Introduction to Alphalist Podcast -- Tobias Schlottke: Hello friends, this is the Alphalist Podcast. I am your host Tobi. The goal of the Alphalist Podcast is to empower CTOs with the info and insight they need to make the best decisions for their company. We do this by hosting top thought leaders and picking their brains for insights into technical leadership and tech trends.
Tobias Schlottke: If you believe in the power of accumulated knowledge to accelerate growth, Make sure to subscribe to this podcast. Plus, if you're an experienced CTO, you will love the discussion happening in our Slack space where over 600 CTOs are sharing insights or visit one of our events. Just go to alphalist. com to apply. Tobias Schlottke: Welcome to the Alphalist podcast. I am your host, Tobi, and I'm. Meet Shyam Sankar, CTO of Palantir --
Tobias Schlottke: really honored to have Shyam Sankar, the CTO of Palantir here today with me in the podcast. And we are talking about how and why Shyam will deprecate back end programming in the few years from now. Um, and for the ones who don't know Palantir, um, it almost got like 70 billion in market cap.
Tobias Schlottke: That's really crazy. Um, on the NASDAQ. And I think most people here know it from, or most people here don't know it. Like Sham, I hope you, you, you, you're going to tell us more about like what you do and how, because you're working a lot for big corporations and, uh, governments as well. Um, and, and, and most people think you're, you're, you're doing basically purely data analytics and you're doing much more.
Tobias Schlottke: And we're talking today about that and also about your experience as a CTO, et cetera. Uh, so, uh, welcome to the podcast, Chum. Well, thank you for having me, Tobias. It's great to be here. Early Days at Palantir -- Tobias Schlottke: So maybe we start a little earlier in your career. I think you, you've been like the. Fifth employee of Palantir, or no, the 13th employee of Palantir.
Shyam Sankar: 13th. Yeah. Lucky number 13, at least for me, . I joined Palantir almost 19 years ago. Really? Uh, I, I was the first person who was not working full-time on the code base. I don't think they thought my code was high enough quality. They didn't, they didn't want me to pollute it. Uh, but I started and built the team of four deployed engineers.
Shyam Sankar: As we, as we began to call the practice of people who would go engage with the customer in the field. And actually rapidly prototype solutions, identify the gaps. And it was an outgrowth of Alex Korps, our CEO's philosophy that, you know, one of the reasons that French restaurants are so good is that the wait staff is actually part of the kitchen staff.
Shyam Sankar: You know, they have a deep and intimate understanding of the food and how to manipulate the food in order to deliver the experience. And we wanted to build something much more like that than a commodity way of throwing software over the shelf, over the wall, assuming it would work or not work and have a much more intimate relationship with both our software, our customers problems and our customers themselves.
Shyam Sankar: Okay. So you're in the kitchen and you're, you're, you're, you're out there in the field. That's right. Great. Impact of Florida 'Space Coast' Childhood -- Shyam Sankar: But, um, maybe, maybe a little earlier, um, beforehand, like, why did you actually get into computing? Like, why, why are you like, I don't know if you still consider yourself a nerd or a geek? Why, why, why are you one?
Shyam Sankar: Oh, absolutely. I don't know, maybe there's kind of a metaphysical sense in which you were kind of born that way, but, um, you know, my dad would always try to get me, even though he could barely afford it, some sort of latest gadget to play with. I remember having access to like a Commodore 64 and, you know, going over to my friends houses and trying to program them.
Shyam Sankar: I was profoundly affected. I was, you know, I grew up in the shadow of the space coast in Florida. So when the shuttles would launch in elementary school, you would go out to the courtyard and you can actually watch the shuttle, you know, and when the shuttle would re enter, usually, it seemed somehow karmically, it'd always be Saturday morning at 6am, you would wake up to the double sonic booms of, of its re entry.
Shyam Sankar: Uh, and so there was this, Definite optimism that was just, you were imbued with of what humans can do with technology and, and why this would be something that you want to spend all your time working on. So you found yourself like curious about how to build software for the shuttle then at a certain point or? Power of Software --
Shyam Sankar: Well, all of the above. Like, I mean, I think software is one of the most accessible, easily manipulated, you're not going to build a rocket in your backyard, likely, or at least maybe baby Elon could. But, you know, I think, um, having access to pushing technology frontiers through software, I remember in seventh grade building a, you know, a simple version of, of, uh, Scorched Earth, that, that, that game where you'd be shooting tanks and getting to play with, like, trig and some basic geometry and, and Encoding that into BASIC at the time.
Shyam Sankar: So, basically, you started off with BASIC then and, and, uh, Went from there. 10, 10, print hello, 20, go to 10. And, and, and, and, and from there, like, what did you, you studied IT or? From College to Silicon Valley -- Shyam Sankar: Yeah, in college, The hubris of youth, you know, I spent my, my high schools as a, uh, as a programmer. So I felt like, okay, I know enough of this computer science stuff, which I definitely did not, but the hubris of youth.
Shyam Sankar: So in college, I spent all my time as electrical engineer, really getting into the hardware Move to LLMs like Analog to Digital -- Shyam Sankar: . Uh, with a focus on digital signal processing and, and uh, electronic waveforms, which was quite useful. I think quite useful in this moment in particular, because I think as we approach leveraging LLMs, I think it's, it's, um, closer to the revolution from analog computing to digital than people realize.
Shyam Sankar: And we're operating as if we're already in a digital world. And there's a lot of, you know, at the fundamentally we model what's going on in the CPU as digital, but really, it is an analog. There's an analog signal that's going through there that we've done a huge amount of engineering to so that we have this wonderful abstraction of zero and one.
Shyam Sankar: I think we're gonna go through the same sort of process with LLMs. I digress a little bit, though. But yeah, in college, I spent a lot of time as electrical engineer. After college, uh, really when I graduated from Cornell. Almost no startups were recruiting there. Google didn't even recruit at Cornell when I graduated.
