Palantir's chief architect Akshay Krishnaswamy on forward-deployed engineering and AI in enterprise workflows

won't get too into uh geopolitics, but we are joined by uh someone from Palanteer. Um so we have Ash here. Uh please uh welcome to the stream. Thanks so much for joining. Uh how you doing today? I'm I'm doing great. Hopefully you guys can hear me. Okay. Yeah, you sound great, look great, everything's fantastic.

With the fake plant in the background, of course. Um I No, it's good to be with you guys. My name is Ashe. I'm the chief architect of Palunteer. I've been here for about 12 and a half years. Lived a lot of lives.

uh things are still moving super fast, but I'm excited kind of given where you guys are taking the discussion and everything you guys have been talking about throughout the the day and morning as well. So yeah. Yeah.

I mean I I I don't want to go too into the whole thesis of Palunteer in the West, but uh maybe we could start with like uh what you're working on today and and and and uh really uh what does Palanteer actually do? I'm getting I'm getting it is how do you how do you describe it?

And I think it's it's funny because it's been this in some ways it's been so consistent in intention but I think the way the technology has evolved has been because like we had to kind of keep pushing the frontier of what we what we were limited by what we built originally and what we had to kind of fulfill and longwinded a way of saying like it was always about how do we help people in operational context make better decisions.

So it was very simple early on of like these are people who are working in counterterrorism and specific defense workflows and it's like they need to be able to do their jobs better to avoid roadside bombs to be able to uncover intelligence networks like these are the actions the decisions they're taking and like you know the the cheeky kind of naive first version of that was like can't we build like a fine terrorist button or do some cool like algorithmic you know kind of analysis to do this and it's like well you first need to build like the the dirty dirty data integration plumbing to be able to then enable people do the interesting stuff, right?

So that's the whole data integration, bringing together disparate systems piece of Palunteer. And then it's like the requirements keep changing on you, right? Because it's like, okay, that works for the first intelligence workflows. What about the next set?

What about then going into war fighting, which is much more kinetic, much more real time. Um, and so you have to keep building new parts of the stack to be able to enable different types of decision-m. And then there's kind of the whole commercial journey, right?

Which is like, okay, if we think about all these pieces of technology being able to reconcile complex systems, enable decision makers. It's like at BP those were now petroleum engineers or at Airbus those were like people working on the A350 ramp up.

And it's like kind of a different problem set, different nouns and verbs, but kind of the same kind of udaloop of like how do we get people to be able to make decisions better in these contexts.

And so then you see kind of more layers of the platform and the architecture evolving to kind of meet all of these different needs. And it's like I joke it's kind of like building something through back propagation.

It's like you keep doing this forward deployed approach seeing what doesn't work seeing where what you built isn't sufficient then having to understand kind of from the coal phase what to build.

Can you talk about the the the potential trap because you you guys uh like made the sort of forward deployed model uh and you know I think startups now uh can kind of like look at your you guys and your success and be like okay we're going to do the same thing but I don't necessarily think that's always going to end up in the same place.

I'm curious how you thought about kind of, you know, as you were rolling it out, the sort of traps, what you wanted to avoid and how to make sure that um, you know, you were actually building a, you know, real software business behind the scenes. Totally.

And, you know, I could make it sound like it was all, you know, predestined and and perfectly planned out, which it certainly was not.

Um, but I think like the funny thing to me also is that it's become such a meme because it's like it was the ugliest of ducklings back in like the mid 2000 mid2010s where it was like you guys are working for the government deploying and doing what and it's like you know it just was the complete antipode of like the consumer B2B SAS kind of or B2B SAS kind of thing that people expected to see and I think now it's become very invogue which again is just very bizarre I think but but to me it's like it is um like I think you're you're intimating here it's like it's easy to say you're doing it and just put a label on what somebody might call like it's sparkling sales engineering.

It's not forward deployed engineering I think was one meme, right? And it's like to me it's like it kind of comes from how the organization is run. So like the first forward deployed engineer was Sham Sankar, right? Who's our CTO and like basically the organization was built around like the field has this primacy.

It's the primacy of winning and showing up for the outcome and product has you know it was Bob Mcgru and folks back in the day and a lot of talented folks on both sides now. But there's always this dialectic between like field and core, right?

And so it's like as the field is uncovering what needs to be built like how is the core responding and how much circulation is there between the two. I think it's very different than the normal construct in Silicon Valley which if you kind of go up and down the the peninsula here.

