Transcript

Transcript prepared by Bob Therriault, Adám Brudzewsky, and Igor Kim.
[ ] reference numbers refer to Show Notes

00:00:00 [Ashok Reddy]

When we talk about data, I mean every query costs money, and everyone is more data-driven and the problem is you just spinning up more things on cloud and running more queries, but you just need to bring what's necessary, right? That's the power, that's why I keep going back to the array and vectors is you just focus on what's needed to answer the question and what's not trying to bring the whole data across the wire.

00:00:23 [Music Theme]

00:00:33 [Conor Hoekstra]

Welcome to another episode of ArrayCast. My name is Conor and today with us we have a very special guest, but before we get to introducing him, we're going to go around and do brief introductions from our panelists. So first, we'll go to Bob, then to Stephen. Then to Adám and then to Marshall

00:00:44 [Bob Therriault]

I'm Bob Therriault and I am a J enthusiast and very enthusiastic about J.

00:00:50 [Stephen Taylor]

I'm Stephen Taylor. I'm a q and APL enthusiast and also the KX librarian.

00:00:58 [Adám Brudzewsky]

I'm Adám Brudzewsky I am an APL enthusiast, I'm also enthusiastic about other programming languages, and but I do APL for a living full time at Dyalog.

00:01:09 [Marshall Lochbaum]

I'm Marshall Lochbaum. I'm the writer of BQN, but I'm I'm really not that enthusiastic about it, OK?

00:01:18 [CH]

And I'm Conor, as I mentioned before, I'm a research scientist at NVIDIA and I'm an enthusiast of all array languages and I switch between the ones that I I write in day-to-day and we have one exciting announcement and then two other less exciting announcements. So first we'll go to Bob for the very exciting announcement, and then we'll go to Marshall and Adám for our two follow up announcements.

00:01:39 [BT]

Well, my very exciting announcement. It's actually congratulations to Adám, who's also very enthusiastic about the new addition to his family. A baby girl who just happened to pick the fortuitous state to be born on of my birthday, so we'll probably be exchanging gifts in the coming years. But congratulations Adám, that's fantastic and I I hope everybody's doing well and and and healthy and happy and, I mean, just say goodbye to sleep, but you already know that because I think you've got two other kids.

00:02:17 [AB]

Well, I mean the important thing isn't so much that it's a girl. I would say that the important thing is a future APL programmer in the making so I was active in the various appeal chat rooms as she was being born and I was asking this question if if does anybody have a good first APL expression for a little girl and people came up with the way to difficult expressions like IOTA 10 and things like that or even 1 + 2 3 and and then Alve was a previous winner of the APL Problem Solving competition. He came up with the perfect response which was a commute or to the diocese just by itself monadic operator, you know the one that can either make a monadic function out of a dyadic function by using the same argument on both sides flipped over the arguments around or can now be used for constant functions as well, but that expression, just that symbol itself. So I tried that on her with great effect, so that's awesome.

00:03:20 [CH]

Bit confused here. Did I hear this correctly? You were in APL chat rooms while the baby was being born.

00:03:28 [AB]

Yeah, of course.

00:03:30 [ML]

Yeah, he was just talking about APL and then in the middle of it he said, oh I'm helping to convert right now.

00:03:36 [AR]

All right well?

00:03:37 [BT]

And then he was trying to answer questions and saying he was apologizing for being slow because he had one hand on the baby and one hand off the keyboard with so that was it was hilarious.

00:03:46 [CH]

Holy smokes Well, for those that think it's difficult to type APL characters, clearly, you know you can multitask while doing it, to the extent that you can you.

00:03:55 [AB]

No, it's quite the opposite, because these languages so terse, so brief, you stand a chance of typing meaningful amounts of code while you're assisting with the birth as opposed to Java or something? How much code could I produce?

00:04:11 [CH]

I guess that's a good point. That's a good point. Well, congratulations from all of us, sure, yeah, it's going to be an exciting next few months and a new chapter. New journey. And yeah, excited to hear updates of the progress of you know how many months are we going to expect? Our first sort of small programs to be written. We'll be waiting here all right. Let's head to Marshall for the announcement and then head back to to Adám for his last one.

00:04:38 [ML]

Alright, certainly not quite as exciting. We're of course, making various additions to BQN. The one I've recently done is something a lot of people have asked for. It's a parser and a serializer for the JSON format [01], so JavaScript object notation. There's now a library in BQN libs. That handles JSON, so you can just, you know, give it a string of JSON Data and it will give you a BQN array out that sensibly represents what was in that string and the other way around. You can take your BQN array and serialize it to JSON it's I believe it's completely compliant on the parsing. It seems reasonably fast. It's not like simple JSON, but it's it's 20 or so megabytes a second, so yeah, it should be a pretty solid tool and convenient to use.

00:05:28 [CH]

All right? Awesome and over to Adám for his last announcement.

00:05:32 [AB]

Well, it's not so much as an announcement as much as a pointer to something that you might find interesting. If you find this podcast interesting and that is a talk, uh, just from a week before we're record sitting recording this, and and it's called change the way you write, change the way you think [02] and it's talking about and getting some ideas on and and changing your thought processes when writing code in general. So I think that's an often overlooked aspect of these array languages, because let's be realistic, the world isn't going to just start using only these languages and discard everything else, but I think there are a lot of good lessons that can be learned from the style of code and the thought processes that are involved in the thought process you learn to have when you're writing in these languages, and so this talk that we'll put a link in the show notes to go check it out.

00:06:29 [CH]

I really think it's awesome, yeah you can find that link in the show notes as always, and I think that brings us to introducing today's guest, which I'm very excited to be having on this podcast, our guest is Ashok Reddy, the CEO of KX, [03] which is the company, of course, most of our listeners will know that is responsible for distributing the q language and the kdb+ database, so a little bit of background on Ashok is that he joined as CEO in August of 2022. So just I think under over half a year ago he has more than 20 years of experience leading teams in driving revenue for Fortune 500 companies and private equity backed technology companies. He spent ten years at IBM as a general, a group general manager, where he led the end to end delivery of Enterprise Products and platforms for a diverse portfolio of global customers. In addition, he has held leadership roles at CA Technologies and Broadcom and worked as a special advisor to digital transformation company Digital.AI, where he helped the senior leadership team devise the product and platform vision and strategy. So it's going to be a super interesting opportunity to ask Ashok questions about sort of the future of KX. Maybe the history of KX, but before we get to that, maybe you can tell us about your sort of a abridged history of your path to being CEO of KX, and whether along the way you sort of stumbled into any of the array languages, or sort of you discovered them upon arriving at KX. And yeah, be super interested to hear what sort of the background that led you to where you are today.

