Transcript

Paul Ford: Hi, you’re listening to Track Changes, the podcast of Postlight, a digital product studio at 101 Fifth Avenue in New York City. My name is Paul Ford. I’m the co-host of Track Changes and the co-founder of Postlight.

Rich Ziade: And I’m Rich Ziade, the other co-founder of Postlight.

Paul: That’s exactly right. And today we’re gonna talk about services — no, uh, Rich, tell them what we do. Let’s get it done.

Rich: We are a, uh, top-tier — here’s no other way to put it — web platform, actually internet platform, not just web. But we build platforms, and the apps and web apps and mobile apps that ride on top of them, and we do it at scale, and we kick ass. We can speak confidently, at this point, right?

Paul: I think that’s true. We don’t —

Rich: It’s an incredible, incredibly talented group of people.

Paul: All right, Rich, so here’s the thing that happens a lot.

Rich: Yeah.

Paul: OK, so people send us an email, and they’re like, “Hey, what do you guys think about machine learning?” Or they say, “What is our crypto-currency strategy, Postlight? Can you help us?”

Rich: Right.

Paul: And I’ll be honest, when they show up with that, it’s often, like, oh boy. OK.

Rich: Yeah.

Paul: We’re probably not your guy.

Rich: Well…I think, I think that’s right, because I think we’re very, uh, real-world and pragmatic, and we’re not bought into the buzz and the catch phrases. I’m gonna go way, way, way back, and many of our listeners may not know what I’m talking about. I, uh, decided out of law school that I was gonna dive into tech.

Paul: Mmmm hmmm.

Rich: And I just loved it, and I wanted to be a part of it, and I wanted to dive as deeply as I could dive, so I started thinking, OK, well what is Microsoft talking about? And Netscape is exploding. It’s the beginning of the internet.

Paul: Sure.

Rich: And they started talking about this standard called SOAP.

Paul: Oh! Wow.

Rich: And before SOAP, to get systems to talk to each other —

Paul: Which I can’t even remember what it stands for. I think it’s, like, “Simple Object Access Protocol.”

Rich: I think you’re right.

Paul: So it’s about that good.

Rich: All right, I don’t want to drag this out. Here’s the thing. SOAP was a way for systems to talk to each other.

Paul: Yeah.

Rich: In a, and use the internet —

Paul: Over the internet.

Rich: Before that, there were these terribly…difficult and bloated protocols, like, I think it was called CORBA?

Paul: CORBA, yeah, yeah. So SOAP, SOAP used web technologies. You could, you could send, essentially, it was like sending…sending a message from one server to another was kind of like sending an email.

Rich: Right. And I looked up a couple of articles on it, and I looked at a couple of examples of what those packets could look like, and I didn’t understand any of it, and I hated it.

Paul: Yeah.

Rich: I hated that I couldn’t pick it up.

Paul: It’s terrible, right? It’s like an 800-page standard. You just get this huge throbbing document in your —

Rich: Yeah. And I didn’t understand it, but it got so hot.

Paul: Everybody talked about it.

Rich: Everybody talked — Oracle gets in — when Oracle gets in the game, it’s really hot.

Paul: I had the same thing going, where I was just, like, oh my God, if I don’t understand this, I won’t understand what’s gonna happen on the web.

Rich: Exactly.

Paul: And everybody’s like, “Forget the web. Throw your browser in the garbage. It doesn’t matter. It’s all SOAP.”

Rich: Right. And then, so I bailed.

Paul: Yeah.

Rich: I bailed. I didn’t want to be technical, so I, I leaned over to the product, the design-driven product management side for years.

Paul: Because you were like, I can’t, I’m not, I’ll never be worthy of SOAP.

Rich: I’m not gonna do this. I was, I was doing, like, a, a, like…serious UX/product thinking, problem solving, which was a huge business, like, it was a good career path.

Paul: Yeah.

Rich: And it was still touching tech.

Paul: Sure.

Rich: And then 20…04 comes around, and I read an article about REST, which was another approach.

Paul: Sure.

Rich: And I read it in like fi — well, before REST, actually, Dave Winer, many people will know who that is, wrote a piece, he said, look, SOAP’s stupid, and he was just one of those people that says everything’s stupid.

Paul: It was a very emperor’s new clothes kind of thing —

Rich: Yeah.

Paul: Where it was like, “There’s no clothes on that.”

Rich: Exactly. So he came out and said, “Why don’t you just do it this way.”

Paul: Mmmm hmmm.

Rich: And I understood it the first time I read it, and I thought it was beautiful. Like, that simplicity was incredibly attractive to me. And then the Web 2.0 article comes out. Tim O’Reilly’s long missive on, “Here’s how things should be. The web is actually an application platform.” Blah blah blah blah blah blah. But it, it was digestible. It wasn’t a bunch of nonsense coming out of a working group. And I started a company because of it.

Paul: Sure.

