Transcript#
This transcript was generated automatically and may contain errors.
Welcome to the Test Set. Here, we talk with some of the brightest thinkers and tinkerers in statistical analysis, scientific computing, and machine learning — digging into what makes them tick, plus the insights, experiments, and OMG moments that shape the field.
In this episode, the Test Set hosts sit down with Tareef Kawaf. He's the CEO of Posit, but definitely not Kadly's boss. Tareef's known inside the company for finding bugs nobody else can, just from using the tools himself. We get into how AI is reshaping data science, and what he's learned about leading through constant uncertainty. Enjoy the show.
Hey, Tareef, welcome to the Test Set. I guess just by way of introduction, you're the CEO of Posit and my ultimate boss.
That's how I like to think of myself, yeah, generally.
But technically not my boss. That's important to recognize.
Technically, Hadley's cut off from the rest of the tree somehow, yeah, but...
That's good to know. That's actually a new piece of information I learned just now.
I think the advent of the new tools made him attached at some point, so that may no longer be true, but historically, he was unattached.
What an incredible bit of juicy gossip to start a podcast with. Some inside baseball.
Does Hadley report to Tareef? I guess we'll get back on that.
And in the meantime, just to introduce everyone, so I'm Michael Chao. I'm the host and a software engineer at Posit, and joined by co-host Wes McKinney, who's a principal architect at Posit and founder of Kenio, and then Hadley Wickham, chief scientist at Posit.
And so we have a very Posit-stacked crew, and honestly, so excited to be able to talk more, Tareef, about, you know, what I think a lot about your background and kind of how you approach things. I think in some of the conversation leading up to this, I feel like I actually learned so much about you that I actually had no idea about, so I'm really excited to kind of dive into stuff. But yeah, thanks so much for coming on.
Tareef as Posit's de facto data scientist
So one thing I noticed is you mentioned Hadley referred to you as Posit's only data scientist for a while. Yeah, it was an interesting moment. We were in a company meeting or something like that, and Hadley made that comment in front of the entire company, and I'm like, I don't think that's true. At the time, we didn't have a data science function at all. And even though we have people who are vastly more capable on statistics, data visualization, software engineers, you name it, across a variety of vectors, I think he was right in the sense that I was the one attempting to use our tools to answer questions using our data.
And so when I first joined the company, I was hyper aware of the fact that I didn't know what I, you know, there's a lot that I didn't know. And so I was started, so that I forced myself to use our tools to try and represent the information to the board. And so in the early days, because I was a first time CEO, they were asking to meet with me every six weeks, which was a very painful process. And so, but that gave me multiple repetitions, if you will, on using the tools and finding a slew of problems in terms of how I want the tools to work and, you know, where I would get stuck and whatnot.
And that is in and of itself an interesting set of stories about, you know, do people believe me when I tell them that this is a problem? Like, no, Tarif, nobody else is complaining about this. I don't know why. And, you know, over the years, I think I've gained more credibility with people that, like when I hit something, it is really a problem as opposed to just, you know, broken because I'm a, I have a CEO title. Tarif is infamous inside Pulsar that they were to find bugs in software that no one else has encountered, no matter how mature we might feel polished people think the software is Tarif always finds a problem.
It's a talent. I think it's like, I call them the sharp edges because you don't see them, but somebody, you know, somebody will, will bump into them. And some people, it seems are better at finding the sharp edges than others just, yeah, they have it. They have a knack for it, even if not, not, not, not, not intentional. Yeah. It's both good and bad. Like the good news is if something's working well for me, then it's probably going to work really well for the vast majority of humanity. Right. But just because it's working really well for our teams or, you know, even many of our customers, it doesn't mean that like, it's going to work properly for me.
And so it's, it's, those would be frustrating nights. Cause sometimes I'm like, all right, I need to get this deck out to the board and I've just hit some kind of weird edge case. And so sometimes I'd be like sending Slack messages to some of our engineers on the West coast that were still awake that I'd be like, listen, I'm trapped. I don't know what's going on. They're like, well, try this, try that. Oh, that's a good bug. I'm going to file that. I'm like, great.
The funny slash true thing. I'm sure for, for both Hadley. I don't know. I know certainly for me and probably also for Hadley, like, you know, especially in the early days of, of building pandas, like I lived in constant fear of like breaking people's workflows or causing, causing bugs. Like I was always worried about, like, I don't want to be in the news about like, there was a bug in pandas that caused like there to be the equivalent of an Excel error that brings down some business. And then it's, it's blamed on me like, Oh, that West, he didn't check the code properly. Didn't write enough tests. And yeah. And it caused a, caused a business error. You know, bugs are unavoidable, but we try to make, try to at least make things correct. There may be usability problems, but correctness come feels like comes, comes first.
How AI has changed the way Tareef analyzes data
No, it's super interesting. I feel like maybe like something that reminds me of, it's like, so you were doing a lot of data science and you still are, but when you came in as CEO, right away you started analyzing data. And today we have a lot of tools like cloud code and AI and LLMs. I saw you mentioned a little bit that kind of like in general, the like choice of programming language maybe has, has become a bit less important. And I'm sure the, the ways you're analyzing data have changed. I'm curious about like, what does it look like today for you to analyze data versus like a couple of years ago?
Yeah. So, I mean, a couple of years ago, you'd have to write the code by hand. At this point, a lot of times what I'll do is I'll, I'll, I usually use cloud and I usually tell, I, I, I've learned that I need to tell, be very explicit with cloud, cloud. I'm like, I want you to use this programming language. I want you to use these types of libraries. And otherwise it'll just give me, otherwise it'll give me stuff that it decides is valuable for it. So yeah, I can be very, I'm very specific. I try to be very prescriptive about like what I want. For example, like if you tell it, give me a table, it'll give you a table, but it won't use GT. And so I want, for me, I like GT. I like the formatting of GT or great tables on the Python side of the world. Right? So like, I, I'm, I try to be more specific now, whereas in the beginning, when I first started playing with cloud, I'd be like, here's the data, here's what I want to do, or here's a, here's an application I want to build. And then if you don't specify that, if you don't specify that, then it becomes more, more complicated to sort of unwind later for me.
But, but for me, I, I now tell it like, here's the directory I want, put it, put the code in here. Cause that's in five Oh, or something like that. It started spinning up sandboxes, which I'm like, I don't want you to spin up a sandbox. I, I have a, I have a local runtime I want to be able to use and, and go from there. But I, I, once you get it to, once you have a, if you have a project that's configured, it's a little bit magical to be able to go and say, okay, I want to do this. I want to do that. And I usually have Positron open now instead of RStudio and I can see, I can see what's going on.
Yeah. Yeah. Cool. And yeah, I guess happy. And what's I'm curious about your workflows too, with things like cloud code. I know like tariff, you mentioned that you're kind of particular about wanting to use tools. I'm curious for you to also like how particular you are about like the language or the tool that gets used and what that looks like for you all. You know, like in the early days of tariff using the tidyverse, for example, he had hit these bugs. And the only thing he could do was basically like, you know, tell me or file an issue. One of the things that's really interesting about open source software today is like totally feasible that today, if he, if, if tariff was doing this today, he could tell his agent like, Hey, I just discovered a bug in deep liar. Like, can you fix it for me? And that, that, that to me feel, I've been, you know, I've been trying to think like, what's the value like is open source software is still valuable today. Is it still meaningful? And that, that, that's clearly like a place where like that, that feels really valuable compared to having proprietary software that you are no longer blocked. Like this has always been like the pitch for open source software, like, Hey, you encounter a bug, you can fix it. Obviously like that, that's causing a bunch of like downstream issues for open source maintainers. But that is like something that's like legitimately cool about using open source software that plus with agents, now you're empowered as a data scientist to fix the bugs and the tools that you use. There's a lot more forks now.
like legitimately cool about using open source software that plus with agents, now you're empowered as a data scientist to fix the bugs and the tools that you use.