00:07:58 [AR]

Thanks Conor. I think that I would like to characterize myself as a lifelong learner and I think that in that context I am largely self taught and started learning about different languages and then it's kind of been a history which kind of got me where I where I am, but I think it's that I started off my programming when I first started many years back in India where we didn't have computers in those days, but one of my uncle gave me a sharp 1500 pocket computer [04] to start coding more of a numerical analysis, so I tried to figure out it wasn't basic at that time, but then I tried to program Newton Raphson based on how do you solve differential equations right? So I started off with that and then it was first discovery of using subscripted variables which was the first array you know when I first discovered, that is when I said wow instead of having separate variable for each thing, I can actually create something as a subscription subscript and that was my, you know, started off that way and then over the years I've worked, you know. And I did. I'm a chemical engineer by background, so you kind of look at engineering as more about solving problems mathematically apply. It's more like applied math, right? You're trying to represent physical systems in mathematically, and then you're trying to solve it. And that's really where I kind of did most of my learning in that context of not just programming, because I think it's really about applying math and how do you represent math in a way where it can. It's the most best way of representing physical systems, and then you end up. I learned FORTRAN and then I went to move to US. I did a lot of things I've worked in you know, creating the first dot net for Visual Studio and Microsoft was introducing .net languages the common language runtime the? So I was the product manager for this J sharp product, which was the Java for dot Net. So I ended up learning and creating type system. And then I ended up working with rational software around UML and object oriented systems programming where we're trying to abstract things, is there a programming at the language level you kind of move up the chain. So I worked with Grady Booch and Jim Rumbaugh [05] and we were the three amigos too. For some quite a while it was all about object oriented programming and you would not program at the lowest level, but maybe everybody is going to program at the objects, but that's you know, the model driven development didn't pan out like the way we thought over the years. But thenI also worked in the mainframe for when I was an IBM rational, got acquired by IBM and. So I ended up knowing quite a bit about COBOL and this is one of the best systems ever designed. In terms of the systems you know, it was written in, started in 1964, but then I was trying to learn about how the COBOL works and then you know, I'm as I said, a lifelong learner. I went back to school and Georgia Tech and actually did a lot of things around AI and ML and that I got into JavaScript and Python and I think the one with KX is really where I kind of started to really understand the value of power of array programming because the whole notion of how you can represent mathematically things like the speed of thought. I heard you know a lot of times and listen to the ArrayCast podcasts, which are brilliant. I mean many of the people like Nick (Psaris). He's the guy he and I've talked about why is the aray-based programming so useful? Because I think he kind of translated that to coding at the speed of thought to getting to insights at speed of thought, right? That comes down to, I think the mathematical representation of things where people can represent their thoughts in mathematics and the last thing, I think one of the things I found with q and when I talk to customers it's a even though it stirs people talk about being terse and how you represent, but it also it represents things so well, whereas all other programming languages. The best programmer in one of the companies where has done the most work in our base programming it was surprisingly a visually impaired person who couldn't read or see and she was doing everything better than anybody else. That's because it's so expressive in terms of what it can represent so that it's where it can open my eyes into what it could do and what are the use cases which I can talk some of it in terms of what type of things we're seeing with customers and where we are going. So that's really my kind of background in terms of where I am, but I'm a technical person. Call me as at First technical CEO for KX, so that means it's just that more I understand technology there's and I'm still trying to solve customers problems. But it so happens having sometimes people will find that if you can understand them because it's not just technology, but how does it solve problems. So I take it as a compliment, not necessarily like my geek CEO for the first time.

00:13:19 [CH]

It's it's actually interesting when you just said that the first technical CEO, it immediately made me think of there was, uh, I won't say which company I was at, but a company I've worked at. There wasn't one time they brought in some consultant from I think it was I don't know one of the big consulting Accenture. You know something like that. And I thought it was going to be a very like consultee buzz, wordy kind of thing like that was just my bias whenever I see that and then very quickly the first thing they said was, oh, I'll tell you about my history. I started off, you know, writing Smalltalk at Xerox PARC or something and I was like wait what! Smalltalk at Xerox PARC, like that's definitely not what I expected, and so it's something similar because you just you know, rattled through, you know, all the way back to, you, know FORTRAN and COBOL and and J sharp. Correct me if I'm wrong, is that the that was the name before C# and then I was at Oracle or some company was going to sue? Or was that a different J, something something?

00:14:11 [AR]

Yeah, so no, no no it is. It is actually J plus plus was Microsoft version of Java. [06] Sun was suing Microsoft that those days that was before the dot.com OK and when they introduced .net Sun would allow them to launch its over Rational as a separate company to develop this and that's why I was able to do that and work on so. J Sharp was really the java for .net and it was the J plus plus to J Sharp so C# was this still exists? But it's the C#. Looks like Java, but it's it's like you know it. It has its own constructs, but it has Java J# has Java syntax, but it leveraged the common language runtime from .net.

00:14:57 [CH]

OK, interesting so J plus plus was the one I was thinking of and J sharp was, uh.

00:15:01 [AR]

Yes, yes, visual. I think it was all this visual J Plus plus which was very popular, but I think it fragmented the whole notion of the write once run anywhere of Java, and it had extensions which only would work very well on Microsoft. But if you write something in J plus plus, it wouldn't run anywhere else, so that's where Sun those days did like you know, they thought that they broke the value prop of Java. And so Microsoft ended up trying to create this J Sharp to kind of do the same thing. Make it easy for people to use, but it looked like Java, but it was really .net language.

00:15:40 [CH]

Interesting yeah C#, that whole space is something I wish I knew more about like I was looking at rankings of languages and C# actually ranked higher than C, which I was a bit surprised by and I looked just at the latest releases and it's all the way up to like C# 11 at this point. And there's just. There's been so much evolution, I think like C#. Three or five was the one that introduced LINQ, which was has become like a really popular paradigm within C# programming. But anyways, there's been like you know six or eight releases.

00:16:11 [ML]

That's the LNQ link, is that right?

00:16:12 [CH]

Yeah, yeah LINQ. It's like a sort of composable algorithms and it's linked to to sort of what do you call it? SQL as well.

00:16:22 [ML]

A little bit array like if.

00:16:23 [CH]

Yeah, it's kind of.

00:16:24 [ML]

It's more of an SQL thing.

00:16:25 [CH]

Yeah, but anyways, the point being, is that like I hadn't realized that there had been so many releases, and apparently this the the functionality that's been added to it, it's it's completely changed from what it was back you know, however, many years ago C# three was. Anyways, back-to-back to KX and Array languages. Well I'm not sure should we pause if there's questions from anyone else first before we ask, maybe?

00:16:47 [AB]

Well, I'd like to ask them, did you bump into any of array languages before coming to KX?

00:16:54 [AR]

No, I think you know. I think to me the starting with what I found with more from I knew other languages in the sense more like solving the problem. I mean, I actually thought. You know when you look at your spreadsheets right? I started with VisiCalc [07] which was my first one, which after I'd it was trying to solve the type of things what we're trying to do in arrays. But you know, using a spreadsheet, but I don't think recently only I've found that Excel and others have introduced some dynamic array. But it's very hard to kind of do that in in any of these spreadsheets, right? So I wasn't really, I mean, I think I learned more about it in the context of when I before I joined KX. I did a lot of research and that's when I kind of like not only q, but went back to looking at K and you know, if I one of the things I found is like for example when I did Newton Raphson and you do a lot of loops and you try to guess you know you have to start with the initial estimate to figure out what the tangent of the slope is when you try to solve a differential equation. But there's a lot of loops and you get caught with it, and I think we don't know what when it doesn't converge to a root and it takes forever. And with this I mean, I, I saw you know some example I kind of wrote myself was to write one line, Newton Raphson. I never imagined you could write something in a something like a q you can write in one line. You can solve the whole Newton raphson, which. This is where I kind of stumbled about the power of it, right?So yeah, no, I've not really worked or, you know, run into it. But I did have the conceptually I kind of understood the, you know this thing and then things like implicit parallelization, parallelization, things like that which I find what brings array programming brings. I think I was kind of fascinated with it, but I had not used any language or I'm not familiar with the any other languages.