Shyam Sankar: So I headed out to the West Coast. I went and did a master's at Stanford as a way of getting to the Valley. Within two months of being in the Valley, I had a full time job at a startup as the fifth employee, a company called Xoom with an X that had been seed funded by Peter Thiel and had just raised its Series A from Sequoia, and I was off to the races.
Shyam Sankar: So I did that for about three years before joining Palantir in 06. So that's basically where Peter Thiel discovered you then, or how did that go? Yeah, that might be an overstatement, but that's where I started getting the gamma radiation from the PayPal mafia would be a more accurate, accurate representation.
Shyam Sankar: So you, you consider yourself as part of the PayPal mafia then, or? I consider myself as someone who learns, you know, voraciously from that mafia, but I, I'm, uh, I'm a generation beyond, unfortunately. But I do consider myself part of the founding part of the Palantir mafia for sure.
A Day in the Life of a In-The-Trenches CTO -- Shyam Sankar: So, um, I'd, I'd be, I'd be curious how your, your, your day looks like, like, I mean, do you have, I don't know, lunch with Elon every once in a while or meet Peter Thiel every few weeks or how does that look like?
Shyam Sankar: Well, certainly at board meetings and, and, uh, things like that. But look, I'm in the trenches. I consider myself someone with my sleeves rolled up in the engine room with grease on my elbows. I'm trying to visit as many customers as often as possible, not to tell them about the roadmap that's to come, but actually to discover what roadmap ought to come.
Shyam Sankar: What are the empirical gaps I'm observing in the field that I know we can solve and that we should be organizing around and how do I synthesize that information? And I spent a lot of time with our teams looking at what they're thinking about and how it's coming together. That interface, I almost think about it as like the, the kind of very product aspect of what's happening in the engineering org and the engineering part of what's happening in the field.
Shyam Sankar: Okay. And, and, and on the, on the engineering part, like when was your last Git commit? Oh, that's, that's embarrassing. I don't think they really let me, yeah, no, no, no, that's, um, yeah, I won't humiliate myself by saying. Okay. Okay. So that's a while ago. Uh, but, but I think it's also. The fundamental shift that more and more technologists these days, um, are, are pivoting to product, um, and, and, and see how much value they can generate, uh, on the product side and, and potentially on the sales side, right?
Engineers in Product and Sales --
Shyam Sankar: I, I think that's one, one of the, the things you're, you're advocating for as well, right? That, um, it can be productive if engineers end up in sales. Is that true? Yeah, I mean, I would we would kind of bristle at the word sales. What does that mean exactly? Like our kind of macro critique, certainly my critique would be that You know, most software is thin, that unfortunately, the strongest argument to push engineers closer to sales is that most software companies seem to somehow be designed sales first.
Shyam Sankar: Like, what is the box that we think we can sell? Oh, let's go build that box. Versus, what needs to exist in the world to create fundamental value? Why shouldn't we build that? And yeah, it might be hard and messy and complicated to figure out how to sell it, but that, you know, you got to create the value first.
Shyam Sankar: And my lived experience with this is like, over the 20 roughly years I've been here, there's always been some company that was going to make Palantir relevant. And it never did. There's probably been six or seven waves of that where investors are like, okay, you're done now, you're done now. And it's this deep commitment to continually pushing the frontiers, the product perimeter.
Shyam Sankar: that you can only observe empirically in the field when you don't define yourself by what box you're in, but what boxes don't exist that the world absolutely needs. Value of Contradiction -- Shyam Sankar: \What you explained initially, basically, first see what needs to be solved and then build it. Isn't that like more like the European way of doing things while like selling first and then building the product is more the American way?
Shyam Sankar: Like, no offense, but that's that's how I as a European somehow see it. Like we are, we, we Europeans were so obsessed with, I don't know, especially Germans, like building, spending time in engineering, building something that might or might not be needed while, you know, Uh, Americans, uh, often, often have an attitude of, of selling quicker and, and spending time with a client.
Shyam Sankar: And I think like the, the, the, the real, uh, or reality or the, the, the, how it should be is somehow in the middle from my perspective. That's exactly it. It's dialectical, right? So, you know, the opposite of one mistake is another mistake. You need to somehow manage to hold the contradiction. And I think that's one of the deeper lessons.
Shyam Sankar: Like, you know, there, there are many valuable things about having a philosopher CEO, but, but this is one of them, understanding Hegel's important role in creating businesses. Um, I think that's, that's a more profound truth that engineers tend to struggle with in general, which is like all the value comes from the contradiction.
Shyam Sankar: and your attempt to eliminate the contradiction, like should we do A or B? These seems to be, these seems to be in opposition to each other. The answer is you have to do A and B. And what makes you a really special company is that you can find a way to do both. You can hold that contradiction. Okay, uh, so what makes you a special company?
Shyam Sankar: I think all the value we create comes from this managed misalignment between our forward deployed engineering teams and our product engineering teams. I kind of think about it as vector math. Two Types of Engineers -- Shyam Sankar: And it, it kind of comes from this observation that there really are, as a caricature, two types of engineers.
Shyam Sankar: You know, and this is a caricature, but I think it's illustrative. You have engineers who are artists. Their fundamental joy in life is what they created and how beautifully they created it, not its utility to the world. Like, the architecture is perfect. The fact that it happens to solve no problems is kind of beside the point.
Shyam Sankar: And on the other end, you have hackers who, the joy they derive is the fact that they solved the problem. Nevermind that it barely works and it'll probably not work the second time, but they actually did change an outcome in the world. And so this is, you can see how this inherent con, how do you, how do you tie these creative forces together so that you're solving problems in ways that are going to scale repeatedly?