Um it's like we have the kind of high priests or the privileged engineers that sit, you know, eating berries in in in PaloAlto or or in New York and then if you're like a sales engineer or like a solutions architect, you go into the field or we delegate that to somebody who's like a consultant, right?

And um I think that is kind of the inverse model, right? Where it's like if that if you're just relabeling that team as being a forward deployed engineering team, that's not kind of the same thing as saying the way you do product development is kind of front front to back.

Um, and uh, like you have to orient kind of your entire organization around that. And I think to your point, it's like you have to think that there are complexities or worthwhile things to uncover through that method to make it worth it, right?

It's not something that's going to make sense for maybe, you know, simpler product schemas. Yeah. C can you talk a little bit more about the forward deployed engineer?

uh it it uh in a lot of B2B SA if you think about like the implementation of like an ERP or even like an email client or email SAS product like Mailchimp uh there are agencies that pop up and effectively act as forward deployed engineers and maybe they're like partners gold tier silver tier whatever uh I can imagine that that doesn't make sense in a security environment but uh was that ever evaluated on the commercial side or is that just completely off the table because notable kind of like have those like automations or sorry uh specifically like specifically like independent agencies or companies that are building an entire business just around Palunteer uh implementation.

Yeah, it's interesting because like that's now coming to life through alums that we have who've actually gone on to kind of make these kind of like more Spartan I'd say smaller squads that are kind of dedicated consultancies now.

I think the joke back in the day was like nobody who was a consultant or was external would like bother to do this because like we're trying to build this plane in midair. are trying to like understand like your job as an FTE is not just to to service the demand or to implement the use case.

It's also to figure out what to build, right? And like you are given kind of that creative power to say like you now have to spend your your nights, your evenings, you know, the the off time you don't have talking with product about what to build.

Um and then showing up the next day or two days later with a new version and like it's much more collaborative with a core engineering org.

And I think that's where like if we tried to outsource that I think there was always fantasies in the painful moments about like we just get other people to do some of this stuff but it's like the real magic is saying like you are all part of that shared engineering core and it's the same by the way like kind of hiring profile of somebody who's kind of a core dev or somebody who's an FTE and all the people who are like core product leaders now were FTEEs before and so I think that that kind of results in a different kind of profile between the two.

Yeah. Can you talk a little bit about um where uh just some case studies on the commercial side? I remember hearing about the Airbus case study around manufacturing logistics. It it whenever when somebody asks, "Oh, what does Palanteer actually do? " It's very easy to say like you have a bunch of parts.

You need to know where they are and how much they cost and whether or not they're in stock and all this stuff. Um but are there other How are you explaining it to like a a non-erospace Fortune 500 company today? And I think that ve that that very much is the kind of progression over the past couple of years.

I think as the business has really grown. Um there's of course like it's called like the industrial like hardcore I love this stuff right it's like it's the new warp speed cohort. It's like it's seronic. It's Anderil it's Eperis and they're doing like like it's like nuts and bolt stuff.

But I think what's more subtle and I think what's kind of unifying among all these things is like what are we actually doing with the software now? Especially with the AI platform with the Gen AI piece. It's like you're you're modeling the world, right? Of course the other Palunteer meme is ontology.

They just say ontology to everything, right? It's like what do we mean by that?

It's like it's it's a shared model of decision-m and it's like you have to model of course all the objects and links like it's the it's the underwriters, it's the claims, it's the movements across those things, but then it's also all the actions you can take and it's all the compliance and security requirements around who can take what actions and then it's like the reasoning or the business rules that govern how all of these like states can change, right?

So, it's kind of almost like a digital twin of not just the asset but the business process. And so in that sense, like you're seeing a lot of really cool workflows now in hospitals.

So like Cleveland Clinic, HCA, Tampa General, where it's like they're talking about nurse scheduling and and what's cool is like you have this kind of teaming between human users like the nurse schedulers and then AI able to actually be useful in those contexts because you can kind of measure when and how you're injecting the AI.

Um you can put guard rails around what it can do and what it can't do and it's kind of right there with you in the same context. So it's not like this weird black box off to the side where you're like I hope the AI does something useful and then if it doesn't like I'm just going to pivot back to doing my other thing.

It's this kind of like incremental rollout within these existing workflows staff scheduling uh you know patient life cycle things like that. One that's really cool is um AIG recently talked about underwriting and like how they're driving automation through underwriting you know with all their core benches and um Dr.