00:18:47 [BT]

You mentioned UML and working with Grady Booch and that, and taking the programming up a level. Do you see a parallel to how q is used that way that it's it's array programming, but brought up that you're thinking sort of on a higher level. Or do you guys look at it that way?

00:19:03 [AR]

Yeah no, I think you know one of the things that we are trying to do as a company is to the power of, I think first of all I think q is less terse than k for example. K is. It's even less you have a dot dot or something like that means something and it's it goes goes beyond the q get because it becomes a little more this it's still terse and some of the people who know like Nick has written multiple books, but it's the power of that. How do you bring that value of helping people to code at the speed of thought? It's not very easy for you know, except for a small population in this world. So what we're trying to do is how we can democratize it, and to get more people to do that to get to the same thing. So one of the bigger areas what we have been doing is to help people to use Python [08] and your wrap q and underneath the k so you can actually use just Python because there's more than several million people using it, and so you can get the capabilities and the power of q and the array based programming just using Python. That is where one area we are focused on and the other one which is recently you know, it's interesting. The whole LMM and the chatGPT, right? We just had some things which had asked the internal teams to go and do it, and they did that. They took some of the q best ways of doing some of this, and you know use chatGPT to help people to kind of OK. This is what I want to do and here is the q code. But how it would represent if I want to do a matrix multiplication, how do I do that instead of knowing you can actually chatGPT generates that right and so we are thinking more of the higher level means that maybe you know there's a play for generative AI to help people to do that and train and learn versus how we're actually doing, you know, there's one way of doing it in Python, but still, if you want to get the full power of q, you'll have to use q because it has certain things where only q you know it's the way you write. It's thinking differently, right? The way I thought people who come from object oriented programming going back to my example they have a hard time understanding learning q because it's thinking differently. You're thinking mathematically. You're thinking right to left because it's everything is terse because there's a reason for it. The way Arthur designed it and it's not easy to replicate that in a Python because it's designed for more, you know, people to understand, but it's not expressive in a very short way to do it without consuming a lot of power. I mean, in memory it uses things which are joins and things like that in memory so I think we can abstract to some extent, but in order to truly get the power, I think people will end up having to, you know, still continue to program in q for some reason or k in some cases. But it's more you know, I think about it as a spectrum, and it's like, you know you have, you want to democratize it. It's like a photography, right? When people add all this, you know, professional photographers only they understand the best you know where to get photographs and the depth and all the things but you know over a period of time like everybody uses iPhone right now to take the pictures, but you still have the high end cameras and the photographers and people who do for like National Geographic only they understand that and you can't suddenly use iPhone to do what they do. Right, so yes, it looks on paper because your eye only can recognize so many pixels, but so that's the way I think about it is more - Can we abstract it to a point where we also cannot? You don't want to take away the power and that I think is the challenge we have I think KX where we went too much trying to democratize it, but we kind of lost focus on the power users and people who thought, hey, what are you doing around q? What are you doing on kdb+? [09] So I kind of brought it back to even we didn't even allow people to download. So we have, you know, kdb+ available 64 bit. We are investing in 128 bit or 256 bit. We're doing lot of things for the power users but also helping more people to get to it through Python. And now with Microsoft is releasing on working with Microsoft. We're releasing a Visual Studio code for q and SQL and PyKX, which will hopefully make it much more easier for people to program, get the value, and that I think is huge. I think our customers are very excited about it because you're abstracting in terms of auto completion, to how do I debug? How do I do all this stuff? And I think part of the challenge we have today complaining people don't understand the error messages and it says the type error and it's the famous type error where you're doing something and you don't know what it is, right? So anyway, hopefully that answers your question, but it's sort of like you can do update oriented, but it's a... it's not expressive and that's what happened with UML, right? People ended up programming and doing reverse engineering to get to the model versus starting with the model and generating the code.

00:24:15 [BT]

It does and I love the photography analogy because I think that's so apt and especially when you think about people using iPhones to take picture. I've heard so many people saying that they'll take an iPhone with them to take pictures because it takes really good pictures, but if they're really good photographers, they've also got a big camera in their back because they know that there's things that the big camera can do that the iPhone can't. But most of the time for a snapshot the iPhone is perfect right?

00:24:40 [CH]

So just just to be clear in this analogy, the iPhone camera is Python And the heavy duty fancy camera is q OK.

00:24:48 [AR]

Yes, yes.

00:24:51 [CH]

Just just so I understood which language is what.

00:24:54 [ML]

Because q is a lot a lot smaller on your disk than Python.

00:24:57 [ML]

Let's be clear.

00:24:59 [AR]

Which is which is also true, right? It's kind of a not intuitive because I think it's the way I think about. That's why you use so much literally you can do a lot more power, but it's just that in iPhone, when we say Python or others, it's just a lot more people can do it, but you know it does take more power, more things to do it. You know, even the energy consumption. We find that with the q it's the smallest foot when thing out there right? It's less than one MB and it's 840 lines of code. People get astonished in terms of what it can do in terms of the edge computing and the there is a statistic I saw Gartner had put out in three years.

00:25:41 [AR]

The data analytics and other things what people are trying to do will take more power, consume more power, all the data centers than all of what humankind consumes today in power. So if you think about if you can do it that way and all the things we you know if you can save 10% of that, how much more sustainable this world will be, right? So to me, I think it's like we kind of got too much on ease of use and getting more people to do things, but it's also creates lots of inefficient systems and coding and others where this is where I started with software engineers. This is one field where people confuse programming for engineering. It's not really engineering, it's just because if you can type you are not a writer, right? You need to know how to write. And just because something makes it easy for you to type, it doesn't make you a writer. So I think it's the same thing with programming. I think to me you have to engineer things and you need to think what's most efficient and I think other programming, I find it forces you to think efficient. It thinks about what's really required versus just making it easy to write a lot of code and then yes, you can debug it, but it just makes it hard for people to maintain and over a long period of time. I think that's a balance, Stephen.

00:26:58 [ST]

Ashok, you would take me a few minutes to go about using q to get the full power of kdb and I'd like him. I might to tease out of that 22 distinct meanings, one is that there are certain things that kdb will do that you can only really get it to do efficiently in q, so that while we democratize the access to it and provide access through Python and so forth. If you want to get something running really fast and hot, certain things, certain kinds of queries, you're gonna have to drop, you're gonna want to drop them into q. That's one aspect of the power, the other is the expressivity that you were talking about, and the way you were just saying that it disciplines your thinking. And I'm wondering if you see a future for q as a general purpose programming language besides the role it's already established for itself or getting into big time series databases.

