Tim Veil:

Well, welcome again to another episode of Big Ideas in App Architecture. I once again, am thrilled to be joined by a very exciting guest. We have today with US Stephen Blum, who is the co-founder and CTO of PubNub. And the way I like to start these conversations, just because it’s a great way to kick off the next 20, 30, 40 minutes is, let’s start by hearing just a little bit about who you are, what you’re doing today at PubNub. And really to some extent how you got here. I think it’s so fascinating to understand people’s journeys. Not only what they’re doing today obviously, but how they got there. So, welcome to the show and let’s start by just learning a little bit about you and what you’re up to.

Stephen Blum:

Ooh. Thank you Tim. Yeah, absolutely. Lots of exciting things. Things that I like are the tech and advanced algorithms, I’m very interested in that. But also banking systems that are stable. I definitely like those things these days. Not having to stay up late at night thinking about those things.

Tim Veil:

Yes. That’s important these days, isn’t it?

Stephen Blum:

Absolutely. Yeah. And the new way of the world, we are starting to see with artificial intelligence and how it’s being remarkably impressive and surprising in some of the things that it’s able to achieve. Have you played with any of it recently, Tim?

Tim Veil:

So, I did. Well, I don’t know. So, let me tell you what I have done. And I’m coming up to speed with some of these technologies, so you can tell me whether or not this is the latest and greatest. But, that was a couple weeks ago we had to go to Miami. We had our sales kickoff at work, at Cockroach Labs, beginning of the fiscal year. We got to kick everything off. And so I was asked to deliver a presentation, who the people we think are our competitors and what they are. So I did two things and leveraged this AI stuff that you read about it in the paper, or in the news. I asked ChatGPT what the difference between Cockroach and one of our competitors was? And I got to read those responses, which was incredibly fascinating, by the way, because surprisingly it was actually very, very accurate. And what was interesting is if I would rephrase the question just slightly, it would take a slightly different tone. I mean, you hear about it in the news and see about it and read it and all this other stuff. But I’d never really played with it. So, I did that. So that’s the first thing. And then, strangely enough, our sales leaders got into this whole thing about being either, what was it, sharks or lions? So I went to, I think it’s Stable Diffusion, which is one of these similar AI-image generator things. And I said something about a shark and a lion and it created this image for me that was like a lion on a shark’s body, jumping out of the water. Anyway. Is that particularly useful, Stephen? Probably not. But, back to the ChatGPT stuff, it actually was really neat. I mean, the power of asking a system like that a very simple question and getting something legible in return, understandable, and really quite honestly accurate in many ways, was pretty neat. So that’s my limited experience with it. But I know I’m just scratching the teeny, tiny surface of it all.

Stephen Blum:

Yeah. Of course, we’re all wondering is it going to replace human jobs? Yes, is the answer, pretty much, at this point.

Tim Veil:

Do you really think so? Is that-

Stephen Blum:

Well, here’s what we’re doing with it at PubNub. So, we have big teams, got just about 200 people. And they spend a lot of time working, developing, building technologies, adding new features, capabilities, making things more optimal. But also product managers and marketing teams that have a lot of content to write. And, guess what? We had one of our product managers ask it to create a product roadmap based on some of the items and it did his job for him, which is amazing. And so, here’s what we can do right now today. Everyone, every business, this is our advantage right now … But I’m evangelizing it internally at PubNub. Use the GPT and the large language models to do as much of the work as you can. Right?

Tim Veil:

Really?

Stephen Blum:

Have it generate as much of your day job and automate as much of your day job as possible.

Tim Veil:

And what’s been the reaction to that message? So, you’re the first person I’ve talked to about this, I don’t want to say in great depth, but certainly somebody who has a business that they run and operate and is actively investigating in this. So, how did that message land? I can imagine some people would be like … But maybe not, right? I mean, maybe it’s a relief. Like, “Hey, maybe this helps me be more efficient and productive and I can do some other stuff.”

