, uh, was more interested in the technology. Anyway, long story short, I, um, went back to Oracle and continued working there for another, five years. So like 19-1990 to 1997, worked as an individual contributor at Oracle.
Matt Klein: What were you working on?
Randy Shoup: Yeah, it's fun. We worked on a graphical, um, query tool. It was called Oracle Data Browser.
Matt Klein: For-
Randy Shoup: And browser, like not web browser, like there w- there was no web browser in 1990. Uh, so-
Matt Klein: Oh,
wow. So it was a, it was a query frontend, like a visual query frontend for-
Randy Shoup: Exactly
Matt Klein: ... the database, basically.
Randy Shoup: Exactly.
Matt Klein: Is that... Yeah. Yeah. Okay.
Randy Shoup: Ad- ad hoc queries. So you would construct your query, uh, graphically. So imagine-
Matt Klein: Right
Randy Shoup: ... an entity relationship diagram, and like-
Matt Klein: Yeah,
totally. Sure, yeah ...
Randy Shoup: draw an entity, entity relationship diagram, essentially. Visually represent the [00:23:00] tables, the columns, the links, uh, exactly how you would in an ER diagram, and then we had a graphical representation for, uh, filters.
Matt Klein: Yeah. I, I, I'm like having some flashbacks of Microsoft Access where-
Randy Shoup: Yeah
Matt Klein: ... where, like I, I feel like it had similar tooling-
Randy Shoup: Yeah, yeah ...
Matt Klein: probably i- in the same era.
Randy Shoup: Yeah, yeah. For sure, for sure.
Matt Klein: You know? Right.
Randy Shoup: We-- I guarantee you, I, I mean, we probably have patents, but like I guarantee you-
Matt Klein: Yeah
Randy Shoup: ... we didn't invent any of that concept.
We put it together and it was pretty good.
Matt Klein: Sure,
yeah.
Randy Shoup: Um, but there were, there were a couple of other con- um, products at the time that did something similar. Cognos, Brio-
Matt Klein: Right ...
Randy Shoup: uh, InfoObject, something like that. I don't know. Anyway, it doesn't matter. Okay. So worked there for seven years as an individual contributor.
Um, and then, at a, at a New Year's party, uh, I met my wife, now ex-wife, at law school. She actually became an environmental lawyer and worked for a law firm. This is relevant [00:24:00] because in, I don't know, '97 my guess was probably '97. We were at a New Year's party for people that she knew through the law firm, and, uh, one of them was another Stanford guy who I-- who was a year ahead of me.
I didn't know him at Stanford. Um, and we were chatting. He's like, "Oh, I do technology, so do you," blah, blah, blah. Anyway, so, uh, I went and chatted with him, uh, later had lunch, and long story short, I ended up joining this security company, leaving Oracle to join this security company called Tumbleweed, which was co-founded by a French guy and a, and a American guy that had met in grad school at Stanford.
Um, and Tumbleweed did... I was there for six years. Tumbleweed did, um, various aspects of like, um Email scanning into and out of businesses, secure delivery of confidential documents. So like anytime you receive, anything from your bank that said, "Hey, you-- hey, uh, hey Matt, you have a new [00:25:00] statement. Click here-
Matt Klein: Right
Randy Shoup: which will authenticate you." Um, you wanna talk about patents and, uh, things, like Tumbleweed holds the patent on that. Not even joking. Uh, n- holds the patent on email HTTP combination. Like, is that obvious? Well,
I'm, I'm not at liberty to say.
Matt Klein: Right.
Randy Shoup: Uh, but, but anyway, uh, that-- they held that patent. And, uh, anyway, long story short, we had, uh, a bunch of clients, a, a bunch of customers that were banks and financial institutions, again, wanting to securely deliver trade confirmations, financial statements, et cetera.
Um, we actually, interestingly, this never went anywhere in the longer term, but we had, we had customers that were most of the post offices on the planet because at that time, '97, '98, people were-- post offices were thinking, "Hey, mail delivery is going to go electronic-
Matt Klein: Yeah
Randy Shoup: ... and like we as post offices should be part of that."
Matt Klein: What's, what's [00:26:00] super interesting to me is I, I almost wonder if at the time that was almost ahead of its time just in the way of-- I, I feel like just like thinking about my own experience back then, I don't think anyone back then thought about like phishing and fraud and like a lot of the computer security stuff that people think about today.
So-- and that might actually be wrong, so I'm actually trying to ask you, like do you, do you think-- because like that product today is so obvious just with the-
Randy Shoup: Yeah
Matt Klein: ... amount of crime and cyber crime and, uh, you know-
Randy Shoup: Yeah
Matt Klein: ... everything that happens. But I just feel like back then, again, maybe I'm wrong, but I just don't feel like there was quite the, um, there was, there was quite as much maybe at least knowledge of that-
Randy Shoup: Yeah
Matt Klein: type of crime or there just weren't as many people. So it's like what, what was it like to work in an area where-
Randy Shoup: Yeah
Matt Klein: ... I, I feel like that's the very origin of that type of technology from when it was useful.