00:28:01 [AR]

Yeah, I think the first one you know. I think the q there are lots of things it was designed to truly help somebody to code like when we talk about coding at the speed of thought is. When customers are talking to customers and what they say is that they can start working on something where some things take, you know, six months a year to write something in other languages and the q allows them to get to it in a way where it it's not just a language because it has underneath. Like if you want to compare things, what happened between two periods? Just like two lines of code which I can write to kind of say, OK, tell me what happened this year was last year or if I'm trying to predict what the COVID vaccine is. I want to compare different populations during different time periods. Just trying to do that anything with any like a Python or whatever and you it. It takes a lot of effort for somebody to define that paradigm of just, like think about like if I want to do something real time and I want to go and compare all these things that we define like these periods and comparing something and how do I iterate through it and make it? Where is it? q allows you to just express it. I got this whole thing happened last year as you can just get to that as a vector. I can just get to that compare with another one without having to go and look at the all of entire population of data, which is a waste of time and effort for example right? Most people when they write a SQL query, it takes the entire data to kind of make sense of something. I can't just pick this table and this period. Those are all first class citizens in q. And that, I think, is the power what you know it's making it, not only it terse, but it is the reason for it, because you are making everything first class citizens to compare. Or if I do joins like when I think you know you heard about as of joins [10] and other thing as of this time, what happens? Those are constructs which are not available in any other language right now. And so what happens is q is we are bringing the some of it to SQL and Python. We actually are making it easy for people to you know, as of join as a one which kind of brings two things together using Python, but it's still underneath it's calling q. But I believe that, or more more like the professional photographers equivalent they can take advantage of the q power better than you know somebody has got Python is still a wrapper around q, but it's not, you know easy to somewhere to run this, and because I think there are look there are lots of libraries and others which people have built up in Python to run in the Python process that's different than what we are doing is to run in a KDB and q process so there we are really wrapping things. Or in Python, but really it's running in q process right? So that's where there's certain things you can do directly in q is what I would say the efficiency and the power comes into picture the things where if you the other example is I would say. If you we do a lot of things in memory and q is it's in memory mapping. So whether you're storing something on storage or in memory doesn't matter it can use that as one space and you know if you processing your stock ticker symbol like if you are looking at a Microsoft and it's coming through. Most people store it as it comes through and with all the other data it's all over the place in the memory whereas with q you're able to naturally because the time and the way it organizes things it's already sorted. It's all in one column. It's in on one single memory space, so when you try to ask a question it can answer in the context of speed of thought, because it's a lot easier, right? You can try something and this is what customers tell us. I think what they say is the power of q is limited by your imagination of the question you can ask. Whereas everything else you need to know what your question is first, then you program it. Then you figure out whether it answers. So that's the difference. I mean, that's where I think the q is to think about it as it really allows you to ask your question, because most of the time people don't know what the questions are until they see something where they get insight, and then you kind of reframe the question, but most of the time, if the data is aggregated in a way where you are already not collected the data stored in that way the queries you can't answer the question. Whereas q is able to go back and because it doesn't throw away any data, it's all in memory. You're able to answer the question without worrying about it. I actually collect and store and do this in a way where programming language can answer and then I think the second part of that question I forgot now what you said.

00:33:07 [ST]

Do you see a future for q as a general purpose programming language?

00:33:12 [AR]

No, I think there is, it's a good question. I don't know the answer in terms of I think there is a the notion of what I just talked about you making it easier for like something like chatGPT and others if you actually provide the framework to people to use it, I think I've found that q as good as it is there are some programming the people who really understand it because it doesn't come with any bound. You know, like frameworks and things where you know people are used to like abstractions, right? If you have a class, you have all these things which are all available for people to kind of programming any other language. You don't have that. The discipline comes from people building their own ways of rules around it to program. That, to me, is one of the challenge which, becoming a general purpose programming language today is. It starts everything from scratch. You you need to know natively need to build everything. It doesn't have that built in capabilities like all the other languages provide so that I think is where we have to think about. How do we make that easier, right? How to do? Is there a set of frameworks and guidelines and guardrails for people to not go and create really bad code because there are people I've seen where there is people who understand the power of you can create really efficient. Things what we it can take advantage of it, but then there are people who come out of Java and other languages. They try to program it the way they know it. And it ends up being in the wrong place and where I will see a lot of problems, the customers are the second class of problems where people who come from mathematics physics background do a much better job of using q than people who come out of traditional computer science and language background, if that makes sense.

00:35:15 [ST]

Yeah, so we need it. It sounds like we need better, better collection of libraries and example code and so forth.

00:35:23 [AR]

Yeah, I think that's my thing is. Also I mean. it's a different way of thinking about the problem. You know mathematically how to solve something versus if I just want to apply programming like the way oh let me give me equivalent of what I do in Java or JavaScript or Python, what's the equivalent of that in q? I think that's the wrong way to go about it. This is thinking about it, solving the problem in a way it's different, right? It's mathematically you're solving something. Thinking linear algebra, thinking matrix thinking you know how do I represent the problem. Those are the people who are successful and that's how we make them easier. What's the starting point for them? The what is the foundational blocks is where I think it could become more general purpose.

00:36:09 [CH]

Have you had because you've mentioned chatGPT a couple of times and I think you even said that you had some folks try and use it to program with q did. Did you actually do that exercise and like find success? Because I know that several people have DM'd me like screenshot or tweeted screenshots to me of attempting to get chatGPT to write some simple thing in a series of languages. First, do it in C, then modernize it, then switch it to Rust, then do the same thing in APL [11] and everything was correct, but when it got to the APL, it did something that, if you knew nothing about APL, like it was very confident about that it had, it had executed and written the code, but it was doing completely something that like wouldn't even compile or I guess run. So like, did you find 1? Did you actually like have people try that with Q&A and two? If so, did you find success with that because I think that's like niche languages. It's like chatGPT, I think it seems to struggle, so I'm interested if you found success.

00:37:10 [AR]

Yeah, no, I think it's a good question. I mean, we are at early stages. What I found is I think if we look at the models, what we can train them on right? I mean from q, it's just the becomes based on what we are able to chatGPT on its own probably obviously doesn't know q, But what we have been able to do it to it is to kind of train based on some of the models and the things that we can feed and so based on that I think it is able to do things. What if I say, hey, do this matrix multiplication or tell me like training or tell me what is my anomaly or you know it's kind of finding using the what's the equivalent code of q to solve the problem right? In terms of the market data is coming in real time and I want to tell you if this particular is somebody spoofing you or whatever, so depends on what we kind of train and what we provide it. I think we are starting to see that, but I just feel we have to do KX in the Community. I want to be able to kind of create that large language model for API like q. I mean that's potentially where I think we can do and I think that's the, you know, it's early stages at this point, but I see that I think is it, Stephen, you're sharing something and it has done some.

00:38:29 [AR]

You know we are. We're doing some things like that. What he's showing here. Where it can give what's equal and of writing something in q where people don't have to know.

00:38:38 [CH]

Is this something you did Stephen? Is it Stephen that is sharing? I can't tell what it is, Stephen.

00:38:43 [ST]

It's a it's one of our colleagues, Brian, and it was the third of the screenshots of we had received and in the first two in the first two chatGPT had answered some questions fairly sensibly, with some simple with some simple stuff that's correct. Then in the third it was asked to write a function to list files in a folder and sort by size. And when you glance at it, you think? Oh wow, that's pretty good. And then when you look at it, you see hang on, that's not actually correct.

00:39:25 [CH]

I don't see it kind of looks q'ish, but.

00:39:30 [ST]

Yeah, so I think.

00:39:30 [CH]

Yeah, I don't see a sort function anywhere.