Stephen Blum:

Yeah. The reaction is polarizing. There is some that our resilient to the new way of things. They see that they have, for example, technical skills that do outpace in many cases. But it takes them a long time to generate it. A long time. So, if we’re creating some code, writing some Rust or some Go lang, the AI will write it way faster. Right? And you can just make a few little tweaks. But the challenge is, “Oh, are we going to run into scenarios that we need to have very specific development in integration?”

Tim Veil:

So let me ask you about that, because actually this is fascinating, because I write in my job. I’m not paid to be a software developer at Cockroach, but I do for various reasons, continue to write a lot of software for side projects. And just to keep those muscles active. Right? I’m not yet ready for that to atrophy and go away for my daily life. But I know, because I see it in news sources that I follow, that people are using it to write code. What has been that experience for you or for y’all? Because I can see it being useful, for example … Even if it’s not perfect, maybe it gets me halfway there, or three quarters of the way there, or helps me just get down in an editor, some boiler-plate stuff, and I go have to add my special sauce. Has it been actually that useful? I’ve honestly never tried it. Now, as soon as we’re done, I may go, “Hey, write me some good old Java,” which I like. But, I mean, are you finding that it’s actually quite useful stuff?

Stephen Blum:

Yes. Absolutely. And it comes down to getting used to the tool. So now it’s we’re in the world today where it’s a tool that you can use to accelerate your job. I call them labor-saving device. Just like when the vacuum cleaner was invented and the toaster was invented. It’s a lot easier to accomplish your tasks day to day and it saves you a lot of time. And the trick is for GPT, large language model, you need to know what to ask it. And you need to know what kind of segments are important. And then you can work through those things and say, “Hey, I need to generate a cosign vector search,” because you want to do some full text search, but it a low volume. “Just create me a little simple algorithm in Rust,” and it will do it. It’s like, “Here’s the embedding function. And then here is the search function,” and it will write those for you. And you’re like, “Hey. Wait a second. Could those be more optimally written?” And guess what it does? It’ll write them more optimally, more fast and more efficient and say, “Hey, and if you use them like this, it’ll be even faster.” So you can have it actually do those algorithmic things for you. And then you piece it together. That’s where we are today.

Tim Veil:

So can I ask a dumb question. Hopefully, there aren’t too many people to hear this and say, “Geez! That was a dumb question, Tim.” But I guess, how does it know how to write optimized Rust? And maybe everybody else who might be hearing this understands this. And maybe you don’t know any more than I do, but I mean, what is happening behind the scenes? Do you have a sense of that? I mean, I get writing a description about the competitive differences between two databases. There’s a lot of material written on that. But, for it to generate the answer to a specific problem using a programming language … I get it’s a language, but I don’t know. Just, do you have a sense of how that’s happening behind the scenes? Or is this just all the wizardry that we just have to accept and understand as the new one?

Stephen Blum:

Oh, right behind the curtain. How does it work? It’s magic!

Tim Veil:

Tell me. Tell me, Stephen. Tell me. It’s behind the curtain.

Stephen Blum:

It’s a pattern. So, it’s all a pattern. Every language is the same, it’s just a different syntax. Right? And so you say, would you rather do a four loop with 1,000 iterations. That’s going to be the same in every language. Or, would you write it slightly differently, and it doesn’t really matter what the language is. Maybe you have your indexing a particular way in using smart hashing of some sort. So, that’s basically what it does. And all those design patterns are identical. And then you can just walk through them.

Tim Veil:

But what is the source material there? I mean, I guess I was thinking … And again, I haven’t done the reading, so I’m exposing myself as actually the true idiot that I am. But, is it taking as input APIs and definitions of these programming languages? Is it scouring the interwebs for publicly available GitHub code that has this language [inaudible 00:09:52]? I mean, where is it-

Stephen Blum:

Yeah. Stack [inaudible 00:09:54] GitHub.

Tim Veil:

I’m just curious. Yeah. Is that the source material that’s helping learn and train the language model for these things?