Randy Shoup: Yeah. [00:27:00] Yeah. So, um, I think y- so you're absolutely right that looked at from a consumer, regular human lens, uh, nobody thought, nobody thought the first thing about computer security.
And also businesses... there, there, there was malware, there was phishing.
Matt Klein: Sure. Yeah.
Randy Shoup: Uh, there were, uh, hacking attempts, like all those things, all those things existed, and I won't-- I don't want to leave anybody with the impression that Tumbleweed was somehow forward-thinking and thought about these issues that nobody else did.
But it is true that in the broader zeitgeist, we had to explain to some, like I have to explain-
Matt Klein: Yeah
Randy Shoup: ... to my family, like, "Hey, you know, like when you use the web, like when you click on stuff-
Matt Klein: Right ...
Randy Shoup: it, you know, look for that S thing and-
Matt Klein: Yeah
Randy Shoup: ... uh, blah, blahdy, blah." So it... I mean, now everybody for some value of everybody has some familiarity, unfortunately, with, you know, data breaches [00:28:00] and-
Matt Klein: Yes
Randy Shoup: identity theft, blah, blah, blah. And those problems absolutely existed then, were widely known within our industry, and particularly within governments and financial institutions. Like we didn't go and go, "Hey, let me tell you about this thing called malware." Like that, that-
Matt Klein: Right. Yeah,
of course ...
Randy Shoup: there were a- But the typical way those were solved were at the edges.
So like you would have antivirus software on your PC, right?
Matt Klein: I remember.
Randy Shoup: So you had Norton-
Matt Klein: Yes.
Right. Yeah
Randy Shoup: ... or, um, there's a bunch of them, um, McAfee, Norton, , and those thing. And so the... I don't know if it was novel, but the thing that we sold that was interesting was instead of having antivirus on every PC in your business, which you would have to have, we would also do antivirus checking into and out of the enterprise.
So again, it used to be personal computer by personal computer-
Matt Klein: Right
Randy Shoup: ... by personal computer, you would install this thing. But like, it's already too late. They've already breached the walls.
Matt Klein: Yeah.
Well, I, I, I mean, [00:29:00] that, that part seems to me like it's actually forward-thinking because-
Randy Shoup: Yeah
Matt Klein: ... that obviously is the way that everything works today.
I mean-
Randy Shoup: Yeah
Matt Klein: ... no one installs antivirus software anymore.
Randy Shoup: Yeah, exactly.
Matt Klein: Like- Everything is done at, at, at the edge-
Randy Shoup: Yeah
Matt Klein: ... and it's all, it's all done
before it even gets to your computer.
Randy Shoup: Yeah. Yeah. And, again, I don't wanna give the, I don't wanna give the mistaken impression that we were the only ones thinking that way, but I-- we were one of the few companies that was thinking that way and selling software too.
Matt Klein: Yeah.
Okay. So like- Makes sense. Yeah ...
Randy Shoup: antivirus checking, malware, uh, data exfiltration, blah, blah, blah, all that stuff at the edge, like we did that. And there were other companies that did the same thing.
Matt Klein: It's also too, I, I mean, that again, you know, talking about dating ourselves, but I mean, p- most people today don't understand that like software used to come from places other than app stores, right?
Which are, which are very vetted places, you know? I mean, and, and-
Randy Shoup: Yeah. Yes, right. Well-
Matt Klein: I think, I, I think people probably, some people don't even have that context is that-
Randy Shoup: Yeah
Matt Klein: ... there, there are-- Like software used to be [00:30:00] delivered by anyone, really, anywhere.
Randy Shoup: Yeah. Well, um, yes. So when, when I worked at Oracle and even many years after that, documentation was shipped as physical books-
Matt Klein: Sure
Randy Shoup: through the mail.
Matt Klein: Yeah.
Randy Shoup: And software was not shipped over the internet. What the fuck is the internet? Excuse my language.
Matt Klein: Yep.
Randy Shoup: That doesn't exist. It came as CDs.
Matt Klein: Yeah.
Randy Shoup: And like over time, it was more and more and more and more CDs.
Matt Klein: Yes.
Randy Shoup: Uh, you know, um, So yeah, so, yeah, so there we are.
I'll back up a little bit, but my time at Tumbleweed was '97 to 2003, so that is the boom and bust.
Matt Klein: Right.
Randy Shoup: So the, uh, the dot-com boom was 2000, and the dot-com bust was, like, right after that. It was like, you know, nine- 1999... No, sorry. Boom, '98, '99, early 2000. Bust, mid-2000-
Matt Klein: Yeah[00:31:00]
Randy Shoup: and forward.
Matt Klein: A- a- and we'll probably get there, but I'm gonna ask you anyway, having, s- since we're talking about that, having lived very directly through that time, h-h-how do you feel about what we're going through r-right now?