Yeah. I actually just did this yesterday. I am, I read a lot of books now on Kindle because that way my wife doesn't know that I'm buying all these books and and I highlight the books and as, as you may know, Kindle highlights are available online. Uh, so when I went to press the download to sync things, it wasn't working. It was like, put me through the off flow multiple times. And so then I just brought up Claude and I said, Hey, here's, here's the plugin that I'm using. I'm having this. Yeah. And it, and it did. And what it did, it actually went out, it found a branch. It was like, there were 300 branches or something like that now on the project. And one of them has the fix on it. And so it's like, I didn't take the fix as is, I confirmed that this was really the problem. And, and now like I've, I've, uh, you can just go ahead and follow the steps. And it gave me the steps to do it and whatnot. And then unblocked me. And I was like, this is really nice. And this would not have been possible. Um, uh, whatever it's nine months ago, 10 months ago that I don't think that would have worked.
Yeah. It's so interesting of just asking Claude to fix the problem and then see what happened. I mean, it reminds me that Hadley, this must've been a year ago. Like one of the things he said was like, take a task you absolutely dread doing and just ask AI to, to fix it. And I feel like syncing something is like, I just want it to happen kind of. And if you have, it's like so frustrating that I feel like AI solving it. It's like such a incredible, powerful thing to see.
Basically, I find it also great. Like I use Posit Assistant as my primary agent harness, mostly, I mean, just cause it's good at the hand, because I want to use our own products. And now like, it's easier for me to file, like to fix the problem that it is to like report, like this, there's obviously this obviously clear vibes that I'm like, this is a mistake. Everyone's going to agree with me. I'll just file a pull request and fix it. There's still some things where I'm like, maybe this is just a distant difference of taste or belief about what this product needs to do. And I'll still discuss with someone, but that's been like, I don't, I know nothing about how Posit Assistant is implemented. I don't know what I don't, I think it probably is implemented in TypeScript. I don't honestly know, but it feels like I can still meaningfully contribute to that project, because I can see what the problem is, know what it should be doing, and then tell my agent to go do that, which is pretty magical.
I've built a growing stack of AI developer tools since December. And I've been largely building all of those tools in Go for a variety of reasons, fast to develop, fast to compile, fast to test, static binary, easy to install. You just build the binary, just copy it, put the binary where it needs to go. But it's been interesting because I don't, I'm not an expert Go programmer. And so I can kind of read the code. But it also creates this weird issue of like, how do I deal with inbound pull requests. And I do receive a lot of pull requests in these new projects. Like AgentsVue, for example, gets a ton of pull requests. So it's more or less like a database, like a system of record for all of your AI session, your coding agent session data, your harnesses. And on the whole, I know there are some open source maintainers that basically just want prompts, like, just tell me what's wrong, and I'll fix it. I mean, I still use Python for data analysis tasks and scripting. And I've got a variety of personal finance related stuff, like just finished collecting all the stuff for my 2025 taxes, because I'm, for reasons, I had to file an extension. And so there was some data analysis involved and data reconciliation. And just to be able to unleash Claude to say, do my taxes for me, and to write the Python code and to handle all of the data reconciliation. And it's, it's, it's, it's truly tremendous also, like to be kind of unburdened from even like the language choice. And, and, and like, if you say, like, oh, there's a library that I need, but it's in a language that I don't know, but I need to solve this problem that I have. And I need this one component. You know, I'm not that good at R programming, it turns out. And so like, I can use R a lot more now, because I have my agent friends to, to write code for me. So I think is like, just an unblocker and like opening up the universe of software to people. And so like any person can use any programming language, any dependency or library and build a solution for themselves. With like, as long as they have, they know what they want to build, like they have the right intent. And they know what the shape of the solution should look like, and like how it should work and how it should feel to use those turns out to be the hard problems. Like, it's
amazing how many people, they might identify a problem, but they can't conceive of what the solution should look like. And if you can't explain to the agent, what you want the solution to look like, you know, you might present the problem, the agent will suggest solutions, but the solutions may not be very good. They agents have to have terrible, terrible taste in general.
Yeah, I agree with us. But I think there's one thing that I like the worry that I have right now is that there are a lot of like you are using the agent as a co worker as a and you're still retaining your ability to judge and think through things and whatnot. And one of the big worries that I have right now is that I see increasingly a lot of people just like abdicating the judgment and the deeper thinking to to the machine. And the problem is that the machine like will make bad decisions or will not even know that it's making decisions. It's like, Oh, great, you can do this. It's like, Well, is that really what you want to do is, you know, like being able to still think and reason about the problem is, is very important. And I don't, I have not figured out a technique yet to force folks to inspect the work and confirm that this is actually the right thing to do that that approach, it makes sense.
And I think if somebody said this, maybe it was you Hadley, at one point, there was somebody who said something like a like LLMs, you know, sort of an exponent on on who you are and what you are. So if you're if you're already a greater than one, then you will that that, you know, that exponential is going to grow you dramatically. But if you're just a little bit under one, then that, you know, the greater power is going to make you a much smaller impact.
there was somebody who said something like a like LLMs, you know, sort of an exponent on on who you are and what you are. So if you're if you're already a greater than one, then you will that that, you know, that exponential is going to grow you dramatically. But if you're just a little bit under one, then that, you know, the greater power is going to make you a much smaller impact.
And so that's, that to me is almost the most central thing that when you think through, and we've been having a lot of debates internally about like, what is, what is the role of, you know, code in this, in this new universe, right? And how do you think about that? And to me, what I like is, is this idea that code ends up capturing the actual lineage of the thinking, if you will. So like other people could look at it later and see, okay, great, I can see exactly where you made, where you made a mistake in your thinking, right? So like, that's, that's really important, I think. And, and a lot of people are just like, I ask a question, I drop it into Claude or chat GPT, or whatever, I get an answer, and I plug it in, and we go, right, which is very much not what we have, as a company, I've been saying is the right way, the right path forward, pre AI, right, just in terms of like humans, putting something in an Excel sheet, getting a plot, cutting and pasting it, putting into a PowerPoint, and shipping it to the board, right? And you're like, wait, how did we get here? Let's go back to that spreadsheet. Let's see. Oh, there's an error there, and so on, so forth.
With code, what I love now is that you could potentially bring up another agent and say, hey, tell me what's wrong. Tell me where what could what could be wrong here. And just, you know, even if it's 70% incorrect, that 30% of it, you know, might give you really good insights in terms of like, how you're representing your code, how you're representing your data, how you're representing the point that you're trying to make. So yeah, I mean, I can, I completely agree. I mean, there's clearly a class of AI users who are more or less disengaging from the details, like of what they're doing, and delegating the judgment and decision making and taste to, to the agents make maybe they've been told by there is a lot of marketing from the frontier labs and from the AI industry to basically tell people like that these models are so unbelievably intelligent, and they can hack computers, and they can do all these things. And, and so if you want to build something that is good, that is exceptional, like it requires engagement with the details.