Karp and Daario from Anthropic were on stage with with the CEO and they were talking about kind of how this incrementally happened by you know first modeling the underwriting process in the software and then gradually driving more automation through it.

So I think domains like that are showing a lot of really interesting and like kind of colorful usage outside of what you might typically associate with Palunteer.

Yeah, it must be so funny to be uh using Palunteer in the global war on terror and then as a as like a field medic and then you get your job as a you know someone in a hospital and Palanteer shows up again and you're like yeah this software defines everything.

Um h how uh do you think that just agents in a B2B context are are sort of um uh overhyped underhyped? H has your thinking around them changed at all?

I mean, when you guys started the company, agents were just called bots, and now now they're they've got a shiny new name, and it's very exciting, but I'm curious how your your thinking around sort of the highle concept of agents has changed from, you know, basically over the last decade.

No, it's it's a really good point and I think it is interesting to see kind of the nomenclature shift on you over time and and I think our perspective is there is something you know unique and significant about the rise of generative AI like all the transformer-based architectures what you're seeing especially with the latest generation of frontier models as being essentially like a a reasoning engine that can be used in concert with decision-m and so like from our perspective it's like what these agents are showing is the ability to actually creatively be able to deal with that decision graph with that model of decision ision making and then actually provide real horsepower that first augments human users then can kind of slide up to automation.

But I think to your point it's like how is this not just hype from our perspective is like is there an actual story you can tell around like it could be 10% useful with maybe the first phase of underwriting like the exploration phase and then I can actually have a smooth curve up as I start to drive more agentic behavior through the workflow.

I think where we tend to see like the hype bounce off of people, especially enterprises at this point in the cycle especially is where they're saying, you know, when when people are telling them just trust us, you know, it's just you're just going to like flip a bit.

It's not going to be a sliding scale, it's going to be a bit flip and and hope that it works. Cuz I think we all know that that's not how it works. And I think what's actually kind of ironic is that's not how it works with any human user either, right?

It's like if you're training an employee or like the Japanese call it like desi, like an intern, it's like you're gradually leveling them up over time, right? And so I think we're very bullish on that and I think we're seeing a lot of very cool use cases in in that sense. Can you talk a little bit about the tariffs?

Uh I'm sure that a lot of customers are uh rethinking their entire supply chains potentially. Uh how does Palunteer play into making great business decisions in uncertain times? Yeah, I think you it kind of gets what you were saying at the very beginning around ERP systems, right? And kind of these rigid legacy systems.

And in some ways, you know, these system shocks, these exogenous system shocks especially are a bit reminiscent of things like COVID where it's like it kind of tests the state of your software, right? It's almost like an environmental pen test against the state of your operations.

And so I think where, you know, we've tried to provide value for existing customers and where we're engaging new folks is where they're saying like, hey, like we feel trapped by the rigidity of the existing system structure, right?

like you know our ERP models things one way our second third and fourth ERP models things in a discontinuous slightly different way same with our MEES systems our PLM systems our ordering systems and it's like what we really need to be able to do is understand a what's going on across everything and then be able to impose strategy and kind of reconnect strategy with operations in a way that can actually be adaptive right I think what we're not doing is going people and saying hey there's a magic wand that will magically fix all the tariff issues uh you know no matter what the circumstance is but can we instead give you like an operating system to be able to say you can now actually like kind of like you're a pilot in a jet be able to navigate through as conditions are changing test things get leveraged through AI and automation and kind of be on the front foot when it comes to this stuff and I think you know we we've had customers who have talked about their supply chain journey with Palunteer with General Mills Tyson others and I think they're great examples of people who have like it's almost like they're retaking control from like all those legacy systems can you talk a little bit about the uh kind of the broader Palunteer ecosystem I know that obviously Palunteer sits on top of a lot of like onrem servers and cloud different mixes of cloud tech and then uh now with AIP uh there's now companies that are building on top of Palunteer.

So they're kind of uh sandwiched there. Uh how are both of those going? Where are the where are the pockets of opportunity that you're seeing in like the high growth uh that's maybe indexed to Palanteer's growth which has been fantastic um but uh but outside of the Palunteer direct ecosystem. Yeah.