Randy Shoup: Uh, yeah. Uh-
Matt Klein: With the AI boom. I, I, I mean, it's like, do you feel it's similar? I guess, like, t-tell us from your experience, what is the same, what is different a-about-
Randy Shoup: Yeah
Matt Klein: what's going on right now?
Randy Shoup: Yeah. It's, you know, history doesn't repeat, but it rhymes.
Matt Klein: Sure.
Randy Shoup: Uh, the, what's it, the names are changed, but the, the song's the same, whatever. The song's the same. We have a... versus then, it's compressed time cycles. But, the the dot-com boom was real things, like the internet exists and is really valuable. And also in the '98, '99, 2000, some companies, many companies got way over their skis and over-invested in a bunch of things that didn't immediately pan out. [00:32:00] A thing that, uh, you might know, but, uh, what might be a surprise to, the younger listeners is amazon.com.
So Amazon and eBay both started in '94, '95. Uh, same with Yahoo.
Matt Klein: Yeah.
Randy Shoup: Uh, Google was actually several years later, but like many of the big internet properties started then. Okay. Amazon had a flat to declining revenue business for many years. Many, many years. Uh, and they were like, exactly like one of these AI companies.
We're like, "Are you guys gonna ever, I don't know, make any money? Uh, uh, legit." Uh, and now, like that sounds, that's crazy talk.
Matt Klein: Sure.
Randy Shoup: Like, what do you mean?
Matt Klein: Yeah.
Randy Shoup: Jeff Bezos is one of the most... the richest people on the planet. Um, uh, but that took a long time, and we don't have to go here in specific, but Amazon took...
Amazon had a massive painful monolith in 2000, eBay similarly. And Amazon took, despite [00:33:00] the fact that they were losing money, uh, Amazon took 2000 to 2005 to rebuild their entire-
Matt Klein: Yeah, sure
Randy Shoup: ... internal infrastructure around what we would now call microservices. They simply called it service-oriented architecture.
Happy to talk on many hours, uh-
Matt Klein: Sure. Yeah
Randy Shoup: ... about that, but probably not the point. What I will say though is what people don't appreciate now is how much Bezos had to like fight Wall Street to like, "No, I'm actually not-- I'm gonna keep growing, uh, revenue. I'm not gonna be profitable, and not for several years, but hold on."
And they're like, "Wait, I want my money." And like, "No, no, hold on." Uh, and if, if the OpenAIs and the Anthropics of the world play out like Amazon, like wow, check, you did it well. Um, but some of them, many of the, many of these companies, and I don't know if it's either of those two, I don't know that, many of these companies are gonna play out like pets.com.
Matt Klein: Yeah.
Sure.
Randy Shoup: Uh, or um- Webvan.
Matt Klein: Yeah.
Randy Shoup: Uh, and if you're young, you don't [00:34:00] know any of those names. Uh, but Pets.com was like, "Hey, we sh- we should... People buy things on the internet, we should ship pet food." And, like, that's not even wrong. You could do that right now, but it was way early because the shipping cost was so high, and pet-
Matt Klein: Yes
Randy Shoup: food is really ex- is really, like, low, low monetary value, high shipping cost. So, like, that wasn't the best
place to start.
Matt Klein: I mean, not only can you do it now, but you can do it from a variety of companies, whether it be Chewy.com-
Randy Shoup: Oh, yeah
Matt Klein: ... or Amazon or, like, I mean, there, there's-
Randy Shoup: Yeah ... it's just- And, and, and W-
Matt Klein: it's a different time
Randy Shoup: ... and Webvan-
Matt Klein: Yeah
Randy Shoup: ... and Webvan in, I'm making this up, but it's probably not wrong, '98, '99, 2000, was grocery delivery to your house.
Matt Klein: Yeah.
Randy Shoup: And, like, that was... They got way over their skis, but, like, everybody do- does that now.
Matt Klein: Sure, yeah.
Randy Shoup: Like, what, what... That was, that's so, like, obvious that that needs to exist, DoorDash and Uber Eats and-
Matt Klein: Yeah
Randy Shoup: Instacart and, you know, like, so Webvan is the predecessor to all those.
Matt Klein: Yeah.
Randy Shoup: And it was just that they were, that the economics didn't work in [00:35:00] that moment, but the idea was absolutely solid-
Matt Klein: Absolutely
Randy Shoup: ... and-
Matt Klein: Yep ...
Randy Shoup: uh, absolutely correct and far-seeing. Um-
Matt Klein: Okay, sorry. So, so to bring you back to your story.
Randy Shoup: Thank you.
Matt Klein: So, um, boom, bust, you were at the security company, and then-
Randy Shoup: Yeah
Matt Klein: ... and then what
happened after that?
Randy Shoup: Yeah. So, uh, so we went public in, uh, so Tumbleweed went public in, on August 6th, 1999. And I remember that because my then wife and I moved to our house in, from one house in San Francisco to another house in San Francisco on August 7th, 1999.