Power Law: Solving th e Problems of the Most Futuristic Clients --
Shyam Sankar: I think the second part of it is we deeply believe that the sort of problems we want to go after in the world. Our power law, like, you know, the sorts of problems that you can solve by averaging all the requirements your customers have and synthesizing that into a roadmap. Those have mostly been done, and I'm not sure they're that valuable, versus if you thought about it as some portion of my customers have a problem that is living in the future.
Shyam Sankar: They are way ahead of the rest of the market. And if I can deeply and thoroughly solve that problem for them, I will have built the fundamental infrastructure required to solve it for the rest of the market when they catch up. And one of the most profound examples of that is the vaccine distribution infrastructure that we built during, during COVID.
Shyam Sankar: We first built that. for oil and gas, you know, it produces 3 billion of incremental hydrocarbons a year. It's a digital twin from the seabed all the way to the petrol pump. Had we not built that two to three years before COVID happened, we wouldn't have been able to deliver in the UK and the US vaccine supply chains in a matter of weeks.
Shyam Sankar: Sounds good. But, um, I think, I think many of my fellow CTOs here, um, still have challenges with understanding like which part of the infrastructure you're building. Um, and, and, uh, which, which piece of the puzzle you're solving. And, uh, this maybe is also a good. That like leads to our topic of like how you're going to deprecate backend in a few years. Shyam Sankar: So maybe you explain that as well. Building the Future of Backend Development --
Shyam Sankar: Well, I think the longer arc of Palantir is that we had to take responsibility for the full stack. And we kind of started at the top of the stack and worked our way down. So if you look at our reality today, we have roughly 150 independent dev teams building some collection of 5, 000 microservices.
Shyam Sankar: some combination of which are deployed to a thousand production environments, a hundred of which are air gapped. That means that we had to create, first, if you go to the very bottom of the stack at the production infrastructure layer, we had to create the necessary abstraction so that any one of these dev teams, any engineer in that team could be held operationally responsible for any of the thousand production environments.
Shyam Sankar: There's no SRE between you and prod. And so then I need to create an abstraction that allows you to not have to worry about the specifics of the environments at each one of these teams. um, places. Then, um, I need to automate the deployment. You're certainly not manually writing CICD pipelines for each of these thousand production environments for 5, 000 microservices.
Shyam Sankar: So at the bottom of the stack, we created a bunch of technology that we've now brought to market and made available in Rubik's and Apollo. In particular, we think of it as like autonomous deployment. You have to go beyond just CICD to autonomous delivery of your software. Then as we go above that stack, we have Foundry. Shyam Sankar: which allows you to, to kind of totally reinvent how you're thinking about how do I integrate data?
No Greenfrield Enterprise Tech: Need to make sure it works with old tech -- Shyam Sankar: How do I integrate my ability to write back and orchestrate with my existing transactional enterprise? Like I've never met really a greenfield enterprise customer. They all have extant investments that run their business today.
Shyam Sankar: And a successful strategy requires you to be able to quickly really connect to and manipulate and orchestrate them. And then above that, the kind of application layer, everything from low code tools that frontline factory workers can use in Hamburg as they're building the A320, all the way to pro code tools that React developers, data scientists who want to use Python, hardcore backend people who want to write in Java.
Shyam Sankar: So that means that you basically build some sort of an ORM initially and, uh, like, like move that to the browser and basically enable people to, like, create an ontology and, and, and logic on top of that, or? That's right. I think the beating heart of what we do is the ontology. And we built it in the reverse order.
Shyam Sankar: Even though I described it bottoms up, because it's maybe more rational to look at it that way, the journey of developing it was the other way around, which is working backwards from our customers problems to understand what infrastructure would we need to exist so that we could not only solve the day zero problem of delivering it, but the day two problems of maintaining it, evolving it, and transforming their enterprise as a consequence of it.
Companies as Decision Chains and Why Data is the new SNAKE oil -- Shyam Sankar: You know, there's been so much historical emphasis, I'd say, over the last decade on the modern data stack and like, you know, data is the new oil. I feel like data is the new snake oil. It's like, it's not really, the data is not inherently valuable unless you're using it to make a decision. The decision is the critical part of the equation.
Shyam Sankar: How am I bringing more data surface area so that we're making better decisions? And how do I get into a continuous learning loop around those decisions? And then if you take one step back, obviously no decision is happening in isolation. This decision has constrained your optionality for decisions that are downstream of it.
Shyam Sankar: And it was, it itself was constrained by decisions that are upstream of that. So you're living within a decision chain and at some sort of gross level, if you kind of squint at every company, that's what they are. They are a decision chain. Some people call it a value chain, but you know, so how do I help you get better and better at that?
Shyam Sankar: And to really do that, you need to recognize the impedance mismatch between how your technology is organized. All these transactional systems are kind of optimized for how they want to store and manage data, Ontology -- Shyam Sankar: but your enterprise, you have some sort of mental model as a human. that we call the ontology of how you, how you picture all this stuff happening.
Shyam Sankar: You have to allow for that fact, the fact that that model will change over time as your business evolves, your own conception of what your business is changes, but you want to be able to actually operate at this digital twin level, like how we as humans think of our business and what we're doing. and orchestrate and manipulate all your transactional systems as a consequence of that.
Shyam Sankar: So the ontology is like a logic layer. We've spent so much time building data warehouses, maybe we need to build a logic warehouse. How do I create an abstraction layer that makes my entire business programmable? Like, here is the API for my business. I don't have to worry about the fact that actually, you know, I can just say, Hey!
Shyam Sankar: Allocate inventory. That's a method call on an object for a product. And then underneath that, there are a series of SAP functions, maybe warehouse management system functions, other homegrown transactional systems that I need to call and orchestrate. I'd like to provide abstraction to the engineers who are driving my enterprise, to think in the higher order concepts, to encapsulate security.