Stephen Blum:

So open AI is definitely the big one in the news these days. They just released ChatGPT-4. And, if you have the ChatGPT Plus, where you pay 20 bucks a month, you can get the preview access to it now. And it gives you a much larger input. So version three gave you 4,000 tokens. Version four gives you 32,000 tokens, which means you can put giant documents in. Or huge pages of code and have it do specific tasks against those. So the bigger it gets, the more work it can do. And that means it can do work on large code bases now versus smaller ones.

Tim Veil:

It’s funny, just sitting here thinking, so I’m pretty active in that, as I said, I write a lot of code. But I end up building a lot of almost utility stuff that I find useful in different things. And I have been working on this tool to auto-generate or auto-populate databases, based on analyzing the schema that a code base can inspect from say, for example, a JDBC endpoint. So, in other words, connect to some database, inspect the metadata, and then fully populate based on rules, et cetera, the full schema, including the referential integrity, et cetera, et cetera. Well, I’ve been working on this for a little while and I have run into this block, where I can’t quite figure out how I want to solve this problem. And it’s one of those things where it’s just me working on it and I’ve hit this wall. It’s like writer’s block, I can’t get beyond it. And so I haven’t worked on this project in a while, because I cannot reason about a solution to this particular problem I’m having. I’m wondering if that’s the kind of thing that I ought to like, “Hey.” And I can’t even remember where I got stuck now. mean, I kind of do, but it’s like I wonder if that’s the kind of thing that might be useful, even if it just gave me a hint at how to maybe solve this problem.

Stephen Blum:

[inaudible 00:12:06] And it can be a good idea to [inaudible 00:12:09]. And Tim, you and I were actually talking last week. We were talking about a neat idea, where you could ask it to just run a query against the database. And now even with large schema sizes, you could say, “Hey, GPT? Here’s the database schema. Can you write me a query that does X, Y and Z?” And guess what it will do? And it will pass you back the query and then you just pass that off to your Cockroach database. And two API calls, boom! You got your answer.

Tim Veil:

Yeah. We talked about it and I kept meaning to go explore it. I think that would be such a fascinating way to interact with the database, because right now, interestingly enough, we ask a lot of questions about our own data, our customers, our prospects, who’s viewing our docs, who’s creating issues in our GitHub. There’s so much data out there. Right? And if we could centralize it, that would be an interesting thing. But then, just having the understanding of the domain model so that you can write intelligent queries and associate different pieces of data together to find interesting patterns. That’s a non-trivial exercise. But think about the power of having something like the data accessible in a domain model schema that’s well understood. But, instead of having to have somebody sitting there and banging away at a sequel editor writing complex queries, you simply say, “Hey, how many critical issues were created over the last quarter?” Yeah. To me it would be … And I’m sure there are many people already working on it. But having ChatGPT or something like that as an interface to a database, much like we treat standards today like JDBC or other things. That’s the gateway to the database instead of these other protocols, would be super fascinating.

Stephen Blum:

Tim, I think you just eliminated a few data analyst jobs just now.

Tim Veil:

[inaudible 00:14:20] I’m not empowered to have those conversations. Not at Cockroach. But, yeah. I mean, wouldn’t it be interesting, though, because at the end of the day there’s some data that you want to have completely accessible. And this is the promise in a lot of ways of lots of tools and careers, is to make data to somewhat extent accessible, so that people not only know what questions to ask, but perhaps spend enough time in the data just to learn what questions are interesting to think about. And sometimes there’s a barrier to entry to that. I don’t know enough to write to explore the data, to understand what questions to ask of it. If we could solve that and make data truly accessible to people who knew nothing more than just common English language, it would be really, be really powerful stuff.

Stephen Blum:

That is interesting. One of the hard problems with data, in general, is even knowing what question to ask from the data.