So on the day that the, uh, French co-founder, the French VP of engineering, and a bunch of other people became millionaires or 10 millionaires, uh, the next day they drove up to San Francisco to help me move. And like that says everything about what's great about working with peop- do you know what I mean?
Like we're all like-
Matt Klein: Yeah, for sure
Randy Shoup: ... these, uh, these people are absolute, uh, absolute rock stars and deserve all their, [00:36:00] all their wealth and success, and they're just regular people to help, that help each other out. Uh, so, uh, that was, that was... That's why, that's why I remember that. Uh, okay, so then, uh, Tumbleweed stock got up to 102, and then a month later it went to three.
So yeah. Uh, and if-
Matt Klein: Ho-hopefully that was after the lockup or was it prior to the lockup?
Randy Shoup: Uh, the lockup expiration was... I could have sold and unfortunately I sold on the way-
Matt Klein: yeah
Randy Shoup: ... on the way down. Um, and so let's just say I learned a lot about loss carry forward.
Matt Klein: Yeah.
Randy Shoup: You're already laughing.
Yeah. Um, loss carry forward, baby. Um, that-
Matt Klein: Don't
we all? Yep.
Randy Shoup: Yeah. That, that, uh, that dot-com bust experience, I'm not even joking, paid for 10 years of future capital gains for me.
Matt Klein: Yeah, I'm sure. Yep.
Randy Shoup: Just paying it back, paying it back-
Matt Klein: Yeah
Randy Shoup: ... paying it back.
Matt Klein: Yep.
Randy Shoup: Uh, [00:37:00] okay, so, uh, in 2003, we'd gone through a series of l- we were peaked at maybe 400, 500 employees and then like 300, then 200.
And so at some point, like the company continued on and certainly the value that it provided is real value. Um, but at some point it was im- time to move on to something else. I briefly went to Informatica, which is the data, data movement company, to go work for my first boss who I, I'd worked for at, um, at, uh, Oracle, a guy named Girish Pancha, who, um, just retired essentially from, his company called StreamSets, uh, acquired by IBM, I believe.
Um, anyway, he's awesome. But then I went to eBay, and I've been at eBay twice. Uh, so I joined eBay in, first, for the first time as an individual contributor 2004 to 2011, and I worked on eBay's search engine. So at the time, uh, search was a thing, like web search was out there.
Matt Klein: Yep.
Randy Shoup: In the early days it was, um, [00:38:00] let a thousand flowers bloom, so it was like Yahoo!
and AltaVista and Ask Jeeves and like a bunch of names mostly people have forgotten. Um, and Google with PageRank just like blew everybody away. Like it was just so clearly the results were better, they were faster, they were cleaner, just better. Um, blew everybody away. Uh, okay this is different now, but in 2003, 2004, eBay had a search problem, but it was a very different search problem than web search.
So web search at the time was very static. Like, you would crawl the w- I mean, so these things still happen, but on a different timescale. You would crawl the web. But like, web pages don't change mostly, like that often. Um, we're not talking about news things, but like most web pages are roughly the same as-
Matt Klein: Yeah, sure
Randy Shoup: they were yesterday or a year, or a week ago. And so there was a lag, but and there... It was in the problem. Like, it wasn't weird to have there be a lag between the crawling and the, uh, the querying, essentially. [00:39:00] Um, and so the classic model of a Google, Bing-style search engine is crawl the web, build an index, do s- querying on that static index.
eBay has a different problem. eBay is a marketplace that changes-
Matt Klein: Right,
everything is changing constantly. Sure.
Yeah.
Randy Shoup: And more, this might even be true now, I don't know, but like Amazon, even today, doesn't change as fast as eBay was changing 20 years ago. Why? Because like eBay is individual sellers.
Matt Klein: Exactly.
Yeah. No, no, makes sense. Sure. Yeah.
Randy Shoup: Auctions. Like so, and um, and not commodity. Like I'm selling this particular coffee cup.
Matt Klein: Yes. Yeah. For sure. Yeah.
Randy Shoup: Right? Which, which my, my son made for me. Um, uh, anyway, so, so we had a real-time search problem, and there were no real-time search engines on the planet. Like no research at all.
Um, now you get it for free with Elastic, so kids are like, "What do you mean you don't have real-time search?" Like you get, like get that for free. But that didn't exist. So, um, I [00:40:00] joined in 2004, but the building of this real-time search engine had started by a bunch of ex-AltaVista rock stars in 2003. And, um, I feel very lucky, frankly, to have learned- uh, joined in 2004.
I learned a ton about building scalable systems from them, and we built a real-time search engine. Everything's in memory.
Matt Klein: Hmm. Super cool.
Randy Shoup: Um, happy to talk about the technology there. It was super cool. Uh, the... 'Cause at the time, no inverted index. Ev- Inverted indexes were built statically, and so like there was well-understood ways of like compressing the index and spending a bunch of time in this batchy way, like making the index all cool.