Shyam Sankar: You know, like, how do I make it as easy as possible for them to move as fast as possible safely in a way that I can govern at the ontology layer? So that sounds like a, like an ORM with finite state machines for each of the objects you store with connections to every possible system out there. Like, is it, is it that?
Shyam Sankar: I think that's a good first approximation. Of course, you know, as anyone who creates something, you'd always bristle about, uh, You know, small differences here or there where we think we've kind of thought about it differently. But I think to a first approximation that that's a good way of thinking about it.
Shyam Sankar: But the there's a heavy emphasis on on the write. So I think with a lot of these systems, you get a lot of emphasis on the read. Like, how do I integrate this data so I can see myself? Great. The best you're going to be able to do with that is a dashboard. So if you have an insight you can't take an action on, what's the point?
Shyam Sankar: It's entirely academic. Instead, how do I build software that's actually going to be used at the factory floor, where people are making decisions? You know, and the fundamental pathology for most institutions is that there's a disconnect between what is happening in the boardroom and what is happening on the factory floor.
Shyam Sankar: And you try to solve that with layers of bureaucracy and middle management, and, you know, you're trying to, like, reduce complexity by dividing up the problem. But you, it's highly attenuated as a, as a consequence. You like, you, you lose a lot of the signal of what you're trying to do. So in the boardroom, they think they have a steering wheel for the enterprise.
Shyam Sankar: They're turning it and they're somehow surprised that actually, you know, the ship is not moving at all. Um, and that, that is a technology problem. /One way of thinking about this is empirically like COVID in the lead, let's say in the five years leading up to COVID. How much did the world's enterprise companies spend on their IT systems?
Shyam Sankar: I mean, many, many billions of dollars. When COVID happened and their supply chains were disrupted, What could people point to and say, Oh, I'm so glad I invested in this tech stack. It enabled us to weather this storm. I mean, I think by and large, the only thing people could point to was Zoom. How depressing, what an indictment of the software industry that we were able, we could not deliver resilient enough software for them to, to, to pivot to entirely new problems that were historically unforeseen.
Shyam Sankar: And I think that, that, that, That's what happens when you focus on the playbook of what is sellable, instead of thinking about fundamentally what is valuable. Palantir's Vision for the Future -- Shyam Sankar: So, what happened at Palantir when Our business went through the roof. Apart from you, like I mean, I think, to use the Buffett, uh, analogy, when the tide went out, you could see who is naked.
Shyam Sankar: Uh, and I think a lot of companies kind of scratched their head and said, like, why is none of this stuff working? Like, I have an exquisite ERP system. I'm so proud of how it's implemented, but somehow I have to shut down my factory because, you know, a 5 wiring harness is no longer coming across the border from Ukraine or whatever the case might be.
Shyam Sankar: And I think it really, it really poked holes at, uh, The seams between all of these systems and the inability to, to, to pivot your enterprise to new challenges and problems. This speaks to this divide. Inductive over deductive problem solving -- Shyam Sankar: The other, the other thing I think that we get pretty right at Palantir is this, you know, most of the world is designed to solve problems deductively.
Shyam Sankar: Any problem you can solve by thinking top down, someone has a little software point solution to go after. But what about all of the inductive problems? What about the things that you, uh, are either not knowable ex ante or can only be solved through iteration. You know, that's what I think when we, when we talk about backwards, the same way that hyperscalers provided us This abstraction to, to deprecate infrastructure, really, you know, it's actually kind of hard to find, uh, young software engineers who know how to manage software on prem, for example, they've forgotten all those skills.
Shyam Sankar: The abstraction is so beautiful, so clean. Deprecating Backend Development -- Shyam Sankar: Well, I think we've built the necessary building blocks to deprecate back end development. And we've done that. Like, the right way to think about the infrastructure we've built is that these are highly opinionated building blocks in the same way that a hyperscaler gives you highly unopinionated building blocks, which are great.
Shyam Sankar: You can build anything you want with them. It just might take you an extraordinary amount of time and effort. And the primitives are kind of at the infrastructure level that you have to reason with. We've given you highly opinionated building blocks, like a schedule is a first class primitive that you can use and manipulate across all the applications, all the services in your ecosystem.
Shyam Sankar: Security is consistent throughout everything you're doing. So it liberates the developers, but just you don't have to think about the bare metal. Well, you don't have to think about your infrastructure. You don't think about autoscaling. You don't have to think about security. You can instead focus on a declarative approach to back end.
Shyam Sankar: Let me just declare, here is the model that I essentially need. So not just the data model, but also the logic model, like how I'm going to do this. I can bring, you know, to use the MVC sort of framework. I can bring my own controllers. I can bring my containers to this orchestration engine. I can write my own logic in the platform in a serverless sort of way.
Shyam Sankar: And then I can very easily drive the view part of this either through low code applications that are seamlessly integrated through ontology to everything I want to do in pro code. So if I want to build an entirely custom set of applications on top of this, you wouldn't even know that Palantir is backing it.
Shyam Sankar: In the same way, you wouldn't know that AWS is powering half the internet. We're providing that scaffolding. So that means like in a few years from now, or maybe potentially now, uh, I don't have to like boot up Ruby on Rails whenever I start with a new idea, but rather call you or just use your interface because you gave me access.
Shyam Sankar: Exactly, that you would just go to build. pounder. com and get your infrastructure to get going. Um, and I, I think we're, we're well on the train. I, we've done a lot of this with enterprises. We're now very focused on bringing this to the broader market so that any company could get started on, on day zero building on top of this. Low Code Gets Complex? --
Shyam Sankar: And how manageable is it? Like, I mean, often no code solutions lead to, let's say, less complexity at the beginning and more complexity later on. Like, I, for example, love Airtable, uh, at the beginning and I hate it at the end. Uh, like, like, how is that? I want it to be pro code. So we absolutely provide the low code tools, because I think if you're being realistic, the factory floor worker is going to be developing solutions in low code, but we want them to be seamlessly interoperable, whether you're building on low code or pro code, the applications you're building are going to work together because they're working on top of the common ontology.