I think another thing that that I worry about a lot is that that it's clear that that the frontier models are getting worse and worse at communicating with humans as time goes on. They're more verbose, like their quality of like explaining what they're doing has gotten has gotten really bad. And so this makes it just a lot harder for the human operator, like even a very skilled operator to decipher, like what the agent is actually going to do in response to your prompt. And so I feel like we're probably, you know, in the harnesses, like with Posit's system, and like, you know, just thinking about our tooling, like, you know, we're probably all gonna have to invest in in tools to help with like, you'd think that the frontier labs would do it for us, but to essentially decipher and like, you know, before an action is taken, like, explain to me in people speak in human language, what is going to happen here so that I can judge because if you're presented with a wall of text, like 1000 words, and Claude ease, like gobbledygook, it's like in this wedge and load bearing seam, you know, there's this or that vertical slice be, you know, it's, it's become kind of incomprehensible in places. And so that makes it just much harder for the person who's just trying to get work done to know, like, is this what I what I want? Is this the right thing? So I don't know the right answer.
Your argument was, is that maybe some of the people disengaging their brain is like, it's just too much. And I can, I think it's good enough. It's likely to be good enough. So let me just ship it.
Well, I think a lot of people have decision fatigue, because in the past, you know, as a as a worker, you wouldn't, you wouldn't have to make that many decisions in a given in a given day, or maybe decisions will be made in meetings or with your colleagues. And so I think what's happening is that that a lot of people are being handed, handed Claude code or handed, handed codex and said, you know, you don't need, you know, all that overhead that we used to have the meetings where we would align, and we would reach decisions, and we would decide to do things like you just work with the agent and figure it out and solve the problem. And so individuals who previously might have to make like a few critical decisions in a day are suddenly being asked to make like 50 decisions or 100 decisions. And it is like, like, for me, it is extremely fatiguing to to have to make that many decisions, like architecture decisions, design decisions.
And there's definitely days where I, I close, I close Claude, I close codex, because like, I, you know, simply cannot think about the software anymore. Like, and I'm not willing to just press okay, and say, whatever you say, Claude, like, your recommendation sounds fine. And a lot of people, if they're under time pressure to get things done, or they're reaching their decision fatigue at some point during the day, and then they're just kind of capitulating. And so just like accepting the recommendations from that point onward, which is that where the danger zone lies.
What Posit is and why AI haunts it
Wow, I think maybe, maybe to really, because I know, I know, we've talked a lot about LLMs and AI. And I do think the decision fatigue is such a big piece. But maybe to make sure we really hit some of the key pieces, I wonder if to put a really like, fine point on what people are, are maybe coming here for to hear from you, Tarif. Like, maybe we could just talk really briefly, maybe you could tell us really briefly about just the basics, like what, maybe just to zoom out, like thousand foot view, what's Posit? What do you do as CEO? Why are you haunted by AI?
To me, Posit is a company that was, that's only really possible because JJ Allaire founded a company called RStudio and said, hey, we want to try and create an organization where creators can come and do really good work. And we build something for the long haul. And we get, you know, we treat people the way we want to be treated. And we trust folks to sort of do really good work and, and keep the sort of the, the whole spectrum of our customers in mind. So like, we want to be able to, these tools, we believe are very important for humanity to have. And so having the ability to fund that for decades is really something that is important to the folks who are here. Right. And so, so that is, you know, and I think we've generally succeeded at building such an organization. I think Posit right now has, you know, contributed probably well over a hundred million dollars to open source software. And, and every year we're able to add a meaningful chunk of change to that equation, which I think is really good and punches well above our weight in terms of a company of 350 people. Right.
So if you think about our history, that's, that's really wonderful. And, and when I, when I, when I was my two job offers at the time, and I told my wife, I'm like, this is what's powerful about this job is that like, even I'm out of work and year two, you've still made a difference in the world. You've added something that is of value. And so for whoever, you know, when people come to this company, they contribute for some window time and, and, and hopefully they feel like they've helped move that mission forward.
My job as a CEO, I, I feel like, I feel like my life is mostly about problem solving, which I enjoy and influencing, which I some mostly enjoy. Sometimes I don't enjoy so much. I think AI is top of mind for me because it has disrupted several things simultaneously. Right. Like, and so one of the most obvious things is disrupted how people think about the value of what we provide as a company. And so that means like, what is, you know, like you tell people like, Oh, we are a company that built the RStudio IDE. And some people will say, well, do IDEs matter? Is that even, is that even a thing? Right. And we're like, all right. Or, or we are a company that believes in code first and you should be able to document things like, well, no, I just ask Claude and I get my answer. I don't know why this needs to exist. And so being able to help everybody understand like, why does Posit exist? What value does it provide? How do you think about the investments that you make on the open source side of the world is, is something that is, is impacting a lot of my time and my energy.
And then the second piece is like, Hey, how do you help people adapt to the, to the rate of change? And so the conversation we were just having earlier about decision fatigue is an interesting way to look at the world. And, and actually, frankly, makes me feel a little bit better about the complexity. Like I have been feeling like I am failing and helping people see what I, what I believe to be very clearly, like here's the new world and we have to live in the new world. And you may say, I want to have the old world, but that's not really available to you anymore. We have to adapt to the universe as we know it. And we each have to accept that our roles change and as, as, you know, meaningful chunks of what we used to do by hand no longer need to be done by hand. And yet I don't want people to also let go of their cognitive powers and, and, and their ability to judge things properly.
And so, so right now I'm like, I think I would describe us as like, we are on this journey to sort of help figure out where and how and when, and how do you provide the right guardrails for people to use this really powerful technology in ways that are actually ultimately, you know, are responsible. So I've been posting about, you know, John Chambers is prime directive and people are like, who knows who John Chambers is, you know, but, but like, for me, that's a, that's a name that's been bandied around in our company since day one, like 13 and a half years ago was the first time I ever heard of it. JJ is like, you have to read the prime directive. You have to understand what that is. And so that has driven all the work that we do. And, and I believe that that still has a role to play in society and, and as part of the tool chain for both individuals and for corporations.
John Chambers' prime directive
Yeah. Do you mind explaining a bit what the Prime Directive is for people who aren't familiar?
I feel like Hadley should explain it because he actually knows John Chambers and it would be weird to explain it. Yeah. And maybe who John Chambers is.
It's funny that you say that, Tareef, because I feel like the Prime Directive had much more play in JJ and Joe's heads than it ever did in mine.
So it was really interesting for me because John Chambers was the person who created S, which is the precursor to R, which was originally developed to be an interactive statistics environment for people who are formally having to compile Fortran code to do analysis. So John Chambers was very, very influential in the field of data analysis and statistics. And I guess I'd read the book where he mentioned the Prime Directive, but it didn't stick in my mind at all. I guess maybe because it was just self-evident, but it was really interesting to me to see that JJ and Joe found it.
So I'll take a stab at describing it, or at least my interpretation of it, which is that I think that the primary goal of analysis is to produce reproducible code that is not only correct, but can be proven to be correct or can be less inspected and critiqued.