Matt Klein: Right, so you would like pre-
Randy Shoup: And we were like, "We can't do any of that."
Matt Klein: You would like pre-,
pre-compute everything once per period of time, and then you would
search against the index.
Randy Shoup: Yeah, we did not do that.
Matt Klein: Right. Yeah.
Randy Shoup: That, that was the classic was build it in batches, serve it static.
Matt Klein: Yes. Right.
Randy Shoup: And there's only so...
It's just like data movement, right? Uh, Lambda architecture back from a million years ago, like batch, batch, micro batch, micro batch, micro batch.
Matt Klein: Yeah.
Randy Shoup: Like there's no way. You c- You have to get to streaming, and [00:41:00] that's what this was. It was s- uh, streaming- Events essentially streaming changes to the index in at multiple thousands per second and taking in the same process queries at multiple thousands per second.
And obviously, you know-
Matt Klein: But would it, would it use like a, like a static index and then put changes on top? Or, or, or would it always, like, would it always compute the active live index, if that makes sense? Like I'm trying-
Randy Shoup: Yeah, it does. Beautiful
Matt Klein: ... trying to, trying to wrap my head around-
Randy Shoup: Yeah. Uh- Yeah
obviously I'm not at all surprised, Matt, but you know your, you know your stuff. Uh, it was the second. So it is, uh... It was lock-free tacking additional... So tacking additional entries onto the bottom of the posting lists and in a lock-free way, querying them at the same time.
Matt Klein: I see. Okay.
Randy Shoup: So there were-
Matt Klein: Makes, makes sense.
Got it
Randy Shoup: ... there were query threads [00:42:00] traversing the index at the same time as, uh, I forget what we called them, indexer threads-
Matt Klein: Yeah, that makes sense
Randy Shoup: ... adding to the index. And then-
Matt Klein: Yeah. So, so, so, so, so you would like, you would be constantly updating the main index, but then taking all of the updates that were in flight and merging it together effectively, like into the final, into the
final result.
Randy Shoup: Yeah. But, but yes, conceptually, but not m- don't think of it as merging.
Matt Klein: Right.
Randy Shoup: Think of it as, let's, let's imagine we have one posting list. There are many, but there's one posting list, and we are traversing it for matches. There's, there's a thread of-- that is traversing it down -
Matt Klein: Right
Randy Shoup: ... uh, to answer a query, and there is also separately a thread tacking new entries on at the bottom.
Matt Klein: Yeah. Okay All right, cool
Randy Shoup: And, uh, yeah, it was sup-super... I, I didn't invent it, uh-
Matt Klein: So you worked on that a-a-as an individual contributor, and then-
Randy Shoup: Yeah, yeah ...
Matt Klein: and then what happened?
Randy Shoup: And, um, I'm... Yeah... so eBay [00:43:00] had a very odd way of doing their engineering at the time, which they since have changed. But it was a very classic IT view of the world.
So, um, the, they had about 1,000 engineers, and they were thought of as a pool of fungible units. Like I-- for this project, not product, for this project, I need five weeks of HTML work, and I need four weeks of database work, and I need three weeks of application server work. Let's assemble a custom team for this tiny little project out of, I'll take one of you, and two of you, and three of you and throw them together.
Um, and you can imagine that that didn't work very well. It didn't produce very good, uh, code, and obviously there's no code owner, no long-term code ownership.
Matt Klein: Right.
Randy Shoup: So like it was a big old mess, even worse than most places. Okay, so the, the solution, which helped me personally, but wasn't the s- isn't the solution I would [00:44:00] recommend, is, okay, well, we have this, we have this fungible group of 1,000 engineers.
What if we hire 20 industry rock stars and call them architects, and have them tell the other 1,000 what to do? And like, I'll say openly for my career and what I learned, that w- I was one of the 20 and, uh, I learned a ton, uh, by like basically-
Matt Klein: Yeah, sure
Randy Shoup: ... I mean, again, don't do this. Please don't do this.
This is not a recommendation. But like I was the architect for the search area, so I learned a lot about the search engine, uh, and helped with that. But also like the frontend of the search, it's like, it's like an L-shaped thing. So like search is a foundational ca-capability that we at eBay used for everything, absolutely everything, anything that needed to change, recommenders-
Matt Klein: Yeah
Randy Shoup: uh, merchandising, all these kinds of things, like we used the search engine. And then there's this vertical component of like people coming and finding things on the site. So I [00:45:00] owned- As an architect, in quotes, like I owned that whole L. Um, and again, don't do it that way. What you really want is autonomous teams with, with cross-functional-
Matt Klein: Yes
Randy Shoup: ownership that own a f- own a domain, uh, in the long term. But anyway, but I, but I learned a ton about building scalable systems and, uh, had a ton of fun.
Matt Klein: Yeah. Cool. Um, s- wait, s- sorry, but you said that you had two stints at eBay.