Rich: And I started a company that evangelized REST in 2004.

Paul: And REST let you just build software like you were building web pages. It was the same set of ideas, basically.

Rich: It was readable.

Paul: Yeah.

Rich: It was clear.

Paul: And that’s where, when you hear about JSON and JavaScript objects and all that stuff flying around, that’s, that’s what that started.

Rich: Exactly. And the people I was reading back then was Dave Winer, Paul Ford, because he was just a weirdo. Honestly, he brought such a, such an outside perspective, even though I could tell he was technical. He didn’t like the conventions, which I thought was really interesting.

Paul: I didn’t, you know, I think I was like you, in that I wanted to belong. I wanted to get in.

Rich: Exactly.

Paul: And I, but I didn’t go get a job anywhere, because I thought nobody would hire me, I had an English degree.

Rich: Yes.

Paul: And the web was like that. You were just building stuff and, and there wasn’t really a discipline around it.

Rich: That’s right.

Paul: And so, uh, I would write on my blog about things I didn’t fully understand, and just, that was my way of figuring them out, and finding out who would talk to me and, like, what was going on. And that, that… [sigh]

Rich: Versus the bullshit Oracle white paper.

Paul: See what I miss is there was a sense back then of people — people used to discover things in public. That was the, probably the best application of internet technology I’ve seen, was when people were blogging and they’d be like, “I don’t know what this is. Let’s poke at it.”

Rich: Yeah.

Paul: You know who still does that? There’s a great, uh, there’s an old XML hand named Tim Bray —

Rich: Yeah.

Paul: Who, not like chronologically old, he’s in our cohort, but like, we are probably actually — anyway, regardless, he’s young in spirit. [laughter] And he still takes tech apart, and just kind of pokes at it.

Rich: Yeah.

Paul: He’s like, “Well let’s look at this thing.”

Rich: Yeah.

Paul: Tbray.org, great website. Smart person, who’s been around forever, and his way of learning is just picking it apart.

Rich: Yeah.

Paul: And if you don’t get it, that is a flaw in the technology.

Rich: Exactly.

Paul: If I can’t tell you what this does and why you’d wanna use it —

Rich: Yeah.

Paul: Which is where I come down a lot on blockchain. People are like, “Ah, it’s worth a lot of money.”

Rich: Yeah.

Paul: And I’m like, great. I mean, so are a lot of things. [laughter] Like, bricks are worth a lot of money.

Rich: Yeah.

Paul: There are people who are really into sneakers. They’re worth a lot of money.

Rich: RIght.

Paul: I can’t build anything for you just because they happen to be worth something, and then everyone’s like, “Ah! The blockchain, and smart contracts and…” It just reminds me very, very much of the late nineties.

Rich: Yeah. Exactly.

Paul: Where there’s just, like —

Rich: It’s just a mess, and everybody’s trying to, everybody’s trying to make it sound like they’re one step ahead of you.

Paul: Ah, it’s so big.

Rich: That you really need to jump on. That was, that was kind of the M.O. —

Paul: I’m not vulnerable to it anymore.

Rich: No.

Paul: No.

Rich: Well here’s the, here’s the thing I was able to do, right? I was able, I made this bet, and I said, I’m gonna start a shop in New York City. And I’m gonna build something that is high performing, that is gonna be considered cutting edge, that is going to file patents against the things we were doing, and it was gonna be based on simple, understandable tech that wasn’t a load of bullshit. And that’s probably one of the proudest things about the shop that I had built at that time.

Paul: Sure.

Rich: A shop called Arc90.

Paul: I mean, you know, and I think we, it’s worth saying, if anything carried forward from Arc90, that’s still basically the case here, I mean, we don’t —

Rich: We don’t like bullshit. We don’t like it.

Paul: You can list the number of things that we use to build, like, it…there’s probably 10 separate technologies total in our organization.

Rich: Yup.

Paul: And after that, you start to get real suspicious.

Rich: Yeah.

Paul: So we encourage, someone came up to me not too long ago from engineering, and was like, “I’m gonna get started, and just try to understand machine learning.” I’m like, “That’s great.”

Rich: Yeah.

Paul: Go get a sense, start to under — because that’s another one, where the absolute…it’s a little different than blockchain, because machine learning has actual applications at a certain scale, like —

Rich: Yeah.

Paul: Google is increasingly basing its, its business on machine learning.

Rich: Right.

Paul: Right?

Rich: And they have the pool, this massive pool of data, to feed it, right?

Paul: And it is explainable, but it’s about data. It’s, machine learning is, like, database-plus, at a real sort of conceptual level.

Rich: Yeah.

Paul: It’s not fundamental, in the way that, like, using an iPhone for the first time was fundamental. It’s not fundamental in the same way a browser is fundamental. It might have unbelievable economic impact.

Rich: Right.