So I've been thinking about this a lot recently while trying to think about what's the difference between software engineering and data science and how AI is impacting them differently. Because I think this is one of the key distinctions. When you're doing software engineering, you were both creating the problem or defining the problem and then solving the problem. With statistics or data science, you're typically trying to understand something about the external world. So while in software engineering, because you defined the problem, you can tell if you've solved the problem correctly. You can write tests to verify that. When you're doing a data analysis, you actually have no way to tell if you've come up with the right number. So the best that you can do is to inspect the process that you used. And that's why having code, that's where the prime director is so important.
Yeah. It's interesting to me that you're surprised at how much it impacted JJ and Joe.
Because for me, it also impacted me pretty dramatically because I was like, okay, I get it. I get it. And the key concept here, Michael, is that he says something about how powerful the tools are. And so the human, when they look at that answer, which actually mirrors some of what we're talking about with AI. You get an answer. You're like, this thing is really powerful. I'm just going to believe it and go, you see what I'm saying?
And so like, there's this, there's, there's an interesting parallel to me about this, this problem, because it's, it's difficult. You see a plot. It seems really believable. You're going to take it and run. Yeah. It's interesting. Like what, yeah. What role does the sort of process and being able to audit what like generated the numbers play and do I, and to Wes's point of decision fatigue, like, do I as a person want to expend that effort or how do I choose where to put my attention?
I mean, I think one thing that's, that's related to this, that's interesting, you know, potentially there's implications for data science, but, but I think like AI and formal methods like theorem proving like lean in particular has created this whole Renaissance and in mathematics and pure pure mathematics research and essentially being able to do like folks like Terence Tao doing AI assisted, like grinding on hard problems and coming up with solutions, using, using lean as the engine to be able to construct and verify, verify proofs.
And so, you know, I think there's, you know, one of the, one of the optimistic in the optimistic views towards data science and in the air AI era is that we can build systems to help us both have like better understanding and confidence in, in this scientific methods and analysis that we're doing and, and to be able to like verify, verify and check that, that like there are not errors in the methodology. And because I think that having the verification systems are even more important for the agents because they're generating, you know, dozens of iterations or hundreds of iterations of there needs to be some feedback, some way to check and verify that, that, that you know, the right invariants are being respected and I'm not sure what, what formal methods and verification looks like for data science, probably a little different from, from like number theory, but something I think about, yeah.
I think the number of iterations is also like interesting because like for software engineering, it doesn't matter how many attempts you make to solve the problem, but when you're doing data science, if you're just trying different technique after different technique, you know, eventually you're going to find the answer that you want. And so the number of iterations is actually like really important. And I think that's something, you know, we start to kind of grapple with, like if it's now so much easier to try 250 different statistical methods, like that, that's going to have a real impact on like science and do we really believe these results?
I think there's a lot, there's a very big difference in saying I've discovered this after trying the most obvious method that occurred to me versus I discovered this after trying 250 different techniques that an agent came up with. And I'm sure you like, you can tell an agent like, Hey, this is what I believe to be true. Like figure out the statistics to, to prove that like. It's like PPP hacking on steroids. Yeah, exactly.
I think there's a very big difference in saying I've discovered this after trying the most obvious method that occurred to me versus I discovered this after trying 250 different techniques that an agent came up with.
This kind of reminds me Tarif of also like the talk about verification and in data science, like it can be hard to even have like a ground truth. Like you might just want to inspect the process and, and try to like do the best with a model.
Imposter syndrome and trusting your gut
I know one, one thing that caught me that you said when preparing for this was that I think, I think there was a question about if you've experienced imposter syndrome. And I think you mentioned something about like a version of imposter syndrome.
So yeah, I, I did not understand that I had any imposter syndrome problems until about a year and a half ago when I started realizing that there's a phenomenon that I've been going through, which I think can be classified under imposter syndrome. So, so for me, when JJ and I met, he was like, listen, you should run a company. And I'm like, listen, I don't know how to run a company. My background is running software engineering teams operations, running a live service essentially. And my thesis was like, listen, I'm going to do two more startups and be done. Right. And, but meeting JJ, we had such a good chat for two hours. We covered everything under the sun. And so I walked out thinking maybe I should attempt because I'd like to work with this guy. I'd like to, you know, I'd like to see if we can create something that can create value for, for maybe hopefully other generations.
But I always knew that I had this problem where I was like, all right, I'm a first time. I'm first time CEO. I'm a first time president of the time. JJ's like, take the CEO title. I'm like, no, no, you keep the CEO title. I'll just be the president. And my title on Slack was like the accidental president, like, because it really was not part of my plan. It was not my intention.
And so I would, you know, I would listen very carefully to people who have done the job before. And, you know, people who are older than me and maybe people who have MBAs or whatever. And and over the last, I'd say four or five years, I've realized that like some of the decisions I have made or pulled the trigger on were not really things that I believed in. I did them because I sort of trusted that like people know more than I do and therefore I should listen to what they have to say, which I think is a variant of the imposter syndrome where you're like, maybe I'm not right. Maybe I should just take somebody else's advice. And in hindsight, I'm like, that was those are all have been very expensive mistakes. And it wasn't until about a year and a half ago that I started putting it all together. I'm like, all right, these are these are all similar. They all have the similar root cause.
And, you know, I hired this person when I didn't believe that they're the right person. We made this pricing change when I feel like, you know, saying like there's a bunch of these little things along the way. I'm like, why? Why did that happen? Why did I do it? You know, I'm free will. Right. Like, so what's going on? So it's for me, that was the moment where I realized, oh, I am not trusting myself enough. And so now I'm like, I have started pushing myself to say, how deeply do I trust? Like, how do we how deeply do I believe this? And am I willing to stand up to everybody and say, no, we're doing this because I feel very strongly that this is right, you know, it's like this level of and then and so then the real danger for me, what I worry about a lot is like, how do you not become overly confident and you stop listening to advice?
But but historically, I've not trusted like where I was before was actually more dangerous because I'm doing things that I'm not really sure are right. My gut's telling me to do something else. And I'm going along with what people have told me. But so now I'm in a world where if I feel strongly about something, I'm comfortable going up against pretty much everybody. Right. And Hadley's, I'd be unfortunate position of watching that. Like, I've gone up against the board in the last year and a half pretty strongly about a variety of things. I've gone up against Hadley and Joe in like the last three or four months where I'm like, listen, I feel very strongly this is where the world is going and I know that we don't love it. But like, here's what I think. Right.
And I don't take that lightly because, you know, I have tremendous respect. The caliber of talent that we have in this company is, you know, is remarkable. And so it's a big thing to sort of stand up to tell people like, I think I'm I think we're going to go this way.
But so now I'm in a world where if I feel strongly about something, I'm comfortable going up against pretty much everybody.
It's so interesting to hear because we've talked about like making decision like data science is so much about decisions. We've talked a lot about like LLMs and how do you know when to like push back or like inspect what they're telling you? And I feel like this is such an interesting variant of imposter syndrome that I'd never really thought about, like that kind of like, yeah, not not trusting yourself, right. Or your gut when maybe in the face of an expert, you feel like, oh, I feel like I maybe need to dig in more or kind of like trust my instinct.