00:39:33 [ST]

I think what I think what you've got here is that chat bots are really good mimics and the mimicry can spill over into bluffing. So to get to get it to guide people, we've got an interesting journey.

00:39:50 [CH]

It's interesting though what Ashok, what you were saying, though, is that. You kind of, to me, it sounds like you're describing like a prompt engineering system. Where basically it's a tool that you can build. That is, you know, ChatGPT flavored. Where you basically just ask it things and then somehow you train the model behind it to basically use q and kdb+ to do like, you know, very fast querying of time series databases to whether it's financial applications, which is, you know, typically what right now what q is used for in KDB+. That could that could be like a very interesting because you could imagine people sitting at like a Wall Street trader desk that are not, you know, technical in nature, but that wanted to throw a bunch of queries at it and the idea that you could just speak to, you know, some echo device. You know, currently today does not exist, but you can imagine combining these two technologies, that yeah, you would need to do a little bit of work. You'd need to feed the GPT model. You need to train it on, you know, a corpus of q queries and stuff. But you could imagine that like in the near future, that would be something that you could hand some text to, you know, internally q to like answer of querying, you know whether it's time series data on Wall Street or something else like that. That could definitely be a thing and it's very interesting to think like what would you do with that kind of technology. I'm sure there's applications way outside of Wall Street that that would be useful for.

00:41:22 [AR]

I mean that I think is exactly, you know, we look at more when we talk about democratizing. It's really the people. What they do with data and how they make decisions. And a lot of times people try to say I read the data scientist or whoever is going to do that, but generally it's a line of business person. It's a trader, it's a, you know, operations person is running manufacturing. They're trying to figure out efficiency or what they can state the problem and what the question is, but then you know translating that it takes the army of people and it's sometimes it gets lost in translation by the time you get the answer right. So I think to me there are some things which are. I mean I would say to answer q and I think that's part of the perception. It's been, it's. Weak kdb is used in capital markets quite a bit, but then what I found is we have a lot of people using in the last year or two on other use cases. We have a capabilities in clinical trials. The things what I find is more of a pattern is: How do I look at the anomalies? Who? Observability is a big area where we look at systems and being able to make sense of data. So there is a custom you know when the customer of ours who does clinical trials for 92% of the companies in the world who do clinical trials. [12] And they were involved in finding the vaccines for COVID. I mean that took used to take six years to find something and they compress the time to one year, and now they're trying to compress it to 30 days. Like with the number of variants and other things going on. So they're actually using. They used to use Matlab [13] and now they're using q with kdb to look at massive amounts of data, but also compare things, which countries are efficacy perspective. And now they're increasingly looking at how do I look at healthcare data as a service to kind of start looking at. How do I solve? How do I find the right profile of people who could... That's one of the problems they have, you know, in the world you know so many people and billions of people. So I think that's something which should we find that medical devices to people like aerospace and defense, like we've got planes flying around and trying to find in enemy territory which ones are missiles versus, you know, public population. Like if you look at Ukraine or Russia is doing to me those are these lot of use cases. What we're finding is they're applying this because these are massive vector hungry data use cases. If you're trying to solve the problem of genetics/ biotech these are things which people run. Whereas it takes days and months to sometimes, and if you apply this sort of a technology in array based and vector based thing, you're able to compress the time to very little where we can get answers or at least you can try different things so that I think is what I'm finding is that... We are seeing a lot of the use cases around manufacturing semiconductors where, you know the fabs I mean, we have a shortage of chips right now. And every time you find a fault in a chip it you have to throw out the whole thing. And that's millions of dollars of cost and then it creates supply chain problems. So I think there's lots of these things, like in telco and quite the experience of customer. You know you're finding the time to make decisions is compressing and then if you want to even take the Wall Street coming back to capital markets to settle trades in you know, six days it went to two days. Now there is a requirement in US and Canada by next year you order settle trades in what we call t + 1 from T + 2 to T + 1. So how do you settle things? It has to you. Not only am I, you know, buying or selling things, you have to transfer the shares. You have to make sure that verify everything you know who bought, who sold and you don't have to have full business day to settle that right. So that adds, compresses time and so think about the amount of volumes of data you have to process. What kind of analytics you have to apply. You're looking at risks. You're looking at foreign exchange. It implies petabytes to zettabytes of data you have to process. And so to me that's I think is where it's going towards. I mean, we really are compressing time and now we have to think about all areas where that happens. Whether it's finding vaccines or, you know, you're trying to find in split second whether I'm targeting a enemy system or trying to do that for, you know medical devices. So that I think is, I just want to clarify that you know, there's more and more use case what we're finding beyond capital markets.

00:46:25 [BT]

Use case that I'm aware of is is Formula One racing [14] which people may not have realized that a lot of KX has been involved with. Many of the teams I think you're still involved. I'm not sure going forward. But last year was at Alpine and Williams were the two that you working with.

00:46:42 [AR]

Yes, I think Alpine, you know, is a good use case. I think what we found in that one is the actual use case is 60,000 different sensors coming in. You know, even during race, they collect all this data and then they have to process and try to determine how to optimize for the race. What I what I need to be looking at. But also, they're also using that to design the next generation of cars, so there's two elements to it. How do I look at all these sensors? It's really a IoT or an edge type use case. And you know now the that's what I think is going from a time series data to machine data sensor data. And now you're getting more and more data and so Alpine, and we we still work with them. They are now increasing the amount of data, but they're also using cloud because that's becoming the way to aggregate data. And we're working with some mothers in that context also. I think that's the very interesting that I look at it as not just solving a Formula One racing problem as much as the telematics. The how do you bring all this data and being able to make sense, and I think using array based approach to solve that is how we can get to very fast insights and then you know you think about how much time they have to react between the laps. And now you can also apply that learnings to the actual design of the cars and self-driving systems.

00:48:14 [BT]

And people may see Alpine is not one of the top teams now. They're sort of the second tier. But one of the things that point out to them was the year before Red Bull, who's a very design driven team, won the World Championship. They were using KX, which my dad was involved with, Formula One and he often said if you want to see how A-Team successful look what they were doing the year before they written, they won. Because that will tell you what the right thing to do is. And KX I do notice this because I saw KX as one of the sponsors and one of the emblems on the Red Bull cars and they weren't the next year, but that was the year that Red Bull One had continued to win. So I think it sets that foundation for those kind of things they got to step up on a number of people. And now other teams are trying to use it as well. I'm just wondering whether there is. When I think about areas like that, that you said there's Internet of Things, and then there's time series. I can see so many things like in terms of energy managing, balancing, electrical grids, those kind of things, more people coming online with their own solar power, and trying to figure out when you're going to feed a system or take from the system. I would think that would be an area that KX is probably looking at as well.