No, it's it's a great question as well.

um on like the stuff below I would say like it's like the infrastructure we deploy on there's of course like a broad strategy to keep adopting more and more infrastructure like substrates we can deploy on right so whether that's metal that's the three major hyperscalers that's Oracle um that's being able to deploy you know in sovereign clouds like we've built our Apollo platform to allow us to be able to kind of do this in a highly leveraged way without having to rebuild those core components in every single environment um that I think like kind of there there's a general trend uh towards more options there um I think We're also to your point like kind of riding the wave around people have gotten to the cloud or they've lifted and shifted a lot of data and maybe some data science workflows but now it's like it's time to operate.

It's like okay like am I actually like are the core business workflows those nurse scheduling workflows those underwriting workflows actually changing and I think there's a ton of like kind of tailwind from people saying I did the first step now like what's the operating system I'm using in the cloud and I think that answer more and more is Palunteer.

um things on top I think that are really interesting there's of course like there's partnerships with the AI providers there's partnerships with you know folks like data bricks and others but I think what's really cool and we actually just announced I think a couple minutes before we went live here is um like it's it's the startups cohort so it's people being able to actually build their entire kind of businesses on top of Palunteer and say I like the leverage I get from the integrated you know data tooling model tooling workflow tooling it's like I don't want to start from scratch like 2016 style with the hyperscaler I want to build like on the 10th floor now and so there's a ton of great uh startups there.

Yeah.

Talk about why companies should be taking that product extremely seriously because I remember there was not that long ago when people said oh you know don't like don't build a Shopify app like you know uh and then you have like all these unicorns that kind of like started as don't build a Chrome plugin and Honey sold for a billion dollars.

Yeah. So I think most often yeah if you want Yeah. If you want to do anything in especially in industries around national security, I feel like building on top of Palunteer is a way to like, you know, potentially accelerate totally trust and things like that. But uh would love to hear it in in your words.

No, I think that that's that that is a that's well put. I think there's two dimensions I think about, right? One is like you said like it's being able to get leverage in kind of the it's access and compliance and these sorts of things. So it's like I want to work in I want to get access to problem sets or mission sets.

And so our mission manager offering on the government side allows you to be able to get high side get into these domains and deploy your products. And there's a ton of great examples of that. And I think people see a lot of value. That's kind of more of the infrastructure enabling you as a startup.

I think there's then building with kind of the platforms that are at kind of the above the infrastructure level.

So this is foundry and AIP and there are folks saying uh like like there's there's Helmgard and others who have presented recently that say like to build the like the AI enabled applications that I want this is the toolkit I want to use to build them right because kind of everything from that data integration layer the ontology layer the workflow layer the SDKs all of that stuff is like now it's the equivalent of like the you know in in the personal computer era like the op it was great to build against core components on my IBM PC or my son workstation but it's like I could really use an operating system to be able to like build at a higher level and build much more sophisticated things.

So I think that's what we're seeing. We had our second developer conference uh 3 weeks ago and they had a bunch of startups there kind of telling that story in effect. Uh I I think I already know the answer but do you feel like you guys are just getting started?

I imagine with these you know customer relationships you have right now and the existing products every single time you you maybe like think you hit the end of the product roadmap for like a certain category it's like you know you you get the next uh you know request from a customer and it's like hey there's this whole area and opportunity that we weren't even thinking about a year ago much less five years ago.

You took the words out of my mouth. It's like it's it's kind of an incredulous thing to say.

I realized a little bit of like self-awareness in Silicon Valley to say like you're 12 and a half years in and it still feels like it's been a blink and there's like it's just it's like the I joke it's like it's the end of Dune book one in like the first book.

It's like it's like we're just getting like we haven't even gotten off of Arachus yet or whatever or like whatever metaphor you want to use and it's like I totally feel that way because it's like a lot has happened, a lot has transpired and I'm it's meaningful and there's so many people that were part of the journey that are now in the ecosystem of folks you guys have been talking to and and others broadly but it does really feel like it's just the beginning.

I I want to know about uh the shift in security mindsets specifically in the commercial sector. We've heard a lot of talks about uh corporate espionage even in B2B SAS but obviously there's uh you know a bigger national story around security even outside of a government context.

Are you seeing increased demand from corporations that maybe don't need to be ITAR compliant just around taking their software and data more seriously in terms of security? 100%. And I think that manifests in a few ways.

One is just like it's it's the core kind of like hygiene or meat and potatoes part of that which is like it does feel like the you know the threat vectors are increasing. The threat profile against any prominent company in America is increasing.