Tim Veil:

That’s exactly right. Exactly. I think that’s what I see just in some of the internal work that we have. That’s what we struggle with more than anything. I can get you the data if I knew what question to ask, right? But I’m not sure what the question is, because I don’t know what the story the data is telling. But I want to pause this train, because I like this. First of all, I like this, this train of thought. We may come back to that. But I want to give you an opportunity because folks listening may not understand what PubNub does. And I want to you to have an opportunity to talk about … because I spent, since you and I last spoke, doing some research on what you all have been doing. You’ve been at it for a little while, but tell us a little bit about what you guys are doing, because I think it’s a really fascinating … Well, your company’s very fascinating, the problems that you’re solving. There’s a lot to it and I want to hear about it. I want you to have an opportunity to tell the story, if you so desire. And then we can go back to talking about ChatGPT. I’d love to hear more about of what y’all are building.

Stephen Blum:

Yes. Absolutely. I’m excited to tell you about PubNub. We’re a developer API platform that allows multi-user collaboration. And so, when you’re ordering your food to be delivered and you want to track the order status, and you want to see where the car is, because it’s lunchtime and you need that burrito as fast as possible. This is what we help with. These are the things. Multiplayer games. Right? When you have people in the same realm or same arena, they need to be able to coordinate with each other and where they are and compete against in different teams. And this requires data communication over a wide variety of networks and platforms. And this is where we shine as well. Games will launch on Xbox and PlayStation and PC. And they can all work together seamlessly over the same network. And we power a lot of experiences. And this is what we do. And we’ve been doing it for 12 years. And, guess what? Back in the day I was so worried that, “Oh, how are we going to get to scale to 10 million users? 10 million users is so many users. How are we going to … ?” We’re over one billion users now on our network. It’s been a journey. It can tell you that, for sure. And this is where the AI comes into the picture, because we have so much data going through our network right now. And there’s so much opportunity to leverage some of these advanced technologies. The time is now, the time to do it.

Tim Veil:

The time is now. Strike while the iron is hot. All right. Right. So, before we talk about some of the advanced tech stuff, because I really do … 12 years, it seems like a long time these days. I don’t know why it strikes me as a long time. Maybe it is, maybe it isn’t. And maybe it’s a blink of an eye. But maybe if you can, what was the spark that got this thing started? 12 years ago, you sat down and said, all right, this is something we need to do. You said it’s been a journey and I want to understand a little bit more about that journey, because I think one of the most fascinating things just from an architectural perspective is, “Hey, I start with some idea and to get to a billion it looks maybe potentially different thing.” But, so the first question really is what was that thing that was like, “Hey, this is a problem we’ve got to solve right now. Nobody’s doing it,” or, “Somebody is and we can do it better.” Can you share what some of that kind of initial spark was to get you and the team going in the early days?

Stephen Blum:

Yeah. Well I could just tell you from personal experience. I like to play video games. I like to wake up late in the morning and not have to worry about anything. And I was worried that I might not be able to do that forever. So, how do I make it, so that’s possible I can do that every day for the rest of my life? Let’s start a business that just takes care of itself. And how do we do that? How do we build a business that’s automatic like that? Well, API company. Okay, that’s great. Let’s do a freemium business model, that’ll attract a lot of people without me having to do a lot of work and also gives you a lot of beta testers and QA testers, for free. Right? So that started to spiral. Okay, well, if we just go this path and then there are multiple business ideas. Let’s go down this connection path. Right? I need something that can allow us to have multiple-user communication. This is when iPhones were just starting. They didn’t even have push notifications back then, when we got started. There was no way to communicate between this device, other than SMS. So, how do we enable those developer-level experiences to build games, on-demand economy? There wasn’t even Uber. The way you ordered a taxi was, you called someone on a phone and said, “Hey, can I have a taxi? I’m at this address.” And they’re like, “Hmm. Maybe we’ll have someone there.”

Tim Veil:

I remember these days. I remember these days.

Stephen Blum:

So we-

Tim Veil:

Painful days. [inaudible 00:20:17] Painful.

Stephen Blum:

So-

Tim Veil:

Stone Age.

Stephen Blum:

Yeah. It feels like now, because it’s so much easier to just press the button, which is so great. I love that. And so myself and my co-founder Todd, we were working on ideas to help build this new world, this new way of communication. And we launched PubNub and API. Immediately got customers, immediately got paying customers. We were making more money than we were … So we had positive revenue. And then we realized that all this effort that we put into it is now leading to more effort. So, the success is breeding more work. So we had to hire more, we had to put more hours in and, years later, it’s still the same thing.