Randy Shoup: Yeah. Right.
Matt Klein: Is that what you mean? You had that, the IC stint- Yeah ... and then the architect.
Randy Shoup: Yeah, so this stint, this stint was seven years from 2004 to 2011. I left eBay to co-found a startup, with another eBay colleague in 2011. As I like to say, we were in the 99% instead of the 1% of startups, so-
Matt Klein: Yeah. Sure
Randy Shoup: ... we didn't really do much. Um, I ended up joining Google as director of engineering for App Engine, so g- early days of Google Cloud.
Matt Klein: Yeah. Cool.
Randy Shoup: This is now 2012 to 2013. Um, so I ran engineering for App Engine, that's the platform as a service. Snapchat [00:46:00] ran on it for many, many years.
Matt Klein: Yeah.
Randy Shoup: Uh, Pokémon Go, three million other applications intuning, including-
Matt Klein: Well,
and, and just for a historical context, this is just a couple years before obviously the rise of Kubernetes and a lot of the open source solutions-
Randy Shoup: Yeah
Matt Klein: that started to, to kinda come into the
space.
Randy Shoup: Yeah, yeah. Google had been using containers internally for many years. That's very well known.
Matt Klein: Right.
Randy Shoup: And, um, very well known, like Google added the containers into Linux-
Matt Klein: Yeah
Randy Shoup: ... originally. Um, but there was, but there was no... Actually, when I was at Google in App Engine, there was not even Docker.
Uh, containers existed in the world.
Matt Klein: Yes, of course. Yeah.
Randy Shoup: Uh, but even Google had been using it for, I don't know, using containers for maybe five, 10 years at that point, had not seen the, what Docker saw, which is you could combine a manifest with a deployment unit with a, code thing, right? Like, it, it was, [00:47:00] I'm, uh, forgetting the words, but like the con- the code container.
Matt Klein: Sure.
Randy Shoup: But combine that also with a manifest of what's supposed to be in there, uh, c- make that the deployment unit. Uh, uh, Docker put all these things together that we at App Engine and other places at Google, like, had not, had not seen. And the only way, I'll just say very briefly, Docker was a fantastic success in terms of its ability to, uh, its mark on the world, uh, and unfortunately not a success as a separate business.
But the only way that we could currently have Kubernetes is because of Docker. Why?
Matt Klein: Absolutely.
Randy Shoup: Because Docker wasn't-
Matt Klein: Yeah
Randy Shoup: ... owned by anybody.
Matt Klein: Yeah.
Randy Shoup: Docker was independent.
Matt Klein: For, for sure.
Randy Shoup: So like Microsoft would work with them, Google would work with them, Amazon would work with them, everybody would work with them.
It's just, uh, in retrospect, you couldn't have-
Matt Klein: Yeah
Randy Shoup: ... planned a better, , a, a better way to develop an industry standard than-
Matt Klein: Yeah, I, yeah. I, I mean, we'll, we'll probably have to do a f- follow-up episode on all of this because there's-
Randy Shoup: sorry
Matt Klein: ... because there's so much here.
Randy Shoup: Yeah.
Matt Klein: But, but, but just to come back, so you were at, um, so you were [00:48:00] working on App Engine, and then what, what brought you back to eBay?
Randy Shoup: Yeah. Uh, I did a few other things, which I, we don't have to go into, Stitch Fix, WeWork, a gaming company in there. Uh, what brought me back to eBay? So I was at WeWork. People may know, that WeWork did not go public and laid a bunch of people off and subsequently went into bankruptcy. I was part of that post-IPO attempt layoff.
I ran about a quarter of the engineering team, several hundred people.
Matt Klein: Yeah.
Randy Shoup: Uh, and I was looking for my next thing to do. I was thinking about maybe going back to Google, talking to other places, and then the eBay CTO, who's currently still the CTO, and the CEO, who's currently still the CEO, this is now 2000, in the pandemic, um, they asked me to come back-
Matt Klein: So you mean 20-
Randy Shoup: and be the chief architect
Matt Klein: Y- uh, 2020. You
said 2000.
Randy Shoup: 2020. Yeah. Right. Right. Sorry, I said the math. Yeah. Right. 2020.
Matt Klein: Yeah.
Randy Shoup: Um, and, uh, asked me to come back and be the chief architect, and essentially-- Well, then not essentially. They said, "Hey, I need somebody to come back and shake [00:49:00] things up and, bring eBay into the modern world of software."
And so that's what I did for 2020 to 2022.
Matt Klein: That sounds
like a very big challenge. I, I-
Randy Shoup: It was indeed a big-
Matt Klein: Yeah
Randy Shoup: ... challenge, and I had to think about it. Um, I had to think a lot about it because I had been out for 10 years, and it's like I left for a reason, right? And I'd been out for 10 years, but I had like way more-- It was like the combination of I know eBay very deeply 'cause I'd been there for a long, long time.