So you know the way that we operate our core workflows, the way we secure our data um I think is more top of mind. It's not just sort of in the purview of the CISO or the CIO. I think everybody's thinking about it now. Um especially given kind of global volatility trends at large.

Um, I think there's also this sense of like if if we're going to have to take these big leaps forward with AI or with building the next generation of digital applications, like these seem like they're going to be big shifts in the business and they're going to affect a lot of, you know, not just the technical users, but the business users, the operational users.

So, if we're building like essentially the next foundational parts of the business, like I want to be very sure that we're building it in the right way.

And I think as you guys probably know like a lot of people have battle scars from the prior generation of stuff not having built been built the right way or there being you know things that people have had to deal with off the back of those designs. So I think pe people are definitely tuned into that.

You're you're totally right. Yeah. Uh culturally what's changed as Palunteer's become a public company? I know you've been there before when Palunteer was private. Now you're public. Uh is it stressful having a stock price out there every single day? Has it have things kind of been business as usual?

What's the shift been? I think honest for those who have been around, you know, I think what's cool is that the the hiring profiles and the people that are coming in seem as good now as they ever have been.

You know, this is the shameless plug for we are hiring all the things, but it's like it's still a very young and vibrant place, which is why maybe like the temporality in my mind is a little bit broken as a result of these things.

I will say one thing that's different for sure is like your friends and your family and your college friends like talking about the stock up and down. It's like people couldn't pronounce the name. people had no idea what it did for like most of the time I've been here.

So, like it's still very bizarre, I think, for anybody who's been along or or been around it to be like people care this much about this.

And it's like, of course, we always cared, but like that's still kind of bizarre when you get these ping like, you know, LinkedIn request from somebody went to high school with and he's like, "Hey man, I'm in the retail investor community. Keep it up.

" Or it's like it went down and it's like, "Hey, can you do something about it's like one or the other, right? " Um, can the devs do something about the stock price? How do you how do you personally think about AI safety? So, not not speaking for for Palanteer on this one.

My my you know, one of the reasons I'm curious is, you know, reacting to AI 2027, which who knows if you even had time to like read through from last week. I I'm not halfway through, but but like I did the mistake of like watching the interview and then reading some of it, so I have to do one or the Yeah.

No, but like in in in the sort of like doomsday scenario, like I imagine like you and the Palunteer team are like in a war room with the president and it's like, "Hey, we got to like figure out how to like, you know, save humanity from the bots.

" Um, but I'm curious like um you know, general maybe reaction to that piece and then, you know, how how you're thinking about it more broadly. Just the shift from pro uh deterministic computing to probabilistic computing injects like just so many risk factors. I I I'm dying to know what you think.

No, I think I again I'm I I'm on the same wavelength as you guys. I think like a I like those pieces. I think even if people dismiss them or try to critique them for being kind of sci-fi, it's like I think it's good that people are doing some definite thinking even if it is a bit forecasted.

Um I think this is true of like, you know, the piece, you know, uh Machines of Loving Grace that Daario put out as well. And it's like I think it's worth having these sort of like future looks just to be able to at least game theory through some of the implications here as you're saying.

Um, I think what you touched on, John, I think is exactly it, which is like if you assume these things are probabilistic or probabilistic in some degree and are going to have more agency over time, like what is the outer shell of asurances that we're building to keep the machine subservient to the human, right?

In like the grand sense of like human agency, right, as as people of the world. And I I think like that's where careful system design from our perspective is so key.

It's like there's kind of testing the models in isolation, but then there's like the live fire exercises of how do these things actually perform in indicative contexts like whether that is a battlefield or that is a production line or that is a hospital.

Um, and I think like we have to evolve the testing frameworks um to take into account more of the operational theaters so to speak and then be able to pull back the requirements from there. There's a lot of pontificating.

I think that's easy to do when you're just looking at the models themselves against benchmarks, but it's like how do you evolve past benchmarks to operational testing? And this is kind of what we've done with other sectors, right?

With like whether it's electrification or it's like other kind of core core assets that make up the digital world. Not in the same level of sophistication, but we had to evolve how we're measuring efficacy, right? Yeah, for sure. Um I I I feel like Palunteer's taken a pretty uh foundation model agnostic approach.

uh kind of a bring your own uh bring your own foundation model if you're a company. Obviously, you do have some partnerships that are bigger than others, but uh that makes sense from a commercial or DoD perspective. Um but uh what has been your reaction to DeepSeek?