Shyam Sankar: But we envision the developer platform being pro code. You're going to be delivering, you're going to be building your React applications or your Python notebooks, really leveraging this backend. Will we need IDEs int the future? -- Shyam Sankar: And does that only deprecate backend development or also, uh, deprecate local development on your machine?
Shyam Sankar: Your IDE setup is like an extension of your mind. It is a bicycle for your mind. So I think a lot of development is still going to happen locally. And that's, that's part of the art is, is making sure that that dev experience is magical, that DevX really works. I think for some of these environments, it's actually much easier.
Shyam Sankar: If we give you a hosted dev environment, which we do. So it's like in the context here, you can go, you can develop, it's connected seamlessly to the APIs. It has the environment, the kind of test infrastructure loop you need to do this. So that's there, but we would never want to deprive you of a local dev environment.
Shyam Sankar: So that, that means you can still use your own version control, et cetera. Um, or is it all happening on Palantir servers? No, you can use your own, but certainly we're just using Git and we're providing, we're making that easy for you. So if you don't want to go futz with it and set it up, yeah, sure. Okay. Um, and this is basically like accessible for startups, like digital native startups out there.
Shyam Sankar: Like you mentioned YC and that you're collaborating there so everyone can just use it. It's not like purely for governmental organizations or, uh, like big enterprises. Yeah, that's right. It's out there. And there's. probably on the order of 100 startups today that are that are using it, they're building, this is like really the core piece of technology infrastructure they're building around.
Shyam Sankar: And part of what I'm excited about is we're kind of making this more generally available and launching it so you can just come there, sign up and get going. Tacking the Application with Clean Abstractions --
Shyam Sankar: So are you then also deprecating AWS long term or? In a sense, yes, like, we're not competing at the infrastructure level, we are running this on top of hyperscaler infrastructures at the application layer, like, how are you building your backend, what are the backend services you're going to need, how much, how much burden is there on you to operate these things?
Shyam Sankar: versus actually you have a clean abstraction that allows you to deliver the applications you want. So I don't have to build scaffolding anymore. I don't have to build like CRUD models, etc. Um, I don't have to build the API itself. Uh, I don't have to build, like, I don't know, the connection to LLMs. Like, how does that look like? Shyam Sankar: That's exactly right. Leveraging LLMs for Enterprise Use Cases --
Shyam Sankar: Yeah, so we bring all the LLMs to you. You can use them self hosted. You can hit the common commercial ones in a very elegant, integrated interface. We're going to keep building in that direction. We think with LLMs, the biggest problem is the bottleneck from prototype to production.
Shyam Sankar: And there's a whole tool chain that we're excited about releasing here that helps you manage that transition. So you can go from a very compelling, charismatic demo to something you can safely put into production to drive enterprise use cases. -- Shyam Sankar: And that gives you some, some features around how do I auto tune?
Shyam Sankar: How do I figure out what LLM is actually right for me on the cost performance curve? The difference between the most expensive model, which today is the highest end anthropic model and, uh, the cheapest, but still most powerful open source model is about a thousand X. So there's going to be a lot of room for optimization and thinking about, well, yes, the cheapest way to prototype these things is with GPT 4.
Shyam Sankar: 0. The cheapest way to deliver this in production is probably going to be very different. How to avoid being a "Jack of all trades but master of none" -- Shyam Sankar: And like building all this, like how do you prevent of being jack of all trades, but master of none? By maintaining a deep commitment to, uh, delivering the software for outcomes ourself, right? So I think everything we built, we didn't build because we thought it was a good idea.
Shyam Sankar: We built because we had to build it to solve a huge problem out there in the world. Uh, and so that I think there's this empirical validation. We had to use this stuff in anger to deliver value for outcomes that we all depend on every day collectively. So from that, how do you synthesize the learning?
Shyam Sankar: That is the, essential inductive journey of what must exist that doesn't exist today. That's why we have so much conviction in what we have. Okay. Yeah. I mean, it really sounds cool. And, um, to be honest, like you, you showed me a demo like for one hour and I was like deeply impressed. Uh, so I think it really makes sense for, for folks here to my listeners to check it out at, at build.
Shyam Sankar: palantir. com. Uh, before we come to the, like, maybe your ideas on leadership, et cetera, Is there like, are there any restrictions, any limitations? Like, can everyone just get started or how does that work? Uh, for complicated reasons, this, the complete self service signup is available in the US, Canada, India, and the UK.
Shyam Sankar: It's coming to more countries quickly. Uh, if you're not in one of those countries and you'd like an account, you can, uh, get in touch with us and you'll get a free developer account through other means, um, but the self service bit is right now in those core markets. So you still have to build something for, for our market here, for Germany and for Europe. Focus on Strenghts --
Shyam Sankar: Obviously, um, coming from, from that, maybe to your ideas on, on leadership and how you actually, um, how you actually, What you learned on your way, um, I think one of your core principles is that you believe in meritocracy. Is that correct? Oh, highly. I mean, maybe even backing up, like the core talent philosophy at Palantir is that we want to be an artist colony, not a factory.
Shyam Sankar: I have no career ladder to offer anyone. I can give you no predictability in what the future is. But what I can tell you is that we are going to treat you as an individual, a unique human who has your own set of strengths and weaknesses. I'm never going to ask you to somehow be well rounded or get better at your weaknesses.