Matt Klein: Yeah.
Randy Shoup: And people don't cycle out very often, so, like all the players, I still knew they were all still there in diff- slightly different roles. Uh, and I had 10 years worth of, we didn't have the word the whole time, but like continuous delivery and modern software in the real world-
Matt Klein: Right
Randy Shoup: ... in the outside world.
Matt Klein: I, I, I mean, but what-- For, f-for what you could share, I mean, what was it like, like trying to bring these practices in? I mean, I would imagine that would be an incredibly, incredibly difficult thing to do-
Randy Shoup: Yeah
Matt Klein: ... I mean, to, to, to try to change an organization like that.
Randy Shoup: Yeah. Uh, it [00:50:00] absolutely was. It's a big old challenge, 5,000 people.
Um, well, I've actually given conference talks about it, so I'm not-
Matt Klein: Okay. Yeah, we can-
Randy Shoup: Uh, it's, it, it's completely open.
Matt Klein: Right.
Randy Shoup: Uh, yeah, you can put that in the show notes or whatever.
Matt Klein: Yeah,
yeah.
Randy Shoup: Uh, I can give the brief thing and then go into whatever detail you want. Uh, so the first half of the story is we doubled engineering productivity.
Doubled. What do you mean by engineering productivity, Randy?
Matt Klein: I was just
gonna ask-
Randy Shoup: Thank you for asking, Matt
Matt Klein: ... because, I mean, I, I f-I feel like from an industry, we're very bad at actually-
Randy Shoup: We're very bad
Matt Klein: ... measuring what, what that means.
Randy Shoup: Yeah. It's, uh, the fancy term is flow velocity, but you can think of it as features and bug fixes per unit time.
Matt Klein: Okay.
Randy Shoup: So, uh- a team before we worked with them, a team would produce N features and bug fixes-
Matt Klein: Right
Randy Shoup: ... per week. And after we worked with them, they could produce 2N. Uh, and this is in 20-- This started in 2020, so we had the accelerate book, we had the DORA metrics. So those were the outcomes we were trying to [00:51:00] drive.
How can we drive-
Matt Klein: Yeah, it, it's more that, and like we're, we're almost out of time. I, I find this so fascinating.
Randy Shoup: We're
gonna
have to come
back,
Matt.
Matt Klein: Yeah. I, I mean, we, we-- like, we probably will. It's more that even those metrics, if you don't link them to business outcomes, like actual conversion, goods sold, like whatever-
Randy Shoup: Yeah
Matt Klein: it's-- I'm being honest, it seems kind of meaningless to me. So it's like I feel like from an industry we're, we're always trying to find something to measure, but I'm just curious how you think about it because it's like measuring those things, like I, I get how that is a unit of measurement that you can track, but it doesn't in and of itself relate to business outcomes, right?
Randy Shoup: Right. Absolutely true. Uh, uh, we might have to pull a, put a pin in this, but like that's the second half of the talk. This, we have-- The, the second half of it-- So first half of the talk is how we de-doubled engineering productivity, which I'm very proud about, and it's removing bottlenecks.
Matt Klein: Right.
Randy Shoup: Uh, working individually, what's blocking you?
Matt Klein: Yeah.
Randy Shoup: How can I unblock you? I had no [00:52:00] novel ideas, I like to say, like we just executed the DevOps playbook. Uh, but it did not make any difference in the, in the ultimate outcomes at eBay. Our theory was that engineering productivity was the bottleneck-
Matt Klein: Yeah
Randy Shoup: ... and the reality was it was not.
Matt Klein: Okay.
Randy Shoup: Uh, and, uh-
Matt Klein: Fair
Randy Shoup: I don't wanna
Matt Klein: No, no, no, no, no, no, that's fine. Um, prior to us wrapping it up-
Randy Shoup: Yeah
Matt Klein: ... I, I honestly think that we should do another episode just to talk about some of this stuff.
Randy Shoup: Yeah.
Matt Klein: But, um, I, I would be remiss if I did not ask you about, you know, we've started to talk a bit about the AI boom and those kinds of things.
Randy Shoup: Yeah.
Matt Klein: Just y- given your long history, um-
Randy Shoup: Yeah
Matt Klein: ... in the industry, just, just wanted to get your short thoughts on what is, what is, uh, happening right now. I mean, do, do you, do you think engineers are obsolete? Do you think they're not obsolete? Like how, how do you think the next five or 10 years are gonna go?
Randy Shoup: Wow. Five or 10 years, I don't think we, I don't think we've ever been at a time when we know less about what's happening in the five or 10 years. [00:53:00] Uh, let's see. AI is absolutely real. The hype-- It's hyped, And overhyped and oversold and so real and so transformative at the same time.
Matt Klein: I agree.
I, I, I talk about it the same way. It's both hyped and real. It's ve-very interesting.
Randy Shoup: Yeah.
Matt Klein: Yeah.