I feel like, too, like one interesting thing here is is we have Hadley and Wes here who are kind of like not not to butter you up too hard, but like world experts in these respective domains of sort of like data analysis with R and Python and open source tools. And I know that you you've met with them pretty regularly. I'm really interested in in what those interactions look like and how how you all like do you all how you all handle, say, like disagreements or things like that. Can you think about times where you all have disagreed about something?
Disagreements with Hadley and collaborating with Posit
There have been many, many times. What does that look like? Yeah, Hadley, Hadley and I remember a story. So I told you about like I was using DeepLyer. It was still early versions of DeepLyer and the various tools of what then became the Tidyverse. And at one point, I remember going to. So actually, I remember like a six month mark, Hadley came to me and says, listen, you can ask me for things, you know, you don't have to be because I was like, I'm not going to bug Hadley. He's like he's, you know, the chief scientist. Leave him be, you know, I'm I'm your guy. I'll fix whatever you need to fix, whatever. Leave Hadley be. And eventually, like in six months, I think Hadley reached out to me and says, hey, listen, you can talk to me. And I'm like, all right, well, good. So then I started asking him, like, listen, can we get this fixed? He's like, yeah, it's on the list. And so like I did it again. I'm like, Hadley, where are we? I'm going to work on this on Q3 of next year. And I'm like, well, that's not going to work. That's a lot of board meetings where I'm stuck with this thing that doesn't quite work. Right. I'm like, how do we solve this problem? And he's like, well, you know, if you could hire and could hire another engineer to work on it, that would work. All right. Well, that's a good way to solve the problem. Like, let's get to these things a little bit earlier.
But but, you know, that that's that was a very, very early days. I don't know if, you know, you like that that that happened. And maybe I told you this before, but but more recently, these are much bigger questions. Like I think I like blows much bigger holes in terms of like how people think about data science, about coding, about programming languages, et cetera. And so those are much more heated conversations, because in some ways, in some ways you're sort of touching on people's self-identity and how they feel about that. Right. And you're like, listen, that may not matter anymore. I'm not that I'm saying that, but like the implication is that like, boy, you know, does that really matter as much? And so people are reacting not so much to what you're saying, but to the implications of what you're saying and how and how that affects them and their transition into this new world.
Are you talking about are you talking about like the kind of shift in what skills are valuable? Is that. Yeah, yeah. And what does that imply in terms of, you know, product direction, investment levels that we have in certain, you know, packages or libraries or approaches? You know, how do we shift how do we shift the organization? What does that mean? What is the interface that you're going to be using to understand data and to communicate? And so, yeah, so that's probably the most recent set of conversations. And then I would argue there's probably the heat, the most heated conversations we've had in the last 13 years has been around this. Like, hey, what is the what is the future of data science? What does that look like? And how do we and how do we reason about that? And what is our what is the role that Posit plays in that game?
It's so interesting to hear. And Wes, you do you and Tarif me also. Is that your chat? We do. We do. I mean, I've I've had some type of working relationship with folks at Posit for for going on going on a decade now. Like I met I met Joe and JJ when it was just the two of them. I think back in 2012, this was even before Hadley joined joined RStudio. And so I I really identified with the mission of building a sustainable software company that can primarily build open source software for for the public good and also solve, you know, enterprise, the enterprise challenges that come along with with doing data science and the enterprise.
And so, you know, I I got, you know, essentially set up a business partnership with with with RStudio before it was before it was Posit to to build the Apache Arrow project. And I was very clear, like, this is going to take a number of years to bear fruit. And I think one of the reasons why, you know, I've, you know, been involved with with this company for such a long time is because there's very few companies that that can think and act and operate on such a long time horizon. So certainly, like, there's a need to, like, pay salaries and be sustainable and grow and grow the business, but also like to be able to think about projects that are that take a number of years to really fully realize and bear fruit. And so it's not just like, what do we need to do to hit like next quarter's numbers or like, you know, the numbers three quarters from now? It's like, what, how can we build the right foundation that can sustain us and power us for the next 10 years, the next 15 years?
I think one of the reasons why, you know, I've, you know, been involved with with this company for such a long time is because there's very few companies that that can think and act and operate on such a long time horizon.
And so, you know, I think one of the things that we've, we've often talked about was, you know, early on was like, it's important for us to bring the Python and our data science worlds together and to create have initiatives that help, like, enhance collaboration, enhance interoperability, like to bring these communities together, because I think there was a period in time when when there was this adversarial feeling of like, I'm an R person, I'm a Python person. And Hadley and I were always publicly being like, the language wars are stupid and counterproductive.
But, you know, I think with with Tarif and JJ, like, it's a lot of our collaborations have really been about, like, you know, what can we do right now to to put ourselves on a better path? And how can we plan better for for the future so that like, we're in the place that we want to be in two years and in three years and but I really enjoyed like those kinds of like, you know, strategic long term thinking and building really ambitious software initiatives. And it's certainly refreshing from having been involved in, you know, other companies in the past where there is like this, the short term, the short term mindset.
But, you know, I think that there is a need for like, I think in any business partnership or business relationship, there's a need for some amount of like, lateral thinking or like dev devils advocacy. So like, if you aren't in some amount of like, questioning the premise, or like, sort of examining like your strategy and saying, like, where are the holes here? Like, what, like, like, how could this fail?
And so it's like, somebody needs to come to the argument, like, sort of with a critical, critical perspective saying, like, what are we not seeing here? Like, what are we potentially like, where are blind spots? Like, where could our strategy possibly fail? Like, what are the things that we could that that could happen that we're not anticipating?
And so I'm happy that we have things like, you know, DuckDB is one of those things that like came out of seemingly nowhere, not quite nowhere. But, you know, we made a huge investment. In making DuckDB work really well in both Python and R and like, just magically like interoperate with the things and share data and all that. And, you know, it didn't happen by accident, like it required like planning and coordination.
Yeah, it's such an interesting partnership, because I do feel like, Wes, like having you around Posit, it's been so interesting to see and useful. Just like all the things you're doing and being able to like chat in Slack about even like AgentsVue or RoboRev, all the tools has been super helpful.
I try not to upset the apple cart, but I'm not, I'm not like known for my, my patience. Like I, I, I like getting things, I like getting things done, like as quickly as possible, even if it takes like, you know, if I can get something done in a month, like I'd like to get it done in a month as like kind of one of the, you know, the, you know, the historically Python people, although I'm increasingly a post Python programmer, as one of like the Python people inside Posit, it's, it's, I've taken it upon myself to occasionally be the provider of, you know, divergent opinions and whatnot. So respectfully.
Yeah, it seems really important. And I, yeah, Tarif, I'm curious. I know, Wes, you mentioned like the value of someone playing devil's advocate. I think Tarif, you, I saw you mentioned to the importance of devil advocate, devil's advocate, or people feeling like they can argue with you or push back. I'm curious as CEO, how often you see that? Like, do you feel like people often are willing to push back or?
There are several people in the company that have no problems at all pushing back on me, right. Which I actually love. I like, I love the fact that people feel comfortable doing that as opposed to, you know, saying something and believing something else.