Tim Veil:

And here you thought, “I’m going just going to start this business and it’s going to allow me to sleep in and play video games.” And then all of a sudden the opposite happened. “Now, I have to work more. Now, I don’t get to sleep in.” Yeah.

Stephen Blum:

Exactly.

Tim Veil:

I’m sorry it happened that way for you.

Stephen Blum:

Yeah. That’s how it works.

Tim Veil:

It sounds like it’s been a pretty good ride.

Stephen Blum:

Has been a good ride. And you get, along the way, all of these opportunities learn how business operations grows and scales. And you get to the big bees in terms of numbers and it ends up … You learn a lot and you gain these skills over the years and you realize you want to leverage these skills more and more. And it is fun and it’s satisfying to do that. And so now, that’s what I want to do on my day-to-day basis. I want to do these things, because I just got used to it. And it ends up being a little more exciting and rewarding than video games. Why? Because video games are like this box where you can just … You go in. Yeah, there’s some puzzles in there. They’re just too easy now. I’ll tell you what’s hard. What’s challenging? Business. There you go.

Tim Veil:

Yeah. Yeah.

Stephen Blum:

That’s the trick.

Tim Veil:

What video games, just out of curiosity, were you a big fan of? Or are you fan now? Or is that not something that-

Stephen Blum:

Whoa!

Tim Veil:

… that you would share?

Stephen Blum:

No. I mean, I’ll share. I love the metro games. I love those.

Tim Veil:

Oh, yeah.

Stephen Blum:

Those are great.

Tim Veil:

Original metro games. That’s like, that’s old school, right there.

Stephen Blum:

There you go. And I like those. Yeah. And just the Megaman games.

Tim Veil:

Yeah? Old school stuff. Right?

Stephen Blum:

Mm-hmm. I was very much into those. And that’s what got me into software development, was gaming. I was like, “I want to make games like those too. I want to build [inaudible 00:22:47]

Tim Veil:

Have you ever built a game?

Stephen Blum:

I have. Yeah. I’ve made several games in high school and college. The internet was different back then, say, say. It’s not a good distribution model. Games was all CDs and things like that. So, it didn’t turn those into businesses.

Tim Veil:

So if we go back to kind of the journey. And again, I find this stuff so fascinating talking to the folks who have started, founded businesses and ultimately made them very successful. I mean, we talk about it from an application architecture perspective. Obviously, you’re sitting here, “Hey, we just got to get started.” I’m assuming, so correct me if I’m wrong. “Hey, we got to get started. Here’s an MVP,” or, “our first release.” Maybe you did, but I can’t imagine on day one or day 90 you expected, from an architecture perspective to support a billion users? I’m sure your goals were more modest initially. Where I’m going is, what were some of hose hurdles you had to cross, where like, “Hey, we built something that supported X number of users. And now, wait a minute. To take it to the next level, we have to get here.” I mean, what was that journey? Were there architectural evolutions of the product that got you from day one to, here we are 12 years later with billions of users,”?

Stephen Blum:

It’s a lot of iteration. Yes. Joining Todd in his dining room on a whiteboard. We drew out what this would look like if it were to scale to the world, scale to the moon. We realized high replication, globally distributed, CDN-style data availability. And there’s a big difference with how PubNub works, versus a content delivery network. We’re a data push. We send data direct to devices. Usually these mobile devices that have networks that can change from Wi-Fi to four and 5G. We need to still get the message to that device. Now, how do we do that? And reliably? Well, high replication, eventual consistency. That’s the model that we chose. So when you send a message into PubNub to be delivered to a target device, that message gets copied multiple times per region. And we have 17 regions deployed globally. And, that means your message is available in all those regions. So, anyone nearby will be connected to that region and be able to receive it directly from that region. So, this gives us the fan-out model, which you can mass broadcast. A lot of events, sporting events, live national television events, use us for that type of capability. And that’s what helped us scale is those types of major media events has built and built our system.