00:49:30 [AR]

No, absolutely I think you know the whole notion of smart grids are big. Every where we're starting to use smart meters and smart grids. So we actually do that for the whole country of Finland. And we have got several countries. Poland is the next one which we are starting to do. And especially they're so worried about the infrastructure. In kind of so, we have this what we call a Fingrid. [15] So we manage the entire Finland grid and it's all powered by kdb and q and it's processes so much information but also improves the efficiency of the whole energy system and then you also are now doing replicating that across different countries. So yeah, that's why I think is something which we continue to look at because to me, the world, the amount of power is being needed. The efficiency, we waste a lot of power and so trying to you know, apply the efficiency and we talk about sustainable, you know, Stephen has done a bunch of things around it to show how little power we can consume, and that's one aspect. But then just looking at all this data and being able to optimize it for entire countries to save power and then you know countries like India for example. I mean there is literally, the country is growing so rapidly and they skip the whole infrastructure problem by moving to wireless. And now they need so much power because everything India is the number one country for AI right now and that's what Satya Nadella was in India couple of weeks ago talking about how they're putting four data centers in India Azure, and it's become the number one country to apply AI. So all it means is you're consuming more power and think about where the energy comes from. And they're dependent on imports and oil and other things. So there is definitely that's a big thing what we can do and not just from a as a KX as a company or product, but more from just the humanity and what we can do to address the bigger problems. The energy and some of the things that we can do around saving lives finding a it's continued to COVID it's new variations and it's rapidly evolving. You know, it's dynamically changing. So it's just like how do you, the people talk about, predicting the landing of an hurricane. Think about that problem. If you want to land predict the landfall of hurricane. The type of things that you need to look at dynamically all the things around this, wind speed and everything else happening dynamically to where it's going to land what happens with systems you bring all of it combination, and that's truly the type of system when you try to predict the next COVID vaccine. Or you know the type of things you can do in energy. So that's the type of problems I think we are involved in and continue to solve because again it's all powered by the vector and array type things.

00:52:20 [CH]

I'll be, I guess, remiss if I don't ask you. You've mentioned now, like a bunch of different. I don't know if you want to call them success stories, but just capital markets, hedge funds, you know, vaccine research. I had no idea about like the the Finn grid and entering Poland. As the CEO I guess you're one of the best people to ask is are there. Is there sort of like a a huge success story? Other than the ones that you've mentioned that like people wouldn't necessarily know about that KDB+ is being used for. That you're able to share. Obviously, you know you're not able to talk about all the people that use your your software they've got confidential stuff but. Like just the finger of one I was like, oh like you're powering you know something for a whole country that's like a pretty massive thing to be able to say about a technology that you're building that I had no idea. About like are there other things like that?

00:53:03 [AR]

Yeah no, absolutely. I think you know one area where we are doing a lot is around semiconductors. So right now the whole world, everybody's investing in semiconductors, and we have a company which basically powers pretty much all fabs in the world for example. It's used KDB and q to support for them to support fault detection in manufacturing of in all the fabs right, these are disconnected wire gap systems and as I said like it's kind of finding those faults and being able to predict things before they fail. And these are massive systems. I mean, these are fabs a billion dollar or multiple billion dollar things world. Right, there's about 160 of these where we power and we find faults, so that's one area semiconductors. Manufacturing you know when you look at people who manufacture medical devices and like, for example, I'll give you an example of one of the largest manufacturer of stents [16] which goes into human hearts and that is such a big critical safety critical system. I mean, if you put somebody a bad stent in somebody finding problems in stents, and how do you make sure... And that company is now not only doing it in one factory, they're doing it across all the medical device across 70 factories worldwide. So that to me is a big thing for us, I mean in medical devices, clinical trials, but generally life sciences, I think, is an area where we are seeing increasingly people are trying to solve the problem of the biotech genetics. I think we have several people who are using this for research in life sciences, so finding the next big drug in pharmaceutical. I think that's and these are. We're hoping to announce we just announced this morning there's a press release around Enterprise Web, which is in the telco space, which is basically between Intel, Red Hat, us and a couple of these companies where think about what's happening in the data, which is the 5G is creating massive amounts of data and networks. And how do you actually process all this in the networks and make sure that you can continue to provide the experience? It's huge. I mean, that's where all the telcos are investing in the networks. And if the network fails in a world, comps stops, you know, think about what happens during COVID. If we didn't have access to the network. My I remember when we lost power in my house in Texas and we were out with our power. My kids were running around you know. They're saying they still had network, and if as long as there's a network, they're. Saying they're connected. They're happy there. Enough power there, enough water, but they're saying I've got network. So I think we are looking at this as you know, the big 5G and telco is a big thing for us, so those are. I mean, I think the some of them I mean we I can name the partners we were working with Microsoft. [17] He's a huge partner with us. We are working with them on many manufacturing things, which is what an Azure. They use us as a strategic capability. We used to be not just on Prem customers, so Microsoft is using us not only in the financial services. Now we are working with them across healthcare, what they call industry clouds, pretty much in all the clouds in the industry perspective. That's where everybody is going now. With the industry clouds means you're solving specific problems, so anything with this shape of the problem, time series, machine data, high speed analytics, real time. We, Microsoft is using us of the. This is the second most valuable company in the world, the hyper scaler. They're using kdb and kx insights for that, so I would say you'll see. The names the top 40 companies, the biggest banks in the world. Are you more stuff, right? And that tells something. But now we are actually saying that the top ten brands. Whether in different industries are working with us right now through Microsoft as well as some of the what we're doing with AWS. So I think that's I'm hoping to share more as we go through but yeah, the big thing we're doing is we're going to announce the new kdb insights at the end of the month, which will have Microsoft Visual Studio Code. We'll have more support. We also want to open source. What we call a PIKX [18] so that we haven't announced it publicly, but that's something which I want to do to open it up for more people to contribute and basically, you know, provide that enable the community to do more use cases versus be dependent on just us. We want enable that we do. We want to open it up. Like that's something that you're massive, you know investments we're making and we are also big partners like Microsoft. Teaming with us to do that.

00:58:06 [CH]

Go ahead Bob.

00:58:07 [BT]

You were mentioning earlier that top end q programmer is developing their own frameworks because that's they can tune things to exactly what they want. And to me, that's really impressive but and to hear you just say about opening it up, I think that does create a challenge. Being able to to configure things so precisely does create a challenge to being able to open things up across industries because as soon as you can be so insular you tend not to share as much because you've worked on that power. But it does sound to me like if you're opening things up by making it wider across companies, you may get the benefit of having a number of different groups work towards a common goal, which I think obviously the business plan for KX is working for KX really well, but I do think it might have got in the way of that opening up processes. Do you see that? As well.

00:59:05 [AR]

No, yeah, absolutely. But I think you know as a company. What I find is we as a part of the broader FD. We also are in the consulting business and so some to some extent I think we kind of looked at everything. Was somebody else did with q was competition to us. Instead of and it doesn't matter who created frameworks or what we they did to enhance that, I think is a fundamental shift. You know what my perspective, which is we really we are going to be a software company to partner with customers and not be doing everything ourselves. And they don't have to come to us and you know, even for like we are cryptic error messages because you know, we people will call us and get support from us and we, you know, that's insane. So we are trying to make it easy for the communities and you know a lot of essays and others which you'll see some increasingly we are announcing some partnerships. Where we are not in the services business, we don't want to be making money off services for KX. It's as much of we want to be the engine to power these sort of applications. People are building to solve the biggest problems and the more people can create this frameworks in a more open source we can do more sharing our customers asking for us they want to share. So we want to enable. How do we enable this sort of frameworks? The best practices and others where people can share and use time to value to get to that you know at the end of the day. I always think about getting to the race is our problem, but once you get to the race we can help people. So let's start make it hard for people get to the race and we want to be the engine. To power that versus trying to build everything ourselves.