Uh it's open source and people think, oh, if I run it on my own hardware, I'm all good. But there's been kind of a sci-fi debate about is it possible to embed kind of a Manurian candidate in the weights that wakes up when it is like hey I'm it turns out I'm in a DoD data center I want to do something different.

Um is that pure sci-fi? Is that something that you need we need evals for? We need extra testing. We need to uh air gap and then kind of test in a in a in a you know offline environment first before we deploy these or do you think it's all sci-fi and it's not worth worrying about?

Well, I think there's probably some like arisatilian meme there, right? Where it's like I think even if it's not all sci-fi, it's like I think there are there are real challenges across like the intersection of different like different considerations.

One is like our like our ability to mech mechanistically interpret these models is still limited, right?

So like even if you assume it's not malicious behavior on part of the deepseek or the or a CCP uh you know you know proxy you could still have a situation where the models are not understandable and therefore you can't trust them for the spectrum of utility.

I think given that ambiguity and also the ambiguity around like h how you know what was the story behind Deepseek's creation. I think there's less ambiguity around the control that the government kind of can exert on them uh in China.

I I think it all points to like why would you add up that composite risk profile and kind of go all in on that if you had open source options um or equivalent options in the western world right and I think this is why it's important for us to have the ecosystem of proprietary models commercial models which are great but also open source models like the llama release that we saw uh and and you're going to have different cases where like the the the abilities the cap the capabilities and the controls provided by each of those different types of models is going to be suited for different domains.

Like for instance, there are going to be certain government domains where a commercial model provider will not be suitable. Um they're going to be domains where it will be and it's like we want to have as the west as the US especially the full range of options to come to the table with.

Um I don't think we want to be in in a in a case where we feel like we're backed up against the option of having to go with a foreign model, especially one from a rival regime. Yeah. Do you think uh every kind of Fortune 500 company needs their own set of evals for when these new LLMs drop?

Because you see what happens on LM Arena and it's kind of driven by uh like the vibe of whoever's voting that day or uh even MMLU and some of these like IQ tests uh don't necessarily translate into better business value.

And so, uh, I'd love to know what you're seeing from, uh, you know, Fortune 500 folks or bigger business folks on how they make decisions about what models to use. Is it just cost to to benchmark or is there more nuance to it? Yeah, I feel like there's a great tweet in there.

I mean, like Eval's vibe like vibes are not enough or something like that. But it's like I I think certain absolutely right.

I think this is part of like the idea or the thesis we have for like this shared model of decision-m that is this ontology system like the how we design eval is literally around that right you could you can say there are sample workflows there are you know unit tests that have to pass every time you upgrade the a you know the agentic logic for that underwriting workflow at AIG or whatever it might be and and and you know it's not just like an abstract set of tests these are actual like invivo tests that you know have to pass and yield certain results either deterministically or probabilistically for us to be able to go forth and and continue with this workflow.

And so even if nothing changing model is shifting on us, we want to know that.

But I think you're totally right like enterprises have to get into a mode I think of thinking about the way they're instrumenting and validating the application of AI as being kind of one and the same with rolling out these emails in an operational sense, right?

You can't you can't measure you can't kind of control or improve what you can't measure, right? And I think it's it's kind of the same concept here. Bring AI into these critical spaces. Yeah.

It's kind of like a the way a hedge fund would back test a strategy and see, okay, yeah, if we apply this to all of our historical data, it's like new model comes out, let's just re-evaluate every piece of debt we've ever underwritten or every policy we've underwritten and see if we get a better performance.

Uh that'll be fascinating. I I'm imagine that there's a business to be built there. Maybe it happens at Palanteer or somewhere else, but uh very very cool uh idea. Uh Jordy, do you have a last question or No, this is Should we let you get out of here two minutes? Thanks so much. Looking forward to the next one.

Yeah, this is fantastic. Looking forward to it, guys. Thank you so much for having me for making the time. Cheers. Uh let's run through some timeline to finish it out.

Um, uh, Cole Rottman says, "Mazletov to OpenAI fund for turning $8 million into $600 million of paper gains from just two deals in $175 million fund based on his math. " Did you see this? They did Harvey. They led the $5 million seed in 2022 and now the series D in 2025 is a $3 billion valuation.