Paul: But you come to us to build a product, and so you come to us to make something that people hold in their hand, or that they use on the web, or the platform underneath it, right? That platform could connect to machine learning. That’s possible.

Rich: Yeah.

Paul: You could give it a giant dataset and say, could we extrapolate some features from this, and use that to be predictive. That’s all good. And, like, we would support that if the client said, like, I really want this for a really specific reason. But it doesn’t actually happen.

Rich: No. Well, here’s…I think it’s worth talking about how it starts to seep into the agenda of someone walking into our offices, right? I think it goes down like this: first off, the name. The name is “machine learning.”

Paul: Yeah.

Rich: It sounds like one of those little robots that look like babies that walk around and learn.

Paul: Critically, it’s not called “hiring people,” which is something everybody hates to do. [laughter]

Rich: Right. It’s called “machine learning.”

Paul: It’s called “robots save money for my boss.”

Rich: Robots that learn. It’s like, oh! He learned Spanish this past summer.

Paul: Again, and it’s so much better than hiring people.

Rich: You’re humanizing this incr — it sounds…so that’s step one.

Paul: Yeah.

Rich: Step two is CNN wrote an article about it.

Paul: Mmmm hmmm.

Rich: Your boss, who is MBA, really smart.

Paul: It doesn’t even have to be —

Rich: Really made the company happen. You walk over to the CTO’s office —

Paul: Doesn’t even have to be CNN, too. It could be MIT Technology Review.

Rich: Yeah, they’re poking around.

Paul: Everybody’s trying.

Rich: Everybody’s trying, right? And they’re like, “So uh…Diane. What are you doing about machine learning?” [laughter]

Paul: Yeah.

Rich: That’s literally, they’re not thinking about it in terms of a context that —

Paul: And now Diane has a problem. Diane’s job is to, like, make sure all the orders get fulfilled for baby carriages from the baby carriage company. And she also has to deal with, like, turning that into just-in-time delivery for all the manufacturing…

Rich: Yeah.

Paul: And so on and so forth, and…

Rich: Yeah.

Paul: Then the boss goes and reads, and he goes, like, “Wow, these people achieved a 30%, you know, they’re just-in-time inventory got 30% faster because of machine learning.”

Rich: Right. So phew!

Paul: Boy.

Rich: What do you — like, so that, that mandate’s given out, and what they usually ask for is like, I want a memo in, like, 45 days.

Paul: Now Diane’s in a pickle.

Rich: She a) needs to get educated, and again, let’s not assume that Diane doesn’t know what’s going on. She may know exactly what machine learning is, but she’s got to wedge it into the strategy of the business she’s in.

Paul: I think also, though, it’s worth framing it this way: no one knows what machine learning truly is, because there are enormous numbers of implementations of neural networks and things that are related to what we call machine learning, but it’s actually a giant buzzword that doesn’t have a single specific definition, and in that is all the opportunity for the businesses that get to stand up and say, “That person doesn’t know what machine learning is.”

Rich: Yeah.

Paul: “I do.”

Rich: Yeah.

Paul: And so that’s where you’re really kind of screwing over Diane, because the boss comes to her and is, like, you know, he doesn’t say, like, “Use these five algorithms,” he’s just says, “I need this, this thing seems like it’s gonna save me a lot of money.”

Rich: Well he’s, yeah, exactly, and he’s thinking about it in the context of, are we in the game?

Paul: Right.

Rich: I, I cannot be in a position where the board sits me down two years from now and they say, “How did you fall behind on machine learning? All your competitors have machine learning.” Right?

Paul: Because Willy’s Baby Carriage Company has machine learning.

Rich: Right. Exactly. And, and who knows what went on over there, right?

Paul: You know, we’re joking, but it’s not that unlike reality.

Rich: No, this is real, right?

Paul: Yeah.

Rich: And, and so…

Paul: So if Diane is, Diane might call us, at that point.

Rich: She might call us and say, “These guys are next-level, and I want to, I want you to think about how machine learning is going to impact…or how it can impact my business in a positive way.”

Paul: Right.

Rich: And now, let’s, let’s, let’s go to Postlight for a second. Let’s go to Postlight’s offices for a second. We are a shop, and this isn’t meant as a selling point, or a…anything else. We are a shop that will look you right in the eye and say, “What are you talking about?”

Paul: Right.

Rich: And we’ll say that at meeting one, and we’ll say, “Can you give me a better idea — ”

Paul: You’ll say that, I’ll say, “Actually hold on just a sec. Can you tell me what you’re talking about?” [laughter]

Rich: Right.

Paul: Because I’m the gentler of the partners.

Rich: You are the gentler of the partners. I just, and I really…it’s not, it’s not meant in a, in a demeaning way. I really want to understand, like, maybe this is the genius in front of me, and I wanna, I wanna unpack this.

Paul: Mmmm hmmm.