01:00:46 [BT]

And I guess the thing is, you've got a big, powerful, fast engine. It's also expensive, but that actually puts at the scale of these larger country sized energy grids and things like that, where those groups can actually afford that to be able to use that engine. It's not something you're looking at. As much for you know, single users or anything, although I think you have created a market for people who might want to learn to program in q by opening up the personal, you know the licenses that you can learn to use the q. But you probably will never be able to afford the engine on your own.

01:01:26 [AR]

Yeah, no, I think you know that's the thing what we are trying to do is. First of all, I think we have this KDB and q now available for personal use for free [19] and now they can use it. You know the license so anybody can go to kx.com and get that. So I found you have hundreds of downloads from different companies, but also the top schools that most people were downloading from Stanford from Princeton from Cambridge, Oxford, which is great to see. I mean we have got lots of PhD and quants and others are using that. So which is great. We also are going to have a trial of the entire capability end of the month where people can use it. It's more once they deploy to production and they start using a big companies is where we want to monetize but not, you know, let people build for a lot of different things themselves they can use. So I think the idea would be to more focus on you know where we want to move up. The higher the stack, but not prevent you know from people and that pricing licensing in the context of cloud, it's much more of a consumption based and what people can adopt and they can start free and then they can get to higher value. And that's really the approach we're taking and you'll see you know more and more of that.

01:02:37 [CH]

One of the things you quickly mentioned in the stuff that's going to be announced with the KX Insights release. What you said Visual Studio code [20] does. Is that mean there's gonna be some kind of extension that is usable with KDB+ or q or something like that?

01:02:50 [AR]

So we have a a partnership with Microsoft where we actually. If somebody can start with Visual Studio Code, there's a plugin just like you have different language plugins right from Visual Studio from Java to J. So you have a Q plugin to which kind of provides debugging to all the documentation and there is a pyKX or what Python interface to KDB and q. All that is built in and they can do everything without having a separate Python environment for example, and debugging, auto complete. To how they can debug. How can they troubleshoot. They can do it all with just Visual Studio Code without getting anything from us, and it's going to also use the standard things like Anaconda and pip install. It used to be everything had to be, you know, kdb or kx versions of these things, and you had to know about that stuff. So yeah, so by everything that we're doing, Visual Studio Code is if you know Visual Studio code. If you're a developer or you're a data scientist or data engineer, you can use it in the same environment and the workflows you're used to and develop q code. Or you can develop in Python or SQL. And be able to generate real time applications.

01:04:03 [AB]

But you said without getting anything from you, but the actual language engine. And it has to be there somewhere, no?

01:04:13 [AR]

We are enabling that as part of, you know you can get the kdb q the free version, the personal use, they can use it with this thing and what we are packaging it up so we'll have that somebody download and use it so that way you don't have to get multiple things and figure out to get a license and.

01:04:23 [CH]

So it's part of the extension.

01:04:31 [AR]

How to set up a queue on your machine and Python so those are all the things that you know we are on January 31, we are going to be in making available and some of it was in alpha and we want to make sure that you know what we have the real customers using already and giving us feedback before we put it out. And then obviously we'll expand on it.

01:04:52 [CH]

That's super we'll definitely make sure to. Well, if we don't go and find a link and add it to this episode. When it gets released, we'll add it to a future episode, but that is definitely, I think something a lot of our listeners will be interested in because. Currently I'm not sure for all the languages, but definitely there are extensions. I know optiva I think was the people that made like an APL one, but like you still whether you're using VS code or you're using the ride, you have to go and download an executable for the language. Which it's not a, you know, it takes 5 minutes to do, but having something that's all in one. And I guess it's. Pretty simple to do for k because or not k, q because the executable executable famously is is quite quite small, so it doesn't actually increase the the size of the extension by, you know a GB or anything like that, so that's we'll definitely make sure to add links when that gets announced because...

01:05:44 [AR]

Absolutely, now we'll do that. I think the you know basically our goal is to make it simple and with one line code. Or, you know, with a few things they can get the value and. That's I think the other thing is, there were some extensions by some other people had written some of it, but it wasn't robust and it wasn't having all. So we have made all our debugging to our libraries, everything available, working with Microsoft. So that's why it's much more robust and it's going to be a first class citizen within Visual Studio Code.

01:06:11 [CH]

Interesting, yeah, that that definitely sounds, I'm sure, if not all of us, most of us will be checking that out once it drops to see, yeah.

01:06:21 [AR]

Absolutely. And also I would say that you want the pyKX or Python. Your performance was a big focus for us, so if you do something in just Python, if it is within 1% of the performance of what you get with q, so I think it's been amazing with a lot of feedback from our customers is we are talking about the largest banks and others who are using it and we're getting a lot of positive feedback on just using Python because a lot of times they may have one or two q people in the company and then they have 95% of the people who just know SQL and Python so you can do ANSI SQL and you can use python without knowing as much on q, but you can get, you know, build really great systems.

01:06:59 [CH]

And the overhead is 1%.

01:07:01 [AR]

Yes, the Python the performance is within 1% of what we get with q.

01:07:05 [CH]

That is wow, I'm saying wow because uh, my former team at NVIDIA that I worked at we had a well, I guess it's slightly different, but anyways we had a Python interface to a C++ and GPU accelerated library and I think the overhead was in some cases like a factor of 10, which compared to previous you know Python, you know it was 1000 times faster when you. Run the C++ so, so dividing that by 10 it's still 100 times faster and it's people are still happy, but to only have a 1% overhead for for the Python is quite impressive.

01:07:44 [AR]

Because, yeah, absolutely. I mean, that's why I think customer service like it runs about 100 times faster in terms of the performance. But it's like overall, you know, 1/10th. The infrastructure or cost of anything else you they need to use because you spin up lot less compute and storage and other things. So that's the other thing which I think q does is the compression. It does because of the way it uses real time. And others we go, you know you wanna somebody's using your street, it's 90% usage, it goes to 30%. But what's not intuitive, because the way we do it, with some of the vector based things. Most of the time when people think I compress, I lose some fidelity of the data. We don't lose anything and we kind of also... When you unzip, it's like a zip. When unzip people think it takes time to zip and then unzip, but it's actually... because the way we do it, the compressed data we run as fast or better than uncompressed data, and that I think is very few people realize that. I think when they actually run this is when they realize the power of it. And those are the some of the things where you know we want more people to experience. So through some other things, because it's not very clear until they actually see that the wow is wow. It takes. I can run a query and so that's why I think we are now working with Snowflake [21] to bring the power of q to that data. We are working with them on making q and Python to run in Snow Park, which is their Python implementation and how can I get more value to the data. Because a lot of people are putting data into Snowflake, which as a partner we are going to bring the power of q and the vector based processing to Snowflake and we are going to do the same with the databricks. So this is something where being an engine means that you want to bring the cord and the power of q to the data not move the data to what you're doing in other languages and programming, so it's a completely different paradigm.

01:09:38 [CH]

I mean, those testimonials are are amazing. They're the same team that I worked on they one of their clients was Capital One and they had a testimonial from them that it was 100. I can't remember if it was the two numbers were 100x and 30X. I can't remember which one was speed and which one was money, but one of them were applied to each and I think it was like, you know: 100X faster and 30 times cheaper or something like that. Which like doesn't really make sense, but when you're running something, even the the GPU that they're running on them costs more, it's a fraction of the cost because it's so much faster, so the the compute that you're paying for in the cloud is. Yeah, it's pay less to go faster.