Randy Shoup: It's, it like, it-- Are a lot of people gonna lose a lot of billions of dollars? Oh, yeah. Are... Just like in the dotcom boom, guys.
Matt Klein: Yeah, sure.
Randy Shoup: Uh, and also, is this absolutely fundamentally transformative for engineering-
Matt Klein: Yep
Randy Shoup: for technology, for the, for hu-human civilization? Absolutely also true. Uh, I'm not a doomer nor an accelerationist. Like, I don't think it's the end of humanity. I also don't think it's utopia. Uh, what it is though is a new kind of power, and power doesn't have valence. Like, nuclear power isn't itself good or bad.
Electricity isn't itself good or bad. Machinery is not itself good or bad. It's what we do [00:54:00] with them as humans, right? So, um, so it's, it's power. It's a new, like, cognition for f- for free or for cheap, uh, which we've never been able to have. Okay, do I think engineers are done? Absolutely not. Jevons paradox, look it up.
I know you know it. Jevons paradox says, uh, when you make a commodity less expensive and there is elastic, by which I mean infinite demand for that commodity, there will be more, certainly more usage of the commodity, but even larger net spend on that commodity.
Matt Klein: Yeah.
Randy Shoup: So let me, let me tie that back. He was talking about coal and e- and, uh, power in the 1850s, uh, Jevons was.
Uh, but it's the same exact thing now where cognition, is now 10X, 100X cheaper. Code is free. Does that mean we need no more engineers? No. We're gonna need more engineers. And there's gonna be some disruption between now and then, just like in the [00:55:00] agricultural lev- revolution, just like in the Industrial Revolution, just like in coal and steel, just like in oil, just like in inter, uh, uh, information revolution.
Like, every one of these revolutions has been deeply disruptive, and the people that are in the middle, like, those jobs change, and those people's, uh, lives change.
Matt Klein: Yeah.
Randy Shoup: And also, in the longer term, we have... Every time we've tr- we've thought in our industry that we are, "Oh, business people can do stuff. We don't need engineers," we just make more engineers.
Matt Klein: Yes.
Randy Shoup: Right?
Matt Klein: Sure.
Randy Shoup: Uh, and by the way, Excel is the, uh, biggest programming language on the planet.
Matt Klein: Yeah. I, I, I, I mean, I honestly feel very similarly is that-
Randy Shoup: Yeah
Matt Klein: ... I think the technology is real. I also think that not only are we gonna have just as many engineers, I think that what is happening right now in the near term, there's gonna be an inflection where we're gonna have a lot of mess to clean up in the next-
Randy Shoup: Oh, yeah
Matt Klein: few years, right?
Randy Shoup: Yeah.
Matt Klein: Which is gonna keep people busy for a long time. Um-
Randy Shoup: Yeah
Matt Klein: ... but that's, that's a whole separate [00:56:00] conversation. Um-
Randy Shoup: Yeah. I, I think all those things are... It, it's absolutely true, but I kn- I know you're not really implying this, but the engineer's job is not gonna be cleaning up AI. Like, it's not gonna be only-
Matt Klein: Oh, no, no,
no, no, no
Randy Shoup: cleaning up...
You're, you're not saying-
Matt Klein: No, for sure. Yeah, yeah, yeah.
Randy Shoup: But-
Matt Klein: I'm just saying that we, w- we are creating some debt right now.
Randy Shoup: Oh my gosh.
Matt Klein: And, and as we learn-
Randy Shoup: Yeah
Matt Klein: ... to use these powerful tools, it's like we're-
Randy Shoup: Yeah
Matt Klein: ... gonna have to figure all of that out. So-
Randy Shoup: Yeah. We, we used to- You know ... only cr- I mean, just not even, I'm not even joking, although it sounds like I am.
We used to only be able to create technical debt at human scale. Now we are creating technical debt at industrial scale.
Matt Klein: Yeah. Um, I could talk about this all day. I, I think that is probably as good a place as any- Okay ... to end. Uh, thank you, Randy. This was an awesome conversation. I think people are really gonna enjoy it.
Randy Shoup: I hope
so.
Matt Klein: Um-
Randy Shoup: Yeah.
Matt Klein: Yeah, yeah. Uh,
Randy Shoup: we can actually talk about something modern.
Matt Klein: No, I know. A- and I, I, I think maybe it's worth doing a follow-up because there is more here, but, but-
Randy Shoup: All
good
Matt Klein: ... it's time, t- time to wrap it up. So, um, thank you so much. That's a wrap for this episode of Beyond the Noise: Signals, [00:57:00] Stories, and Spicy Takes.
Huge thanks to Randy for joining and sharing his story. You can find this episode and all past ones on the
bitdrift
YouTube channel. If you had fun, drop us a review, tell your friends, or yell your favorite hot take into the void, and just make sure to tag us. I'm Matt Klein, and I'll see you next time.
Thank you, Randy.
Randy Shoup: Thank you.