Tim Veil:

So, are you saying in this dining room, in this dining-room setting in front of the whiteboard, you guys initially said, “Look, we know we’re going to make this,” or, “We know we want this to be global. Here’s what this looks like.” So you started out thinking big, as opposed to like, “Hey, let me just, I don’t know, 100,000 people I want to serve. Let’s get something.” Well, I mean, that is smart. Right? Because I think that’s the challenge we see a lot of times. Certainly from our perspective at a database company oftentimes working with people who are transitioning from one stage of growth to the next, for a variety of reasons. “I’m here and I want to be here. I want to be global.” A lot of times people don’t take the right steps or design really for that kind of growth early on. And that transition from an early vision to what my next vision is, is often a painful transition for folks. So, it’s really good that there are people out there that are like, “Hey, wait a minute. Yeah. We got big plans. We’re going to start early.” In Looking back are the things, though, are there things you could go back and tell yourself, back then? Like whisper in your ear? Like, “Hey, architecturally we had the right idea, but this is something we shouldn’t have done or should do.” I mean, anything that you look back on like that and say, “Boy I wish I knew then what I know now,”?

Stephen Blum:

Oh, man. From an architectural perspective, we had a really good start. Everything had to follow the model that we had built with our message bus. So, global distribution can operate in an always-on model, like the concept of no backups. Right? I don’t want to pay for hardware that just sits there as a backup. That hardware is active, I’m paying for it right now. It needs to be in the pool of live, active traffic and say you need to build your software so it’s fault tolerant. If one piece of hardware fails, it needs to route to the next one. That’s a lot of the code that we wrote, is built for that level of tolerance. That was the hard, tricky IP that we wrote for our system, for this message bus, and for high reliability of message delivery. That model had to copy to all our other subsystems, which are present detection. I need to know if that device is still active right now, this very second. Is that device online? Can I send that mess? Can I send a message to that device? Yes, I can. It’s online. So, that’s the model that we took and that’s how we started off. Do I have regrets? Did we miss anything? There’s a couple of things. But, we spent too much time building one particular component. So, we have a component called PubNub functions. Now, this is a JavaScript environment that allows you to program the messages as they’re going through the network. The requirement was minimal to no latency. Right? And so our latency in processing messages means that that processor has to be right at that same region where the ingress is. So we have to have globally distributed compute for that to happen. But, when we got started, the team chose an event bus called Kafka, which is a lot better these days. But back then it was brand new and it was like the way to get going. But the design didn’t match the latency requirements. It took 26 seconds to process the data, which is a little bit slow. And we spent too long on that, a year plus. And at the end of the day it just couldn’t meet the needs. So, we rechanged the model to operate in a JavaScript environment that set side by side with the message bus in the load balancer. So, I guess, just time was lost in that case.

Tim Veil:

One thing I sometimes wonder about … I don’t think this is part of it, maybe it is. I observe not only sometimes internally, but certainly externally when we’re talking to customers, is that if people get stuck, … And you’ll hear this term a lot. “Don’t let perfect be the enemy of good,” kind of thing, where you spend an awful lot of time … In coding, we used to call it like pre-opt optimization. But is this something y’all have struggled with? I think a lot of people do. I certainly have in different roles I’ve had where it’s like, you so desperately, because you have a grand vision … So desperately want to get it right and make it perfect now. But to your point, you can. You can lose valuable time, whether it’s over engineering or overthinking solutions when actually the best course of action is just to get … Because there is no better teacher than real world, real people using the system, real experience. But sometimes I think that gets lost. I see teams, I see people sometimes just spending so much time trying to get it perfect. And they lose that. A window closes on them when that happens. Is that part of it or have you seen that? I mean, am I-

Stephen Blum:

Yeah. There is opportunity loss. If you don’t get to market quick and you don’t iterate quickly, you don’t get customers. I mean we did. That was the PubNub version 0.1. It wasn’t the globally distributed bottle. It needed to be after we got a lot of customers. But it started with one region, one data center, not replicated. And uh-oh, yeah. Not that fault-tolerant. But it did deliver messages. And it did the trick. And so, we needed a system that was much from more robust, much more scalable. So, that was a second system that we had to build.