Shyam Sankar: That's a waste of time. If you have superpowers, like Superman, you know, there's nothing you can do to to somehow, uh, not get hurt by kryptonite, other than to avoid it. So I want to create an environment that allows you to be honest about what you're not good at, not have to engage with those things, and instead to focus on your true superpowers, which is how I think you self actualize. Shyam Sankar: It's how you create the most impact in the world. --
Shyam Sankar: And then, continuing that sort of comic book metaphor, we get things done in teams. Like, how do I get the right sort of chemistry in the rock band so that the music, the jazz band, sounds amazing? All right, so how do I, by enabling you to be like, look, this is really what I'm good at.
Shyam Sankar: This is my superpower. That's a journey of self discovery for lots of reasons. Our classical education system pushes us to be well rounded. It pushes us to try to do all these things. It also makes us think that The thing that we're good at must be hard for us to do. Otherwise it doesn't seem that like intellectually rewarding, but actually often the things that we're world class at are intuitive, they're native to us.
Shyam Sankar: They happen at some deeper level in our, in our brain. And I think there's a journey for, especially for early talent, you know, to discover what, what really is that. And part of that is observing it observationally. Because you look at it and you see, wow, this thing was really easy for me, but everyone else is kind of struggling with this.
Shyam Sankar: And that's indicative of an area that you have unique strength. So let's start there and say, how do I create an artist colony that rewards that sort of behavior that allows people to not, I'm not trying to fit you into a role. I'm actually developing a role around you. And then how do I wield that?
Shyam Sankar: throughout the organization to go then deliver maximum impact out there in the world. So when you start there, you have to be meritocratic, right? Like the whole point is you're developing people based on who they are. Uh, and then you, you, I mean, a lot of things come as a consequence of that. Like how would you organize yourself?
Shyam Sankar: Well, you, there's no consistency actually, because almost everything is itself bespoke. You want to build the right team. You want to organize around the problems that you're actually facing. Flat Hierarchies and Organizational Structure -- Shyam Sankar: And I think there's a way in which The right org structure is quantum. Is it a wave? Is it a particle? I don't know.
Shyam Sankar: It depends on your moment of observation. Is it flat? Is it hierarchical? You know, it's like, to the point of the opposite of one mistake is another mistake. I think a perfectly flat org structure seems like complete chaos. A perfectly hierarchical structure is locked in and sclerotic and won't adapt to meet its moment.
Shyam Sankar: The question is how malleable is it to fit the moment? And your ability to reorganize around a problem is indicative of how flat it would be. And, you know, is your institution the sort of place that could tolerate having the person who's been here three months being in charge of someone who's been here three years? Reward --
Shyam Sankar: You know, I think that's one of the central strengths. And how do you ensure that you reward the right people, the right talents, and the right behavior? And how do you actually measure the right behavior? I mean, it's really hard. I mean, I think a lot of it is qualitative. So on the reward side, you know, the fundamental reward that you're signing up for is Getting to work on bigger and bigger problems, having more and more impact, uh, and, you know, there's a human piece of this where humans kind of, they're thinking, how do I communicate my progress to people outside the company, to my family, to myself?
Shyam Sankar: Uh, and they're wrestling with that so that you kind of come to terms with like, my progress is really about how good do I feel about my work? And when I look at the things in the world that I'm powering, you know, how much joy do I get out of that? How rewarding is that? Shyam Sankar: In terms of measuring it, I think it is highly qualitative. Diversity, Meritocracy and Rebellion --
Shyam Sankar: And, and, and how can you still make sure that you getting the right. amount of like diversity and especially like diversity and thoughts. Let's, let's, let's look at that. Like, how do I, how do I make sure you, you, you have that set in, uh, in, in, in this culture? You need to have enough entropy. Encouraging Rebellion and Creativity -- Shyam Sankar: So, uh, you know, I think one of the, the good thing when, when, when you bias towards flatness, You can't overmanage things.
Shyam Sankar: Actually, lots of things are happening. There's lots of people with heterodox ideas. So one of the things we encourage is, is rebellion. You know, so yes, I have opinions of where I want the product to go, but I also empower you to tell me to fuck off. It's actually the first thing I tell at every AMA with new hires.
Shyam Sankar: I make them practice. telling me to my face to fuck off out loud. Because if you're not willing to do that, it's not actually a flat company. And so I want you to prove me wrong where I'm wrong. Like, of course I have strong convictions, but I'm not going to stop you from going out there and, and, and developing heterodox ideas.
Shyam Sankar: And that's how you ensure that there's enough space and enough room and creativity. I mean, I think a lot of creativity comes from that sort of process of saying, like, I know this is wrong. I want to go prove it. I want to go do this. And having a culture that encourages that rather than tries to sequester that or tries to stamp out, you know, we're fundamentally looking for the heretics, looking for the heterodox people. Hierachy at Palantir --
Shyam Sankar: And how does that look like? On a hierarchy level, like which, which kind of roles do you have? And where do people report? And how flat is your hierarchy? It's, it's pretty flat. I mean, I think to most people, our org structure makes no sense. At some point, we had root nodes. I really like that concept. There were people who like, didn't report to, you couldn't, you know, there was not one node at the top of the tree.
Shyam Sankar: There's just multiple nodes. And it made no sense. And I think part of that is getting people to not look at that too seriously. How can you maximally, there's some things that you're going to need a little bit of hierarchy to, to make easier and, and, you know, whether it's like a review process or who do I go when I have something confusing.
Shyam Sankar: But if you're really looking at the creative process, hierarchy just gets in the way. So when you're doing the creative work, that process, you, you want it to be, you want it to feel completely flat. And I think you, you kind of get over the entropy. There's a human bias towards hierarchy and structure and organization.