Um, it is also remarkable for me to see how difficult it's been for people to be deprogrammed from other companies that they've been at from like to speak up. So we have what we call, um, uh, AMAs, ask me anything that are anonymous, uh, uh, or they can be anonymous. Like the default is anonymous and you can sort of self-identify if you want. But, uh, and you'd be amazed by some of the things that come up in those questions in those settings. And I'm like, listen, why don't you use any of the channels that we have so that we can have a dialogue about it? And it turns out a lot of people are scared. They're worried that if they were to speak up or if they were known.
So the problem of communication, you know, it's interesting. I've, I've been thinking a lot about this, you know, like how we all are, learn a lot about context for an LLM. Well, it turns out the same problem exists with humans. If they lose part of the context, then they will give you completely incorrect answers or they make, you know, completely incorrect assumptions. And so I, um, you know, I'll go to, to Joe Chang, for example, and I'll be like, you're my chief skeptical officer. Tell me what is wrong with my thinking here. And sometimes he'll be like, you know, he'll be like, well, I'm worried about this. I'm worried about this. I'm worried about this. I'm like, all right, great. Like, I understand. I didn't explain myself. Well, let me try it. How about if I explain it this way? It's like, well, if you explain it that way, then I don't have any issues with that. All right. Well, that's good. Sometimes it's like, no, no, I still have these issues. I'm like, all right, great. Why do we have this disconnect? And that allows you to refine your thinking.
Uh, but there's something tremendously powerful about having people in the organization that can use their collective intelligence, uh, to poke holes. But also there's a time I found now where like, maybe people won't be able to get there. And so there's a, that, that, that puts the onus on you as a leader to sort of say, I get it. This is unknowable. Like, we don't know what the right answer is. It could be this, it could be that, but I'm willing to take the risk. Let's go down this path.
Um, and, uh, and do something because acting like not acting, standing at that fork in the road and not doing anything or sort of navel gazing too long is going to cause you to lose the game altogether. So you're better off take, make the commitment to a path and go and make the experiment small enough that if you were to get it wrong, it won't cost you the company, but, but, you know, iterate your way, uh, as fast as you can towards the answer, which is not always what people want to hear because they're like, you know, we're going to, that, that means that you have to accept failure as something that is going to happen. And, uh, but, but you sort of try and manage that appropriately. And for whatever reason, people think that like, if you're the CEO, you're not allowed to make mistakes, which doesn't make any sense to me at all. Like we're all human beings and we all have imperfect information.
Like we're all human beings and we all have imperfect information.
And so you have to be open to being wrong. Um, which is one of the things that really attracted me to, to JJ when I first met him was that he was very comfortable with, you know, he's like, here's why I believe what I believe, but then if you come in and say, what about this? What about that? He's like, well, Oh, that's plausible. That could be it. So the fact that you can be that smart, that capable, uh, and that disciplined and yet be willing to reexamine your assumptions when you see new data. Uh, I think that is, that's what, that's what you ultimately want. And that gives you the best fighting chance of, of, of competing in this world.
Becoming CEO: ambition, immigration, and self-made success
You're so driven. Uh, but I'm almost curious, like if we go back to like high school, you know, and, and college, do you feel like you were like, I'm a be a CEO, you know, or like what, what was going on? Like, what got you to where you are? Like, take us back.
I started writing on LinkedIn starting in May, right? Like that was the, that was a two year ramp up to convince myself, like I should do it and I'm going to commit to it, to, to sort of writing twice a week, which is hard, it turns out, but, um, but I'm an immigrant, so I'm an immigrant, which means that there's a lot of things that you walk in when you walk, when you come to this country and you're like, okay, I don't really have, uh, a real safety net. And so you have, so that creates this dynamic where you're like, I am going to be successful here because I don't have too many other choices.
Um, and success was defined as frankly, power and money, right? But not necessarily CEO for me. That was never something that I aspired to. I didn't want to run a business. I knew that startups would be an interesting way to make money. Uh, so you take these bets and you hope that they pan out. And I've been in the fortunate position where I've made some good bets and the companies that I've been, that I joined relatively early went public and, uh, and then, uh, actually all of them that ended up getting sold.
But, um, but so, so for me, I knew I wanted to be independent. I knew I wanted to have power. And so that has driven me and that's driven me to learn a lot of things. The CEO thing came about, I think mostly because, um, even as head of engineering, I was very curious about a variety of, uh, disciplines and topics. And so I read a lot and I would grab books and sit down and like work through them and talk to people and whatnot. And so I could get myself to, to, you know, a decent level of knowledge, you know, after reading three or four books and talking to a bunch of people. So like most people are not willing to make that level of investment in things that are not related to what they like to do. Uh, I have.
For whatever reason, the ability to sit with pain, uh, for a long time. For longer. And so that means like, I can work through harder problems for, for longer. I don't know what kind of experiences end up, uh, panning out for you down the road. But, but you'd be surprised when, when you're like, Oh, that experience that I had 25 years ago is actually the lesson, you know, that I needed to, to make a different decision now. And so a lot of things about how Posit works are reflection of my observations of companies over the years and saying, okay, great. Like we need to avoid this class of, uh, uh, issues. And so we can make new mistakes, um, and, uh, and go from there.
But yeah, what's, what's, what always drives me is I love to win. I love to win. I love to win. And I also hate to lose. So like, I have like this double whammy where like, I can't help but want to win. And which my wife didn't appreciate because I wouldn't let my boy beat me at anything. And so if they beat me, they beat me because they deserve, you know, they actually earned it. That's kind of how I was raised. And I generally hate to lose. Like, I've learned to accept losing if I feel like I've given it my best shot.
Choosing computer science and the road to Posit
Okay, so you mentioned like, yeah, immigrating was a really kind of like impactful, important experience. And I'm curious, like how you chose software engineering, like, I'll say my dad's an immigrant from Hong Kong. So he was kind of like, your choices are like doctor, lawyer or accountant. And I became a psychologist, which was a big, the most useless degree. I'm curious how you how did you settle on computer science?
it had 10 megabytes of storage, and 128 k of RAM, and which you could extend the RAM a little bit. But, and so that's, and so that we brought that to the Middle East, I started playing with that. And I enjoyed, I enjoyed what you could do with computers. In the early 80s, when I came to the high school here in the United States, I was younger than my class, I was shorter than everybody else. And I spoke with a funny accent, you know, because my English, English was a second language, still is, but, but I don't have as much of an accent now. And, and, and that's where I got introduced to like Pascal, and eventually I got introduced to C and, and whatnot.
Reading as a superpower
Yeah. And as I understand from Hadley and from some of what you mentioned, part of how you really try to fill the gaps is you're a prolific reader. Is that right? That is correct. Yeah. Yeah. Yeah. I try to, I try to, yeah, I, I used to, before getting married, I could finish like four books or five books per month. And now I'm like down to two per month. We'll see. Once my boys all leave the house, then I'll have more time. Maybe, maybe that'll, maybe that'll climb up again. We'll see. But, but yeah, books, books are very powerful, I think as a tool and they're underestimated by the next generation of folks, because people just want to get the synopsis. And I think sometimes just working through working through a book is challenging and forces you to try and remember over multiple sessions, which actually plays into the whole, like, how do you, you know, how do we retain information?
rate that was unprecedented before, right? Like, and yeah, you have to know to ask whether that is true, but but it's, it's, it's an, it's a fascinating time if you're motivated.
I, I know I'm trying to really jam this in with the time we have left, but because you, I feel like you left so many interesting details.