Tim Veil:

It’s an exciting journey, isn’t it?

Stephen Blum:

It is. Yeah. And it’s just as many highs and as there are lows.

Tim Veil:

Speaking of lows, one of the things I think … I think you and I talked about, but certainly it’s been a topic of conversation for me even this week. I mean, I’m here for those of you who’ve maybe listened or watched previous podcasts. I’m in a different location. I’m up here in our New York City office, using the studio that we have here. Actually, interestingly enough, the office that I’m in and the studio I’m sitting in used to be our office in New York, used to be Peloton’s New York headquarters. And so, I think they used to shoot some … I think this is true. I could have this wrong, but I was led to believe they used to, at least in the room I’m sitting in now, shoot some of their promotional material, et cetera, et cetera. But this is the first time I’ve been back in New York in a long time. I mean, not since the pandemic. I’ve been here after that, but certainly it’s been a while. World’s been a crazy place over the last couple years. And what’s interesting, I think about your story, because some of the folks I’ve talked to more recently have been at their organizations a lot less time than you have. So, you had what would be eight, nine years of PubNub before the pandemic. Right? And then the pandemic hits and everybody’s world changes, in sometimes tragic ways. So, you guys had the opportunity, I would imagine, to have built a culture and a way to work. And I don’t know much about that. But, how has it been navigating the impact, being away from each other, not traveling, all this stuff? Not only maybe on company culture, but just perhaps even … Because I’ve talked to about this with other people. Just the ability to build a product or manage a product. It can be challenging when you’re used to doing it, for example, maybe in person. Can you comment maybe on how this has impacted y’all culturally. And then maybe from a product? Maybe it hasn’t, but-

Stephen Blum:

It’s different. So, going to a remote-first work culture from before it was an in-person first work culture,

Tim Veil:

So, what? So, y’all were in person, that was your thing. It was like, “Be there or be square,”?

Stephen Blum:

Yep.

Tim Veil:

Yeah. And I think we had a very similar thing, certainly in New York.

Stephen Blum:

Yeah. We’re here in South of Market, San Francisco, every day going to the office. Still had the meetings on the calendar, still met and still talked. There’s a lot more in-person, that walking over to desks and things like that. But there’s a facsimile to that. Today is Slack. We Slack, like use direct message, hey, walk over to their desk-type situation. It’s not the same, because you’re that human bandwidth, that emotion is gone. It’s basically gone. And that that’s a challenge. And so it becomes, “What words do I use to convey the right thought?”, as opposed to body language, which is a lot easier to read and a lot more familiar for humans. So that piece is missing.

Tim Veil:

I totally agree with you, by the way. I think it’s something I honestly, I don’t know, that we talk enough about, which is the importance of body language, subtle cues that you pick up on when you’re in person. Obviously, I see your face, you see mine. But I mean, there is so much that is missed, so much nuance over video, which is better than not video. But all the stuff that you lose in Slack. I agree. Slack’s been a great way to communicate and it does. There’s some things I like about it better. But, boy! The nuance. You can’t infer, or sometimes you shouldn’t infer tone and temperament for messages. And there’s part of me that thinks it’s created more havoc than it should. But, certainly reading someone’s body language, understanding and being able to respond that in an effective way, when you’re not an effective communicator, or communicate effectively with people, it can be incredibly difficult. And I know there’s all this stuff thrown about, are people more or less productive? I don’t think that’s the question I want to answer, because it’s a hard measurement. But I mean it’s almost like, “Do you like it better, or worse, from when everybody was there every day, to when we were all remote?” And maybe the answer is a little bit of both, but I’m just curious. I know it ends up getting distilled down to this productivity conversation, which doesn’t to me seem near as interesting as just, how does it make people feel?