Shyam Sankar: And that's why you have to also find some way of institutionalizing rebellion. You know, it's the, in, in the Hindu tradition, There, there, you know, there's three roles of the God. They call it the Trimorthy. There's creation, there's maintenance, let's call it O& M, and there's destruction. And the only way to deal with all these, this entropy you accumulate over time is to destroy it.
Shyam Sankar: You know, you need this, this process of tearing down the things you built in the past to rebuild them. You know, it's a loop. Uh, and I think there's a lot of wisdom. I'm not a religious person, but there's a lot of wisdom in that. Yeah, that really is a cool idea. So I have to have to admit, but on the like, I find it hard to really like see how it's practically working and how you ensure that there is flow. Marker Flow and hierachy --
Shyam Sankar: Like, uh, I really believe in like, uh, natural flow and hierarchies often standing in the way of, of, of flow. How do you ensure that you have a, like a constant flow and, and, uh, like not very hierarchical, um, organization where you even encourage people to stand up or to rebel? Well, the first step is, uh, driving transparency and communication. Reduce Informatiojn Assymtry --
Shyam Sankar: You know, so it's almost like a starting condition is you, you really want aspire for organization where at wherever you are, you can know what's going on. It's knowable. Yeah. Now there's a, there's a second question of like filter failure where that can be overwhelming, but you don't have to fight for access to what's happening here or there.
Shyam Sankar: Or how do I have as much information as possible so that I'm on a level playing field when I'm making decisions, I'm fully informed. That's step one. And there's lots of ways you can do it. Like you can think about, um, broadcast comms or. Probably one of our core secrets, you know, if something's happening, there's kind of an expectation that you're going to write it up so that everyone who wants to have access to it can say, I now know what's going on here.
Shyam Sankar: So that reduces the information asymmetry, which becomes one of the unofficial ways that hierarchy gets propagated. Like forget about what's in the org structure, you know, it's when the org structure starts mapping to the information flow structure. that you start having lots of pathologies. If you just said, look, you know, as a thought experiment, everyone's going to know everything at all times.
Shyam Sankar: I think you would find the org structure matters substantially less than, than anything, right? Okay. So now if you, if you start addressing problems of information flow, then it's like, how do you organize the work units? And so how do you get the people who can solve the problems next to the problem and then get out of the way?
Overmanaging and Decay of -- Shyam Sankar: That, that's, you know, there's some, I think a great part of the decay of, of like our ability to do things in the West is a function of overmanaging things, you know, how is it that we started the Inflation Reduction Act and we somehow have eight chargers after spending like 8 billion, you know, it's just all overmanaged or, you know, Elon put 300 plus rockets into orbit for 9.
Shyam Sankar: 6 billion, 9. 6 billion. And California spent 11 billion to build 1, 600 feet, 1, 600 feet of elevated, right? It's just where, where, where is the effort going? Are you getting the engineers next to the problem, then getting out of the way? And that can be very scary because it's, you know, it's not managed. It's not controlled.
Shyam Sankar: There's going to be variance in the outcome. Some of these things will be a disaster. Some of them are going to be truly spectacular. But if you, if you constrain the variance, that means like, sure, you probably have less disasters. You also don't have anything exceptional that comes out of that process.
Shyam Sankar: So how do you kind of lean into that authentic, um, reality? And then you have to go get the best people and you have to empower them. And you have to kind of, the metaphor I use is, you know, you don't turn into the incredible Hulk by lifting a little bit more weight every single day. Right. There have to be these crucibles that you go through, this blast of gamma radiation that might've killed you, you know, maybe it was a 50 percent chance you're going to survive that experience, but if you do survive it.
Shyam Sankar: You're going to have learned an enormous amount, like, and there's no, you can't make up for that. So how do I continually maximize putting my people in a position that, that grows them? And I remind them that your position of maximum growth is also likely a position of maximum pain. Like, no one is telling you this is going to be a pleasant journey, but it is going to be rewarding.
Shyam Sankar: Well, and if you, like, have the right intuition, then it can be also, like, rewarding and not so painful, right? As said. Yeah. Yeah. Look, there's Greg LeMond, the famous cyclist. He said, You know, it doesn't get easier, you just go faster. So, I think one of the pathologies you kind of see over time is like, we continuously, we look at the experience we went through, we reflect on it, and we say, how can we make that less painful?
Shyam Sankar: And there's a very fine line between what's a medicine and what's an opioid in that reflection. What actually, by blunting the pain, ensures you will not get to the outcome? Like, what part of that pain was necessary to the journey, and what part of that pain was superfluous? Certainly, there's often superfluous paint.
Shyam Sankar: I'm just saying it requires, you know, a scalpel to really think diligently about, um, where am I blunting the outcome versus actually creating efficiency. Great. Like, I, I like your thoughts. I, I, I'm not sure if it can be applied to every organization out there. Um, and it, it seriously, means like rethinking how you structure.
Shyam Sankar: Um, when, when did you start with that? Like, is that like from the start? Did you set the values and then, uh, like you, you just went for it? Or was there like a fundamental shift when you, when you thought, Oh, we have to change this. Like, this is too bureaucratic. This is too hierarchical or like, what, what was your, your moment of, of shifting?
Shyam Sankar: I mean, I think we were blessed with starting with kind of this kind of contrarian culture and. Alex really enforced it. I would say to the point of entropy that like there is this human desire for predictability and structure. So at various times, I feel like we've accumulated Um, structure we've had to tear down.
Shyam Sankar: And I've never met a structure we tore down that I didn't really appreciate tearing. You know, when you're looking at the abyss and you're like, but here are all the things we're getting out of this structure. It seems so comforting. It seems like there's so much loss aversion. But actually then in tearing it down and reinventing it, it only becomes strictly positive. Learing form big leaders --
Shyam Sankar: As initially said, you're working with many big leaders from tech and, uh, like people like, I don't know, Satya Nadella, Peter Thiel, Elon Musk, et cetera, Alex Karp, obviously. Who was the most influential person of those, uh, or like even more, I didn't mention? And, and, and why did they influence you so much?