01:10:22 [AR]

Yes, yeah.

01:10:22 [CH]

Sounds like a scam, but. But it exists out there.

01:10:26 [AR]

Now I think that's why there's a whole discipline of finOPS now. With the economy slowing down, everybody's focused on when we talk about data. I mean, every query costs money, and the more people you know everyone to be data-driven. But the problem is you're just spinning up more things on cloud and running more query. But if you don't think through it, you're gonna bring massive volumes of data to just to answer question like the query brings to 20 terabytes of data to answer something a trader will want to do something and he wants to say with, but you just need to bring what's necessary. That's the power. That's why I keep coming back to the array and vectors is you just focus on what's needed to answer the question. Instead of trying to bring the whole data across the wire. And then the cloud, and that costs a lot of money. So the whole discipline of finops where people are trying to figure out how to optimize my finance and being data-driven is going to cost a lot of people a lot of money. But you need to optimize it to what questions you're trying to answer and don't need to move the data around just to do this thing and that's the paradox.

01:11:28 [CH]

Yeah, it's important to get the. I guess designing it correct otherwise you end up spending a bunch of money that you don't spend. I realize that we've, as always, flown by the at the hour mark. Are there any sort of pressing questions that panelists have we got one from Adam? Before we before we close up shop here.

01:11:48 [AB]

And we have to ask the k question, don't we?

01:11:52 [CH]

What's the k question?

01:11:55 [AB]

Well, all this talk about q mean the company is even called kx, which I suppose it takes the name from the programming language k but now you're in a position to maybe have a say, or at least an opinion. If you can. Say that what's with the k programming language in your future there at all? Or is it just going to be q?

01:12:17 [AR]

So I mean, I think it's a good question. So one thing is based on my understanding the k name came from, k stands actually for keys to the Kingdom. So they're actually giving people keys to the Kingdom is what I understand they came from, but KX actually is true that it's supposed to have been.. Initially the company was formed where everything was named on the k versions of it, so that's why the X was instead of changing the name of the company every time there's a new version of K, you end up playing KX, so you you know it. Represents multiple versions of k.

01:12:52 [CH]

Really, I didn't know that that's that's would be X&K X wait, Steven, shaking his head no.

01:12:58 [ST]

I heard a different story. But it doesn't mean that mine's true.

01:13:05 [AR]

So yeah, but. I think you know to your point. To me, I think it's more q. I mean I'm allowed to meet with Arthur one day. And we what to look at how do we? Kind of look at what makes sense. And it's really I always start with the customer problem and what we need to solve versus you know. So technology is there to solve some problem. So I think we have a really brilliant team in... Charlie and others who are working on Q and they've made it practical and we are, we're solving real world problems. So if there are things in k and the type of thing that we need to bring forward to kind of enhance and go forward. That's where we're starting to see things like in machine learning and AI they're doing matrix multiplication and trying to do some of the things that we want to do. We are not very efficient today. I think we can do things better. So I think I would like to take a look at it. I wouldn't. I don't have an answer in terms of you know what we would do with q or k and. It's not necessarily q is the answer as much as what's the problem it solved, and then how do we go about thi k. The new versions of k can give us some things to do differently, so that's the way I would think about. And you know, at some point, maybe in the future, we can look at that and I don't have an answer beyond that at this point.

01:14:18 [CH]

Any other pressing questions that we panelists want to ask before while we have? The CEO of KX. Maybe we'll maybe. On stay tuned to next episode where we get Steven Taylor at the top of the episode to tell his story of. What the X in KX stands for?

01:14:34 [BT]

I'm beginning to think we should have had a show-con, as part of the announcements.

01:14:39 [AB]

Yes, this sounds like all announcements here.

01:14:44 [BT]

Hopefully we don't fall under a non-disclosure agreement or something, and this will actually be on the air before, but I think we'll we actually may have a scoop in a number for areas.

01:14:53 [AB]

Areas you heard it first here on the arraycast.

01:14:57 [CH]

Yeah, the number one array podcast in the world. All right well. If we don't have any other questions, I'll say thank you so much for coming on Ashok. This was awesome to be able to ask you questions and to hear a little bit of thoughts from from you personally and where the future of kx lies and definitely thought provoking to think about the possibility of a, you know prompt engineering system powered by, you know, chatGPT + Q + K and or kdb+ and yeah, hopefully in the future. I mean, we've had on Gita, who's the CEO of Dyalog Limited. I mean jsoftware, I don't think officially has a CEO, but that would either be Eric or Rich who we've both had on. And if there's a CEO of BQN, that's probably Marshall. So I mean, we. Could say now that we.

01:15:39 [ML]

Not a company.

01:15:41 [CH]

It's not, but if we if we were going to point one, so we basically now had all the all the CEO's on see and maybe in the future. Whether that's a year out or a couple years out. We might like to have you back if you're willing and sort of see what's changed in the last two years and sort of. You know, especially if you get this chatGPT system up and running, or you know probably will be called something different by then, because this is the model Now. But I'm sure there's going to be a A GPT 4567 and then they'll call something different, but it would be awesome to have you back in the future to follow up on some of these future plans to to see how it went.

01:16:12 [AR]

Absolutely. Thanks for having me and it's been an honor. I mean, I think I've listened to many of the arraycast. And I know that the audience is, you know, I always feel it's a brilliant audience in terms of the type of things people have done and continue to do so. So I would love to be back and I think I do think that it's a pivotal and I it's a we are at inflection point and I think as a company and what we're doing with some of the hyperscalers and hopefully you know more customers will get benefit and the ecosystem.

01:16:39 [CH]

Yeah, awesome and we will. As always, we'll leave all links in the show notes for folks that want to check out and be sure we'll all the things that are getting announced in Jan 31, we'll try and come back and link them here, but if not, we'll we'll link them in the future episode end over to Bob for the closing. Contact us at.

01:16:55 [BT]

The usual contact@araycast.com [22] if you want to get in touch with us and we've had lots of good suggestions and we look forward to interacting with our community. It's always a lot of fun and those emails get sent off to the slack channel that we're all part of. So can they get discussed there and sometimes you get answers back. But know that they're read, and they're appreciated. And also appreciation note to Sanjay and Igor, who help out with the transcripts which are also available on the site. So if you I know there are some people who as much as they love you, know, our voices are wonderful voices to listen to. I know there are some people who actually just go through the transcripts and. And read them, which is fine any way that you would like to consume us we're happy to be part of. And I think that's all. I'm going to say.

01:17:42 [CH]

Yeah, well, the transcripts are are awesome. We we think both of them and I actually heard the other day that some people would listen to podcasts, but they don't because there's no transcripts for them to search for the part that they're interested in, which I thought is interesting, is that like having transcripts actually probably increases the listenership, because a lot of folks don't have the patience to listen to all 60 minutes, they just want to. Search for kdb+ and and go straight to that one thing. But yeah with that I guess we'll say once again, thank you Ashok for coming on. This has been awesome and we'll say happy array programming.

01:18:14 [ALL]

Happy array programming!