Rich: Which sometimes happens, but usually, it’s…I just have to, they’re walking in with the buzzword slung over their back, and they wanna know how it needs to be applied in their world, and that’s weird to us. We, like, we, we want to, we want to apply it to the world, and we want an example that really, truly resonates in a meaningful way, and has a dotted line to human beings.

Paul: And I think this is really important, right. We…when we call ourselves a product company, what we’re doing is actually eliminating a huge range of technologies.

Rich: Yes.

Paul: OK, we’re actually saying, what we’ll do is we’ll make an experience that someone can have with software that will be reproducible, that they will enjoy, that will lead to a good outcome.

Rich: Yes.

Paul: And we’ll use every sort of, like, both scientific and also very kind aesthetic set of skills —

Rich: Oh yeah.

Paul: That we have.

Rich: Oh yeah.

Paul: But what I refer to a lot, what I talk about a lot is, it’s not science. Like, when people come in, a lot of times they want science, they want, like, prove to me that we will get these results, and give — you know, Google is famous for this. They had like, they tested 43 shades of blue to see who’d click on things.

Rich: Right. Right.

Paul: And some of that’s great. A/B testing is great, there are all sorts of systems that are really useful and give you data, but ultimately humans need to interpret it, and make decisions, and that’s how you get a product built.

Rich: Yes.

Paul: So when machine learning enters our world, it enters not as, like, this big abstract thing where we’re gonna tell you some great strategy. It enters our world as software that we might use, APIs we might call externally.

Rich: Yeah. But we also wanna unpack it —

Paul: Absolutely.

Rich: And see, does this apply here? Because like you said, the world of machine learning is very far and wide, and people are using it in very loose ways, so what do they mean?

Paul: That’s right. And also, a lot of times, you know, when they come in, Diane might actually just be asking us to make the app a little more efficient, and do some relatively intelligent queries on a large set of data.

Rich: Yep. That’s right.

Paul: Things that we’ve been doing for 15 years.

Rich: Exactly. And we’ll hack it. If I’m able to pull it off without hitting Google’s APIs to get — like, I gotta tell you: we did a Bloomberg project, we were teasing out the key people out of an article.

Paul: Mmmm hmmm.

Rich: Very hard problem.

Paul: That’s right. So if it says that John Smith played tennis against, you know, Sam Smith.

Rich: Exactly.

Paul: It would know that John and Sam were two human beings and that, possibly, that they were tennis players.

Rich: That’s right. So we tried this on Google’s API. It was spectacular. I mean, that was more, the term here, we’re gonna throw out another term, sit tight: natural language processing.

Paul: That’s right. Which is also related to machine learning. It’s actually an application of machine learning.

Rich: Right. Right.

Paul: And here’s what’s tricky, is all this stuff is open source. Like, there are ways to really get involved and use it. Or you could go to Google and pay them some money and just use their amazing API as if you were hitting a web page.

Rich: It’s incredible. Like, the quality of the returns — but we couldn’t, we couldn’t afford it.

Paul: Yeah.

Rich: We couldn’t, that wasn’t feasible for our project. So we started hacking. And we just —

Paul: We should actually break that down for people. That’s a great, it was a great product if you were, like, doing a big research or analysis project, but if you wanted to let millions of people get millions of results in basically real time it would’ve added up to tens, hundreds of thousands of dollars really quickly, to use that API.

Rich: A month.

Paul: In a month. Yeah.

Rich: Yeah. Yeah, yeah. It was very expensive.

Paul: So this project couldn’t sustain that.

Rich: Correct. So we figured out a hack. I’m not gonna go into the details of how we hacked it. But we spent two and a half weeks.

Paul: Right.

Rich: And it’s really good. It’s not great. It’s not great. Like, it stumbles on itself sometimes, maybe 15% of the time. But it’s really good. And I can tell you, in terms of the, the logic that was used to do that, it’s not machine learning. At least, it hasn’t earned —

Paul: It’s not machine learning.

Rich: It hasn’t earned the title of machine learning.

Paul: No! It’s, it’s —

Rich: It’s clever.

Paul: You use —

Rich: Is the way I would —

Paul: Yeah, it’s clever —

Rich: Would characterize it.

Paul: There might be a little statistic and model in there —

Rich: It’s clever, it’s clever. But it’s not —

Paul: Machine learning, to me, I mean, I’ll give you my definition, it’s when you give a computer…when you give a system an enormous amount of data, or a large amount of data, it doesn’t have to be —

Rich: Yeah.

Paul: Whatever, some amount of data. And the system uses a set of classifiers and some, some actual rules, and some actual algorithms to go through all of that data and extrapolate features in that data that can then be understood and applied in new ways. And this is something that —

Rich: Patterns.

Paul: Yeah.

Rich: That can, can be teased out, so that…it gives, effectively, the illusion of thinking.

Paul: Well, and what it’s about is about the connections between different things, like, oh, you know, like two pictures of a person with a red shirt.

Rich: Right.