Stephen Blum:

Well, I can tell you from my personal experience that working from home is too good. I’m going to keep it like this forever now.

Tim Veil:

Well, this is coming from a guy whose only reason for starting a business was to play more video games and wake up late. So, okay. Maybe you’re not the right guy to ask.

Stephen Blum:

It is fantastic. I do feel more productive personally, as well. I know the team feels the same. In fact, we have talked about this a lot in the team. They all want to stay remote, forever now. It’s just the new way of the world.

Tim Veil:

And is that fine for y’all? I mean, I know some people are really struggling with that right now. Like, “Hey, get back to the office, or stay.” I mean, is that going to be the new culture or is that still something that’s being debated hotly at PubNub right now,

Stephen Blum:

Just the managers want everyone in the office. So, I could tell they like that better. They like that better. I like this one company. They took all meetings off the calendar. Do you remember about this? They deleted every meeting. All meetings, all recurring meetings gone. The entire company, all meetings are gone. And so the developers and the product managers, they all knew what to do and they were very productive. The people that didn’t have anything to do with the managers, they’d be like, “Well, what do I do now? What am … ? There’s nothing to do. I’m not in a meeting. That’s a problem.” So there’s catches on either side.

Tim Veil:

Well, I know I’ve borrowed a good 40 minutes of your time, and that that’s probably far more than I deserve. So, maybe just as we wrap up here, I’ve been doing this with everybody else that we’ve had on the show. And perhaps it’s just because for us, this is the beginning of our fiscal year, so we’re very much in this mode right now at Cockroach, at looking ahead. Right? Looking at the year ahead of us and doing some planning and trying to get excited for what’s in front of us. What’s exciting you about the year ahead? What are you looking forward to? Are there big plans at PubNub? Are there big plans, personally? I mean, what’s the year look like for y’all? And I recognize that maybe for some, this is a question that’s two or three months old, but for us it feels like we’re just getting started for the year.

Stephen Blum:

Yeah. This is just beginning of the year. We have many, 12 product initiatives that we are planning to launch, at least eight of them this year. And we did get one out already called Events and Actions. And this allows you to bind on different events in the system and then shepherd those, that data, over to your system. So, you might have a Kafka bus or you might want that data to be sent to S3. Or it might want to trigger a web hook that hits an API on your server. So, as you send data through our system, we can then trigger data off onto your system. And that was a great release. We saw a nice uptick in usage when that occurred. But that’s not all. We have a lot more coming. We have a lot more.

Tim Veil:

But just wait. One more!

Stephen Blum:

Just wait. There’s more. And, we are doing some deep investment into our data. And I want to be more descriptive, but I’m being purposely vague. There are some significant advantages that, as you become a customer of PubNub’s platform, that we are going to expose to you that you can’t get even yourself. And there may be some things that we talked about at the beginning of this episode that’ll be exciting to watch out for.

Tim Veil:

I like that. A little mystery. A little excitement.

Stephen Blum:

Yeah.

Tim Veil:

Well, Steven, again, I thank you at the beginning for coming on. Thank you again for spending time with us today. I found this conversation incredibly exciting and illuminating. And I feel like I learned something. I also feel like I embarrassed myself by asking dumb questions, but that’s okay. It was really great having you on the show. I think what y’all have built is incredibly interesting and exciting. And if folks who are listening haven’t had an opportunity to go visit your website and read a little bit more about the product that you’re building, the problems you’re solving, definitely do that, because it’s really interesting stuff. So again, Stephen, thank you so much for joining us here on Big Ideas and App Architecture. It was a pleasure having you on. Yeah,

Stephen Blum:

Tim. Yeah. Great to be here. Thank you.

Tim Veil:

Thanks for listening to the Big Ideas and App Architecture podcast. If you’re a fan of the show, please subscribe on your favorite podcast platform to get every new episode. Also, subscribe to our YouTube page, linked in the description to watch every episode of Big Ideas. And if you want to join in on the conversation, reach out to podcast@cockroachlabs.com. We’ll talk to you next time. Thank you.