Shyam Sankar: Uh, wow, that's, that's a tough question. I feel like I've learned so much from, from, from all of these, these folks. And I'm constantly trying to absorb their mental models and see how you would apply them. But if you're, if you're looking narrowly in the context of Palantir, it'd be hard to escape the enormous impact that both Alex and Peter have had in terms of how we all think about everything that we're doing every day.
Shyam Sankar: And it's motivationally too, it's like reminding us to stay true to, painful journey of discovering secrets that are out there, of not absorbing the comfortable playbook that everyone else is doing, of looking completely crazy, um, managing to be misunderstood. and, and having confidence in what it is that you're actually doing.
Shyam Sankar: But there've been many people along that journey where, um, you absorb deep lessons from before my time at Palantir, people like Keith Raboy and, and Roelof, both, uh, um, when I was at Zoom all the way to the present moment, looking at, uh, the customers we're actually working with and how they think about problems.
Shyam Sankar: Um, it's actually really, you know, we spend all of our time, the work, the creative work we do is around bits. You know, but most of the 50 percent of our business that's commercial is about helping our customers bend metal. It's about atoms and recognizing that the very physical world that I often I think inside of technology companies or particularly inside of the Valley, you kind of forget about the rest of the world that really exists and how crucial it is in this moment.
Shyam Sankar: And that's led us to invest a great deal in reindustrialization broadly, a team, a theme. Arguably, German doesn't have to reindustrialize, but it's an opportunity to keep going with industrial and its industrial might.
Reindustrialization and Software-Defined Production -- Shyam Sankar: Uh, but the West broadly does need to reindustrialize and, um, it's unlikely that we're going to, we're going to drive that industrialization the way we did in the forties, fifties, and sixties, you know, we have unique strengths in software and.
Shyam Sankar: We're going to have to think about how do we do software defined production. Uh, and we have unique strengths with AI that we are the leading labs in the world. We are doing the interesting work. How do we leverage that to drive the physical economy? I think is one of the most interesting challenges right now.
Shyam Sankar: One of the bigger things that still needs to be solved, right? Like, how do we move AI in the middle of everything? And actually, how can we trust it enough to derive actions from it? Right? Personal Productivity and Final Thoughts -- Shyam Sankar: What was your last, let's say, game changer move in personal productivity? Like, is there something you can recommend to all of my listeners?
Shyam Sankar: Like something that made you way more productive in the last months or years? Yeah, actually, in the last year, I recommitted myself to having a small number of things I drive as an individual contributor. Uh, and you can't do that for everything, obviously, as time goes on. But there's an immense amount of value in being so deep in the details on a specific set of projects.
Shyam Sankar: that you actually have to be part of the creation process and the kind of inductive dots that gives you on how you can apply that against the broader portfolio. So I think it's a real mistake to get to the point where all you're doing is managing a bunch of things, managing priorities, thinking, you know, pretty abstractly in the high level.
Shyam Sankar: Uh, and, and so finding some way being connected to at least one project that you are deeply in the weeds on, I think makes you better at everything It sounds counterintuitive, and I think people resist it because of context switching. Like, the context switching is very hard at different levels of abstraction, but it's worth it because of how much more productive you are.
Shyam Sankar: Shyam, thanks a lot. Um, coming to my outro question, so I still have a little surprise for you. Um, Alex Karp told me about a secret Easter egg in the core of Palantir. Uh, in your, in the, in the deepest core of Palantir and the, the, the secret that he, uh, doesn't tell many people, it is actually a functionality that's, uh, called Time Machine and, uh, it actually helps you to travel in time and travel back in time or also forward in time.
Shyam Sankar: Um, and, uh, I actually applied it, uh, on the instance that, that, uh, you gave me, uh, in AIP and build. palantir. com. Um, and. With my newly built model, we can travel back in time in your life, uh, to the time you worked as a developer at Zero Chaos in the year 2000. And you now have the chance to whisper something into young Shyam's ears.
Shyam Sankar: What would it be? Focus on my your strengths, go figure out what your superpowers are. Um, don't worry about the rest of the world, right? It's, it's really that some, all the value comes from the internal creative process and knowing who you are, not on being what you think the world needs you to be. That's my simple advice.
Shyam Sankar: That's a good advice. Thanks a lot, Shyam, for being my guest here. Um, and yeah, to all the people listening, check out build. palantir. com. Um, and, uh, I can even connect you to, to the folks at Palantir. Don't hesitate to, to reach out, uh, so that you get a proper EU instance, right? That's right. Uh, so yeah.
Shyam Sankar: Thanks a lot for being my guest. And I hope you I can I can actually meet you in real life because the Wi Fi was so shitty in today's recording. Excuse for that. So whatever I travel over, I'll ping you. Yeah, my apologies on that. Well, yeah, we should do some sort of physical developer event in Germany.
Shyam Sankar: Yeah, yeah, let's do that. Let's do that. Thanks a lot, Sham. And let's meet up. All right. Thank you, Tobias. Bye. Thank you for listening to the Alphalist podcast. If you liked this episode, share it with friends. I'm sure they love it too. Make sure to subscribe so you can hear deep insights into technical leadership and technology trends as they become available.
Shyam Sankar: Also, please tell us if there is a topic you would like to hear more about or a technical leader whose brain you would like us to pick. Alphalist is all about helping CTOs getting access to the insights. They need to make the best decisions for their company. Please send us suggestions to cto@alphalist.com.
Shyam Sankar: Send me a message on LinkedIn or Twitter. After all, the more knowledge we bring to CTOs, the more growth we see in tech, or as we say on Alpha List, accumulated knowledge to accelerate growth. See you in the next episode.