Paul: Teaching it that that’s a shirt and that it’s red, right?

Rich: Right.

Paul: Like, being able to extract that and be, and connect that to a big model of language and go, OK, this…this is the red shirt thing. Now at no point does the computer actually know, in the way that you or I do, that something’s red, or that it’s a shirt.

Rich: No.

Paul: It just knows that there’s an enormous matrix of language over here, and a, and a set of pixels over here, and as it looks through the pixels and it sees certain values that align with red, and certain, um, outlines that align with shirt.

Rich: Yeah.

Paul: That it can kind of go look those up and then say, “I think I got a red shirt picture.”

Rich: Right.

Paul: And a human being can see that and be, like, “Oh man, I searched for a red shirt, and all the red shirts showed up.”

Rich: Yeah.

Paul: And it’s essentially like a giant database where things connect and knit together not just in the way that we usually think of databases, which is, “get me all the results that match the letter ‘z.’”

Rich: Right. Look em up and get em —

Paul: Yeah. It’s much more like, it’s, it’s pre-computed and figured out with an enormous amount of those connections before you can make them yourself.

Rich: So we know, we kind of know what machine learning is.

Paul: In a rough, in a rough way. I’ve been reading the TensorFlow documentation, because it’s how I relax. [laughter]

Rich: I mean, whatever works, Paul.

Paul: I just want an excuse to say ‘matrix’ all the time.

Rich: Yeah.

Paul: And ‘vector.’

Rich: Yeah.

Paul: Vector’s a really good word.

Rich: Vector is a good word.

Paul: Yeah. ‘Multi-dimensional’ gets in there.

Rich: Yeah. And, which is impressive stuff, and the thing is, we don’t dis — like, we’re not being arrogant and discounting the technology, which is interesting and impressive.

Paul: Right.

Rich: We’re scrutinizing the application of it, because very often, it’s just, “Could you please just inject this — ” It’s like squirting chocolate sauce onto your spaghetti.

Paul: That’s right. That’s right. And it’s great on ice cream. [laughter]

Rich: It’s great on ice cream!

Paul: So why would, why wouldn’t you have it on spaghetti?

Rich: Yeah. It might fit. And let’s scrutinize this, and really probe whether it’s a good fit or not, rather than just tossing a buzzword into, into the mix.

Paul: Do you know about the Gartner Hype Cycle?

Rich: I don’t.

Paul: So Gartner, which is a big…what the hell, how would you define Gartner?

Rich: They’re a….

Paul: Research firm.

Rich: I guess. Think — think tank…

Paul: Yeah…

Rich: Not think-tank, because that’s usually political…

Paul: They do, like, industry research, so you call Gartner and they’ll tell you, like, they’ll give you a quadrant, four-quadrant chart that shows you —

Rich: Yeah.

Paul: Uh, you know, the top consultancies for…

Rich: Yeah.

Paul: Digital —

Rich: They talk about trends, and…

Paul: Yeah. Who are the major providers of internet of things operating systems?

Rich: Right.

Paul: That’s a very, like, they’ll give you a white paper —

Rich: Yeah.

Paul: And you pay them a lot of money.

Rich: Yeah.

Paul: So they have this thing called the Hype Cycle, which actually shows how, and it’s, it’s a very sort of funny poppy piece of tech culture, and it…it shows how things get hyped up and then they kind of go into a trough, while everybody gets disappointed with them, and then they slowly get useable.

Rich: Right.

Paul: And this actually happened with the web, like, the web was like, “Oh my God, it’s changing everything.” And then everyone in, like, the early 2000s was, like, “I’m gonna…smoke cigarettes and think about why America is in trouble.” And then, uh, Web 2.0 showed up.

Rich: And it was a…

Paul: Suddenly it was a useable set of technologies that really started to replace desktop software, and everybody was like, “Oh my God, here we go.”

Rich: Yeah.

Paul: There’s a consultancy called ThoughtWorks, too, and they do a, a uh…a chart of what technologies are hot, and they have an idea — I think with them it’s more, like, adopt, you know, explore, adopt, buy, ignore, you know, concepts like that. What are you gonna do with this new thing?

Rich: Right.

Paul: They’re very granular. They’re, like, you know, Apache Matrix Database 7, and they’ll be like, “Well, don’t, leave that one alone. We don’t see a lot of successful adoption.” So there’s a culture actually out here of people going, like, “Nah, that’s too gee-whiz.”

Rich: Yeah.

Paul: A great example, it’s pretty nerdy, but it’s real, is React, which most people won’t know, but React is the, ti’s the toolkit for building frontend web and increasingly mobile apps, right? And I would say about two years ago, I started to really pay attention to it, and it was one of those things where you see it and you’re like, “Yeah, that one’s gonna win.” It was backed by Facebook, it was open-source.

Rich: Yeah, it came out of Facebook.