Sam Alman is is pretty good at being a VC when he wants to be. It's insane. And then cursor, they led the $8 million seed in 2023 and the series B is at a 2. 6 valuation and uh might be getting like a 4x markup soon. I think we see we saw some leaks around.

So uh could be you know a 5x fund off of two deals in just three years which is remarkable remarkable uh you'll love to see it. Uh anyone who's in that fund any exposure those deals they're going to be sleeping well but they could be sleeping better. They got on an eight nights that fuel best days.

Turn any bed into the ultimate sleeping experience. Go to eightsleep. com/tbpn. Um Lindy man says there's nothing to fear. Smart people adapt. In the 2010s they went into software. In the 2020s they will open up factories. It's all the same thing. Finding angles and edges.

I thought this was a kind of interesting white pill you threw in there. Um it it is interesting that like that's the nature of capitalism. It's uh it's you know water finding cracks in the middle midst of a jar full of pebbles or the way sand flows through a jar of pebbles.

Um if tariffs go up and there's a bunch of incentives like you know the hadrians and the base powers. They're ahead of this curve but that's not the end of the story. There will be plenty of people that see the news if the tariffs stick and say hey this fundamentally changes the economics.

I'm ready to write checks if I'm a VC. I'm ready to build a factory and yeah, maybe it will be a slightly different in work environment than building B2B SAS, but uh I'm ready to make it fun.

I mean, a lot of the one of the coolest things that's happened over the last few years has just been like the Elsa Gundo movement and like the Aaron Slodovs and the re-industrialized folks just making, you know, that that meme obsolete of like everybody wants to re-industrialize but no one wants to work in a factory.

Well, like a lot of people are down to work in factories now because it's not that dirty. It's actually like, oh, there's just a big machine over there. There's still anybody that headphones, you're fine.

Anybody that has ever made made something, you know, in the in the context of a business, doesn't really matter what it is, right? Even if it's a startup and you get a little bit of revenue and you make make some merch and you can suddenly hold, you know, a product in your hand. It's amazing.

So I actually do believe that people many people would rather you know work in a factory or work around factories uh and and be involved with producing things not just software. Well after you produce things and you start selling it online. What are you going to have to do? You're going to go pay your sales tax.

You're going to go to Numeral HQ. It's sales tax on autopilot. Folks spend less than five minutes per month on sales tax compliance. Go to Numeral HQ. And 25 states are now taxing software sales. Yeah. get in front of it. Just do it. And we had a really cool launch video from Zipline.

This company's been around a long time. Uh we should ideally up the video if we can. But regardless, Zipline is a drone. Uh they're kind of vertically integrated drone delivery. They make the drones, but then they also uh run the service. And so they were in Africa doing blood deliveries via drone.

So it's a very important to get it there very fast. um someone's bleeding out, you need a blood transfer, send it from the central hospital as fast as possible. So, even though drone delivery is expensive, totally makes sense in that in that case.

They built out a ton of technology, ironed out a bunch of problems, and now they're ready to bring it to the United States, and they are launching in a few states, Texas and a few other places. Uh, seems really cool. And we were talking to somebody here. We can watch this video because I think this is so cool.

I love to work on things I'm passionate about. I think when you're passionate about something, it doesn't feel like work anymore. You might observe it as like type two fun because often if you're passionate about something, you're willing to do things that other people would find harder than anybody would want to do.

But it turns out when you're passionate about it, it's it's just, you know, it's fun. Yeah. It's just such a different vibe than the normal nerd vibe reel that's like really intense. Play a role and build over the top of the future that we will be proud to hand to our chilling is about as cool as it gets. Warm.

I know it's been a lot of um sacrifice for a lot of people to make this happen. I don't think any of us thought it would be sci-fi future. Yes, the drones drones are going to be a little bit noisy problem.

The the magic now of pressing a button on your phone and getting an item totally in uh you know and we were talking to someone who I think was an investor in Zipline and he said that uh something like 80 to 90% of all of the goods in a Walmart fit in that drone. Yeah.

And so, yeah, you're not going to drone delivery like a couch, but like for most things, you can actually use this service and hopefully the economics pencil out in in a in a viable way. Uh, I definitely want to have the the team on the show soon and talk about it.

But really, like above all else, it just stuck out that they went with such a different style and aesthetic for their launch video with something much more ethereal and and and it just feels like summer and it just feels like you're just relaxing and a drone comes deliver and it's like, yeah, this is cool.