Playing the oud
I also know you mentioned you play the oud. Is that right? I do play the oud. Yep. Yep. Yeah. I don't know how many people know what the oud is, but it is a, it's a fretless stringed instrument that like the, I think the Muslims brought into Spain way back when, and that became the lute, which if you listen to the word lute, it sounds like it sounds a little bit like the oud. And, and that then became obviously the guitar over time. And so, so it's an instrument I've, I've always loved. My mom was a concert pianist, is a concert pianist. And so she taught me the piano at a very young age. And when I went to college, I was like, all right, I'm not playing the piano anymore. I don't really like the piano all that much. I mean, it's nice, but it's not my favorite instrument. And so I picked up the guitar in college, which I think a lot of people do in their freshman year or whatever.
So the oud is fretless, which means that you can hit quarter tones. You can, you can play, you can play anything you want because there's no, there's no real tempering of the, of the, of the, of the notes. And so it becomes very much a, yeah, you have to listen carefully because you can hit quarter tones or half tones and, you know, eighth of a tone, depending on what you're doing. So it's, it's, it's beautiful. It's beautiful.
And is this something you played with other people or what? When would you play it? Yeah. Yeah. I used to like, yeah, I tell people like before getting married, I was much more, I had more time. And so I would go out and perform with others. There might be some YouTube videos of me performing in certain places, but, and I would, I learned how to dance the line dancing of, you know, the, the, the Middle East called Depke. And that was fun too. But yeah, music, I love, I love music. I love singing. I love, I love that world. And it's, it's very different than my day-to-day life now.
Yeah. Yeah. It's such an interesting, neat, I feel like, activity though. And do you play today or what does it look like today? I, well, so it's interesting when, when my, I was working remotely for a few months last year. And so I was able to play more in the mornings since coming back here and I'm back and back in Boston right now, it's, it's been difficult to fit everything in. So, but I think once my, once my second, once my eldest is out at school and my wife is back from the Middle East, then we will, then I'll have a little bit more time, I think.
Waiting 40 years for a cat
Okay. Not, not to lightening around all your personal interests, but I also understand you have a cat that you're obsessed with. Is that? Obsessed? Maybe it's not too, it's too strong a word, but I've wanted a cat since I was five years old. And my mom was like, you can't have a cat. And then when I went to, when I came back from college and got my own place, they're like, you can't have a cat. Then I got engaged to my wife and she was, she was in the Middle East. And I'm like, listen, I'm so excited for you to come, you know, we'll, we'll, we'll get a place, we'll get a cat. And then I just like, here's silence on the line. She's like, it's either me or the cat, you know? And at that point I already was already committed, right? Well, I guess I'm not gonna have a cat. So then I told her like, well, hopefully our kids will want cats. And so it took a while, but then during the pandemic, you know, the kids were like, we should have a pet. My wife is like, let's get a cat. So we started researching cats, but I'm like, you know what? Let's not get a cat now. It's doesn't feel like the right time. And, and then, and then eventually like about three years ago, my wife, I had like a milestone birthday and my wife is like, you should get a cat.
So, so more than 40 years after wanting a cat, I got a cat. So my cat is, is keeps me company all day long. My sons have left me, you know, there to go to the Middle East and now to California. So I'm like, this is, this is somebody that will hang out with me while I'm on calls all day long. So, so yeah, I enjoy my cat. I'm glad though, that at least, even if it took a four decades long arc that you, you made it, you know, that you got the, you got the cat eventually.
Yeah. Yeah. I'm a patient. I'm a patient man. Like Wes was talking about this whole idea of like, can you build for something for the long haul? Right. And I, I like, I believe that really good things can take time to, to sort of perfect and get, get right. Seems like a huge lesson, you know, like impatience and kind of like being able to go with the flow or kind of like have the long view. So.
Playing the long game at Posit
Yeah. Play the, play the long, play the long game. Yeah. Yeah. I mean, some people like I, I used to, I would chafe when people say, Oh, this is a lifestyle company. And I'm like, this is not a lifestyle company, man. We're very competitive. You definitely want to win. Like there's no lifestyle, but it's just like longer time horizons. And how do you build something that's ultimately sustainable? That's the thing that I'm interested in. So like we got the profitability in the first two years and we've grown the company to be 350 people with a meaningful bank account because we're very disciplined about how we approach the work, what we're doing and where we're investing. And, you know, we're willing to call it when we're not doing well in a certain area. I'm like, great, let's move on. Let's, you know, give it a, give it, give it enough of a shot, but if it's not going to work, then call it and then make sure that you're willing to adapt.
Yeah. Yeah. It's a really, I don't, I interesting thing to hear it in powerful lesson. Cause I do think going back even to what, like Wes said about DuckDB, you know, like that positives posits, like sowing seeds and supporting a lot of different things over like really long timescales. And it can, I do think even for me, it can seem surprising or kind of like irrational, like not always clear, but it, it does. I think when you zoom out to the like decades long, like how can we support open source? What, what from the ecosystem is going to be around in decades? Yeah. It's the first principles thinking. So like when we added support for Jupyter notebook and Jupyter lab into our products, you know, I was like, before we released this, we're going to, we're going to like, what's the organization that, that, that supports this work? And it's like NumFOCUS and like, great, let's go and get introduced to NumFOCUS and let's, you know, let's donate, let's make sure that we are part of their roster.
And I remember thinking to myself, I'm like, great, like we're going to do the best we can. So we made a donation and that donation ended up being like at the platinum tier, which I found surprising because you know, many more people, many bigger companies are, you know, using Jupyter notebook or relying on Jupyter notebook or Jupyter lab or whatever. And I'm like, this is, this is part of the problem with open sources that most people just want to use, but they don't want to pay. They don't want to sort of like support it and whatnot. So the hope, the thesis here is that if you can create an entity that believes in the open source work focused on a particular domain, because you can't just say like open source in all forms, then hopefully you can create a sustainable way to continue to contribute to it. But yeah, first principles thinking it's probably the greatest gift that my dad gave me. And, and that's also what I really appreciate about who we have in our company. Like we have that our leaders are willing to engage in like, Hey, let's go back. What do we think are the first principles on this? And then derive from that what is correct.
But yeah, first principles thinking it's probably the greatest gift that my dad gave me.
So when, when I met Wes which was later than obviously JJ and Joe, you know, the conversation was very straightforward. Like, Hey, do we believe that arrow should exist in the world? You know, is it a meaningful thing for there? And it was very obvious. I'm like, of course it needs success, right? We can't, the idea that you can only do this kind of analysis if you are on a massive cluster or spark cluster or whatever, it doesn't make any sense, right? And as you should be able to leverage the, the power of, of modern chipsets and data's formats and whatnot to, to drive value, because I still believe there's a lot of insight that you can get from data sets that are less than a hundred megs. And the idea of, you know, or, and certainly like a hundred gigs will cover the vast majority of what people are likely to have in organizations. And so making that accessible to the masses, no brainer.
No, I, I really appreciate you coming on and talking more about like your role at Posit and, you know, Posit's long timeframe and, and also like a lot of what kind of like brought you to where you are and, and thinking about data science and the ecosystem. I feel like it's, yeah, it's, it's such an interesting, um, I, I find Posit to be such an interesting organization with such an interestingly long perspective that it's, it's really helpful to hear.