Paul: Yeah. And it was just, it maps really well to how people build web apps, as opposed to pages.

Rich: Yeah.

Paul: And so it got a lot of hype, but then you started to see, like, just humans show up, going like, “No, this is going to be how it’s gonna go.” And now we see it everywhere. It’s in finance, it’s in…music apps. Just whatever the hell.

Rich: Yeah.

Paul: It’s built in React.

Rich: It took hold, and I think…

Paul: So that’s a kind of technology where it shows up and you’re like, “OK, here we go.” Machine learning is different. Machine learning is like a whole concept without a specific implementation, where you’re not sure if you’re gonna be able to use it to help somebody or not.

Rich: Like you said, it’s this big, foggy ball of ambiguity. I mean, that just is what it is, and there are interesting things happening, and isn’t a lie.

Paul: No, no!

Rich: It isn’t one big lie, but I think…

Paul: Like Bitcoin.

Rich: Like Bitcoin.

Paul: Which may possibly be one big lie.

Rich: Which, I think it depends on being impossible.

Paul: Well machine learning you actually give it data and you get something out. There is, I mean, you can go and use a service and get great stuff to happen.

Rich: Yeah.

Paul: Bitcoin is all about humans.

Rich: Yeah.

Paul: But what’s funny is people come to us almost with the same level of, like, “What…should we do about crypto-currency?”

Rich: Yeah.

Paul: And we’re like, “Don’t…don’t do anything about it. What are you talking about?”

Rich: Yeah. “What…do you want?”

Paul: Yeah.

Rich: Very often the response, by the way, to these sort of dropping, like, dropping a hot tech trend or term is, “Well what do you want?”

Paul: Yeah, that’s what we say, right?

Rich: Because that’s an end — like, that’s a means to an end. These things, like machine learning and…Bitcoin are a means to an end, so what is, what is it you really want, and let’s work back from that, for a minute.

Paul: Right. And the reality is there’s lots of people who will swoop in before they get to us. If they want, if people want to be exploited — [laughter]

Rich: Yeah. Someone will take your money —

Paul: By some insane blockchain-related technology —

Rich: They will take your money.

Paul: Yeah, we’ll never even see them.

Rich: Yeah.

Paul: Because there’s people marketing to them right now, right?

Rich: Yeah. Yeah. Like, make it more confusing.

Paul: Exactly.

Rich: Make the, make the, um…offerings page, or the services page, way more confusing. So they, they think, “Oh my God, here are the experts.”

Paul: I have to say, though, like starting with SOAP and up to now, to blockchain, and that doesn’t mean that the fundamental ideas underneath these technologies are bad, there has been some damn thing that’s gonna blow it all up and you’re gonna have to deal with from the beginning. And the thing that does seem to work is stable software built along reliable platforms that’s pretty well-tested. Like, that, that seems to be what people actually want?

Rich: Yeah.

Paul: And if you keep building that —

Rich: They still want that.

Paul: They keep coming back!

Rich: They’ve been wanting that for a very long time, and they still want that. If somebody comes in and says, “I want to blow a hole in Netflix,” and I wanna do it, you don’t need an R&D lab.

Paul: It’s — yeah.

Rich: Like, you can do this. The, the scalability opportunities —

Paul: You wanna build a video platform, you can build a video platform.

Rich: You can do it. I mean, you can, you should think about how you’re gonna differentiate, but that’s a different…

Paul: That’s a product question, right?

Rich: A content question…

Paul: And marketing…

Rich: And all that.

Paul: Yeah.

Rich: But in terms of, is it achievable, if they come in and say, “Look, the way I wanna beat Netflix is through machine learning.” That’sssssss…….

Paul: Yeah —

Rich: Rough.

Paul: [sinking balloon noise]

Rich: It may be, in a corner of the experience, meaning it’s way better at recommendations.

Paul: Mmmm hmmmm.

Rich: Than Netflix, but don’t phrase it that way.

Paul: It’s one part of the puzzle, right?

Rich: It’s a very — yeah.

Paul: But I think if you were to go and get a lot of money for a content-based startup right now, you might have better luck than saying, “We’re gonna get the best content,” if you said, “We’re gonna be machine-learning driven.” That’s because they’re willing to make that gamble, right? That maybe you’ll hit it out of the park and there’ll be a 100X or 1,000X multiple.

Rich: Right.

Paul: But the vast majority of people who have that idea, and other people — lots of people do, are gonna fail.

Rich: Right.

Paul: And VCs are willing to play roulette that way. Going back to what you just said, people come with these new, exciting ideas, which boy do I understand, I like a new, exciting idea as much as the next person does. Shiny, exciting, wonderful.

Rich: Can I just say here, I mean, we’re making it sound like all of our clients and the prospects that walk in are dumb. Sometimes they come in and it is a fully-crystallized vision, and it’s incredibly exciting to hear it, hear about it, and talk about it. Some are more technical than others. They’ve put a lot of thought into how to get there, and those are exciting as well. In my head, I can think of a couple of clients who get it.

Paul: Right.

Rich: And understand what they’re chasing, and understand, uh, the tools and the capabilities —

Paul: This will sound, this will sound a little patronizing, right, but just about everybody who we now are working with, you could have a conversation about these technologies and what they, what they could accomplish for them.

Rich: Yeah.

Paul: But what, going back to what you said earlier, right, the fundamental problem is that most people are facing is not, how do I apply technology X to get, you know, incredible yields. That’s a very, like, startup-y, West Coast kind of problem. The problem most people have is: can I get a good enough piece of software shipped that people want to use?

Rich: Yeah.

Paul: That’s — that is it, and that is the fundamental, that’s the, still the fundamentally hardest thing that most people can pull off.

Rich: Sure.

Paul: And especially at an organizational level. I think if you’re a…

Rich: Within an organization, you mean.

Paul: Yeah.

Rich: Yeah.

Paul: If you’re in a big org, just try to get good software out the door.

Rich: Yeah.

Paul: 90% of your users probably hate everything that you’ve shipped.

Rich: Yet…getting them to change their patterns and habits is not easy.

Paul: No.

Rich: So it really has to be appealing. Which is also hard. But that’s…I agree. I mean, they are your barrier. Now mind you, we are not a shop that builds the bridge between two systems to talk to each other purely, though we, we go under the hood, but.

Paul: Sure.

Rich: As a pure sort of integration project that no human will ever see…

Paul: Yeah.

Rich: That’s not our game, really. So.

Paul: You don’t call us and be like, “I have this 30-year-old database that’s running over here and I need it to talk to this thing.”

Rich: Exactly.

Paul: We’ll do some of that to make your app work.

Rich: Yeah, in a relationship context we may do it.

Paul: But I think my, my larger point out of all of this is that there’s always a ton of bright and shiny out there, and if you’re out there and people are saying, “Well, don’t even bother, only learn about machine learning, that’s the hot new thing, only care about blockchain, you know, and Bitcoin-related technologies,” and there are a lot of people saying that —

Rich: Yes.

Paul: You —

Rich: What’s a, what’s a future…hot term?

Paul: [long pause] Phew. This is the funny thing, it’s machine learning for a while, right?

Rich: You know what’s funny? The word “machine”…

Paul: Yeah.

Rich: It’s got, like, a retro vibe to is. It wasn’t computer learning. It was machine. It’s machine learning. Which is kind of cool, like I’m thinking steampunk a little bit?

Paul: You know what comes back a lot is, like, massive parallelization, like, where one computer has 8 million cores.

Rich: Yeah.

Paul: That one shows up from time to time.

Rich: But that one doesn’t resonate with, like, the masses, right?

Paul: The smart —

Rich: Machine learning —

Paul: Yeah…

Rich: Is special.

Paul: The smart contracts, you know, and all the Bitcoin stuff. People really got into Bitcoin because it’s money, right? Like, it’s a pure, it’s a bubble.

Rich: Coin.

Paul: Coin.

Rich: As in Bitcoin, right?

Paul: Yeah.

Rich: Coins are gone.

Paul: Yeah.

Rich: There’s no such thing as coins anymore. Nobody has…do you have coins?

Paul: Sometimes I buy, like, a cup of coffee, and I get…25 cents.

Rich: OK. So machine learning. It makes me think washing machine, and on the other side, I’m thinking a six-year-old in the first grade.

Paul: Yeah, exactly.

Rich: It’s incredibly…

Paul: Like an abacus.

Rich: It’s beautiful, right?

Paul: Yeah.

Rich: It’s warm, it’s, it’s, it’s human. They’ve humanized a very, very complex technology.

Paul: But the hardest thing is still shipping good software that people wanna use.

Rich: That’s….

Paul: Like, all of this technology is essentially pointless.

Rich: Yup.

Paul: Unless there’s some product base that a human being can apply themselves to.

Rich: Yeah. Exactly.

Paul: So I think it’s just, like, a good framing. I would say 15, 20% of my brain might be given over to bright, shiny, new things, but 80% has to belong to how we build apps today, in the modern world.

Rich: Or hardware, or anything. Uh…we didn’t get to talk about the notch in the iPhone X.

Paul: Let’s get that out of the way. Let’s get that done.

Rich: So I don’t want to talk about it. I don’t want to talk — I understand the, the designers are walking the streets and whatnot. I wanna give it a different name. I’ve heard notch.

Paul: OK.

Rich: I’ve heard eyebrow.

Paul: OK.

Rich: Which is weird. It’s not really an eyebrow. It’s like one eyebrow. Unless you’re a unibrow.

Paul: OK.

Rich: I have one more. Do you have any, a name for it?

Paul: I…care…so little about this one.

Rich: Yeah.

Paul: I can’t…

Rich: I wanna close it with toupée.