I mean, that's, that's what was attractive to me about the organization in the first place. Like with that, JJ had that kind of ambition. And so like when I, when I joined, I think it was like the second or third week, I was like, boy, this would be fun to do for five to 10 years. And he's like, no, think 20. Like he didn't even blink. He's like, think 20. And I'm like, okay. And then he got up on stage on 2020 when we announced that we'd become public benefit. And he's like, we want to be here for a hundred years. And like, everybody's clapping. And I'm like, oh shit, that was my reaction. I'm like, I don't think everybody understands what that means. That changes the game even more. Like, how do you build a company for a hundred years? And how do you think about that and whatnot? And in tech in particular, I'm like, give me a break. You know, very few tech companies make it to a hundred years, you know? Uh, and, uh, the, the history littered with examples where that doesn't happen.
What it takes to be a hundred-year company
Um, but, uh, but it is, it does inform how you think. So I know we've, we've gone pretty long, but maybe, maybe we could close out. You mentioned like Posit being a hundred year company, and maybe we could close out just by asking like, what, what do you think Posit needs to do to be here a hundred years from now delivering on its mission?
Yeah. It's interesting that the a hundred years thing was hard in 2020. Now the complexity of that is much, much higher because like, I don't even know, it's very hard to see what the world is going to look like three years out from now at this point. Like, I just don't know. I do. I think that to have a shot in this world, you need to somehow, um, have flexibility in your thinking and a willingness to be curious and try different things and, you know, adapt yourself and let go of ideas of what you thought your life was going to look like and what you thought your day-to-day was going to look like so that you're able to, um, you're able to, uh, adapt.
So like the, the, the biggest worry that I have for, for everybody, honestly, but certainly for folks in, inside of Posit is like, are you going to be okay if 70% of what you used to do is now no longer being done by you? And so therefore, what are you going to do to add value to sort of, uh, uh, to the mission of the organization? Right. And, uh, and that's a hard thing because like I said, it touches on the whole like identity question. Uh, it, it, it puts fear into people's hearts where like, I'm like, listen, I'm not trying to scare you. I'm just trying to say like, this is reality, right? Accepting reality and then adapting to reality is really, really important.
Accepting reality and then adapting to reality is really, really important.
So if we can be a company that continues to sort of, uh, uh, where people are, are comfortable adapting themselves to change, uh, embracing things like when Wes was breaking, you know, uh, uh, scaring people, like we can go faster. We can do this better. We can. I loved it from where I was sitting. Cause I was like, yeah, you go boy. And then he's like, are you okay to read with me? I'm like, yes, Wes, go. Like, as far as I'm concerned, that people, everybody needs to understand, like the world is changing and we have to accept that and, uh, and adapt to it. So I'm a, I'm a big, I'm a big believer in stoicism as a, as a philosophy and, uh, as an approach to, um, to sort of, uh, uh, accepting reality and being able to sort of act in, in, uh, in ways that allow you to express yourself completely.
So, so I'm hoping that's what we become. Like we're a company of three 50, you know, maybe we're still a company of three 50, five years down the road, because we are able to sort of scale the business without having to, um, add a lot of people. But the, but what those 350 are doing probably looks very different in three to five years than it does than what people are expecting today.
Yeah. Hadley and Wes, I'm curious if there's anything you would add to that or thoughts about that you have this sort of, what does it take a company to be here a hundred years from now? Yeah, it's so hard to know. I mean, like, like when, like how many companies have survived a hundred years? It's crazy. I mean, I do think like one criteria is like, we have to be lucky. We have to be good and we have to be, be lucky and there's things outside of our control. Uh, you know, it feels like having, like when I think about organizations that have been around for over a hundred years, you know, I think of universities and churches, and there's something about that, like that mission that, and I think we, we have that. It's maybe the mission feels a little battered by AI because we have to think like, what is our mission around data science and scientific publication mean as data scientists, data science and scientific publication are changing so radically. But I do think there is that, that core belief, that mission, that feels like that, that we have something that can carry us the distance. Yeah. I will say, I will say if you studied those companies that have been around, cause there are, I think there are like 6,000 something companies in the world that are over a hundred and many that are over 150 years old. And in almost all cases, they're not doing exactly the same job that they were doing before.
So like, it's actually really important for the company to reinvent itself in a way that is still, you know, remotely, uh, tied to, I love the example of Nintendo because, uh, I mean, they were, yeah, I just looked it up, you know, founded 1889. Um, they were an almost 100 year old company when they went from, you know, making playing cards to building, you know, the Nintendo, I guess the Famicom original Famicom. And so, you know, I mean, if you kind of had a time machine and went back and so look at like what Nintendo today versus like the way it started out, I mean, so there, there is, I think, yeah, there is a need to like, for companies that exist for a long time to be able to, to reinvent themselves and to say like, this is what we did in the past. Like this is what worked. And we built those things in the past because like we were, we were solving the problems that existed at that moment in time.
And you know, 20 years, you know, things will be, things will be different still. I mean, I I'll probably still be building software in 20 years. Like I think somebody was saying to me, like, you know, even if I, you know, become semi-retired in the future, like I don't really have an off switch, like I'll still be working on something. Uh, I don't know what coding is going to look like in 10 years. It feels like it radically changed about 18 months ago. Um, but, um, you know, maybe it will be, you know, I mean, mostly just like all voice command, who knows, you know, hard to imagine. Who knows, who knows.
Closing thoughts and encouragement
Triph, any last thoughts? Uh, so many thoughts, so many thoughts. Yeah. I know I'm, I'm, uh, I'm, I think it's, I think it's an interesting moment in time for all of us. And I think it is as stressful as it can be. I think there's a way to sort of look at the world and say, okay, great. I'm going to, I'm going to figure out how to adapt and thrive. And I think that's, that's a decision that you can make as a human being to say, listen, I know things are going to look weird and, uh, uh, confusing, but I'm going to choose a path forward through this.
That's going to give me some fulfillment. That's going to allow me to do my best work, whatever that means in whatever that context is. And so, uh, so that's my encouragement for everybody else. It's like, let's don't, don't sit and don't worry about like how things were because like, that's not coming, but you're not going to change that. Like that is what it is. That's a fact of life. And, uh, and now let's go and solve the next set of problems that are, that are facing everybody. And I think there's still a lot of problems to be solved and, uh, there's a lot we can contribute to society.
It's like, let's don't, don't sit and don't worry about like how things were because like, that's not coming, but you're not going to change that. Like that is what it is. That's a fact of life. And, uh, and now let's go and solve the next set of problems that are, that are facing everybody.
Yeah. Yeah, for real. Well, really appreciate you coming on and I feel like laying out Posit's vision and I know we've gone pretty long, so covering just the whole gamut, you know, really appreciate your, your insights. And, um, you know, uh, I was gonna say everything you do for Posit, that's very cheesy and I'm not trying to butter you up. Really appreciate, you know, all Posit's mission and, and all this work and all the work ahead, I think is so interesting to hear about. Yeah. Thanks for coming on. Thank you guys for having me. Yeah. Thanks Tarif. Thanks Tarif.
The Test Set is a production of Posit PBC, an open source and enterprise tooling data science software company. This episode was produced in collaboration with creative studio Adji. For more episodes, visit thetestset.co or find us on your favorite podcast platform.