Paul: You know what I — oh, toupée is good.

Rich: Toupée is real. I don’t want it there. You look ridiculous. Take it off, and just shave your whole head. It’s a toupée.

Paul: You know what I think about, did you watch the Apple event, where they announced this thing?

Rich: I did.

Paul: OK.

Rich: I saw bits and pieces.

Paul: The scene in the Apple — I watched some of it, too — and they’re, they’re showcasing the new Apple Watch that takes phone calls?

Rich: Uh-huh.

Paul: And there’s this woman on a paddleboard.

Rich: Oh, I saw this!

Paul: Who’s out having to take a phone call, and there’s a long lens on her, so she’s clearly just in the middle of some body of water.

Rich: She lost the coin flip.

Paul: She really did. She had to go stand on that paddleboard, and it’s the most awkward paddling, and she’s very balanced. She’s quite good at standing on the paddleboard. I would’ve been in the water. And she’s taking a phone call.

Rich: It’sssss….tough.

Paul: And I just think about that paddleboard a lot. Just, Apple sometimes doesn’t quite get to the marketing breakthrough.

Rich: I mean, the brainstorm was, what circumstances would you be in where you’re not gonna have your phone, but you really wanna talk to someone?

Paul: Paddleboarding!

Rich: Paddleboarding! Let’s do it.

Paul: Because it used to be, like, doctor emergency room, but now it’s just, everything’s fun.

Rich: Yeah, yeah.

Paul: Everything’s fun, and you just spend money and drive nice cars and go and get coffee.

Rich: Yup. Exactly.

Paul: Yeah, the toupée is good.

Rich: Toupée is incredibly strong.

Paul: Here’s the thing: you know, the, the $8 million iPhone X is an absolute marvel of every kind of possible engineering, and I’m just about this excited about it.

Rich: That’s OK.

Paul: It’s fine.

Rich: That’s OK.

Paul: It’s just part of the infrastructure of Apple —

Rich: It’s incredible. I don’t know who is doing it better, I mean, let’s put aside the business —

Paul: Nobody —

Rich: In terms of integrating engineering and software.

Paul: No, it’s great. What a great company.

Rich: It’s the best, right?

Paul: It can do such good work.

Rich: Yeah.

Paul: It really, it’s really amazing.

Rich: But like, on with life.

Paul: Yeah.

Rich: Everybody’s drowning right now.

Paul: Yeah. Yeah…

Rich: All right, listen, I wanna close this with a thought, Paul. We, we like talking to everybody. If, if you wanna talk to us about machine learning —

Paul: I think people know. People know.

Rich: Please come talk to us. We’ll talk to you about whatever you want?

Paul: People —

Rich: I don’t wanna hurt the business with this podcast.

Paul: I know, I know — but we’re not gonna hurt the business. People know, you know, that I think we’re allowed to say that sometimes people come in with mixed-up ideas about what they really wanna build.

Rich: Yeah.

Paul: And we’re happy, I will talk —

Rich: We love brainstorming it and working it out.

Paul: But it is good to clarify for people that we do a thing that is actually just sort of a specific craft inside of the giant world of technology.

Rich: Yup.

Paul: And that some of the new and shiny things, even though they’re on your phone, are actually built using disciplines that were well-established, you know, 15 years ago, and you continue to apply them, and when something new shows up, we’re often a little bit suspicious. Not new to — anything that will help us build an app faster?

Rich: Yeah.

Paul: We are on it.

Rich: Yeah.

Paul: Like, React is a great example. That accelerated people’s development times, and we went crazy for it. But things that actually are new and exciting and seem like they’re the future really slow you down, and that is very dangerous.

Rich: Dangerous is a strong word, I think.

Paul: Well until they get bundled up in a way that you can actually apply, use, and understand them, but when they’re in that foggy zone of here’s the future —

Rich: Don’t go drifting off into God knows where and do it for the sake of doing it.

Paul: Have a couple —

Rich: Yeah.

Paul: Go to a couple panel discussions…

Rich: Yeah.

Paul: See what’s up. So if you wanna talk to us about machine learning…

Rich: Talk to us about anything.

Paul: Uh…

Rich: Talk to us about anything. We’ll talk to you.

Paul: hello@postlight.com.

Rich: Yes.

Paul: And you should know that this is Track Changes, the podcast of Postlight, a digital product studio at 101 Fifth Avenue in New York City. Five stars on iTunes? hello@postlight.com. I’m gonna go paddleboard.

Rich: Just gimme a call.

Paul: I will.

Rich: While you’re out there. [laughter]

Paul: My LTE Apple Watch.

Rich: Yes.

Paul: With the paddleboard-resistant band.

Rich: Oh Lord.

Paul: OK, thanks everybody.

Rich: On that note —

Paul: OK.

Rich: Have a good week.

Paul: Back to work.

Rich: Bye.

Paul: Bye.