This is sci-fi. This is tech, but it doesn't need to be cyberpunk. Like, it's the future. Feels like the future. There's something futuristic about this, but it's still very pastoral and it feels like something that integrates with uh just like the kind of vibe of like hanging out on a ranch in Texas.

It just it was just a really nice video and and and it was just it was just cool to see someone take a very different direction from what we've seen that it's been a cool meme. We've played into it a bunch. Uh but it's maybe at the end of that road and maybe something like this is what's next.

So uh go watch the full video and uh sign up for if you're in the the delivery area I guess. Uh Will Nitsie says the only career question that matters is what can you be world class at? Everything else is noise. I like that.

It kind of paired with what we were talking about yesterday with the uh the intersection of two uh novel ideas or areas of expertise being something that LLM struggle with and I think this will be a a continual big opportunity for humans. Big opportunity for humans.

Uh yeah, I mean I I think AGI in many ways has been achieved, but very few people would say LLMs are world class at really anything. And they kind of go through this galman amnesia thing where if you don't know about a topic and you ask an LLM to talk about it, it can kind of get you up to speed.

But when you talk to somebody who's much deeper, much more aggressive, much more um definitive in their worldview around that topic, they're going to give you much more insight. And that's why why I would disagree. They're world class at creating Studio Giblly copycats for, you know, pennies on the dollar. That's true.

They're they're they're world class at creating viral content as I tested it because the actual Studio Giblly photo I posted from the real movie Spirited Away, not a banger. Flop flop. But when I took Oenheimer and made it Gibli, it got like 30,000 likes. So people like that. Uh let's go to MDA.

They say VCs told me build a pitch deck, design a road map, hire a dev team. But no one told me post Tik Toks, trigger FOMO, make it look like a trend, not a tool. That's when I started thinking like a drop shipper, not a founder. What do you think? Insight? Uh I don't know. I don't know where to take this.

Um, but I I think if a lot of companies posting Tik Tok's good way to get customers.

I talked to a uh a Gen Z founder who said that uh my distribution strategy is if I'm in a if I'm building something for consumers, I'll make viral Tik Toks and if I'm uh selling something B2B, I will hire Nepo babies because the Nepos will have uh parents who are involved at businesses. Yeah.

and and they will be able to get my idea. You know, one one thing this uh made me think of is uh this guy Cormarmac who's built a company called Oasis. He started posting information about uh water quality.

Uh started posting Tik Toks and you know eventually built out a database where you can sort of get access to different data. And he basically tapped into something that was like very much a trend of people wanting to understand what's what's in their water, what's in other products that they're consuming.

So yeah, I think in general powerful to um tap into it, you know, make sure that your tool is part of a bigger trend. Otherwise, it's going to be really difficult to get any type of real traction. Totally. Yeah, the hype can get too much and you can get lost in the sauce and you can just become like a Tik Tok company.

Uh but there's something about the communication and casting a wide net that clearly lets things ramp. I think we probably saw that with the the the uh the Craya team. they had like what 20 million users or something like you don't get that without going viral.

Uh I'm not sure if they did it through TikTok, but they certainly became a trend and that enabled a lot of fundraising and stuff. Speaking of fundraising, uh uh 25-year-old police drone founder uh Blake from Brink Drones just raised a $75 million round led by Index. I met him camping like a year or two ago.

Uh he's teal fellow, great guy and uh fascinating. I love this photo. The drones just hovering behind him. Yeah. Yeah. No, he's a fantastic guy. They come with me everywhere. Yeah. Yeah. Yeah. And he has a very interesting uh use case.

So I I really want to have him on the show to talk about like how do we build the American DJI, you know, obviously like how do you get to market with something reliable? He's going after uh drones for police because the police legally cannot buy DJI anymore or or in many jurisdictions they've been banned.

Um, but if there's a broader ban, um, I'm kind of getting burned out on the defense tech stuff and I want products now. Um, I want I I want there to be, you know, world peace and then everyone's like, "Yeah, actually, we can pivot back into hardware, but we did this whole thing with with Nuros, right?

Saurin was a drone racing pilot. I imagine he wants to, you know, save the world and then get back Yeah. into his original passion of racing drones. So, you can imagine Nuros doing something in consumer. Totally.

And it's just like a it's just a nature of the market that um if Saurin at Nuros had said, "Hey, I'm this fantastic drone pilot. I'm going to build a consumer drone FPV product. " People would be like, "That's too capital intensive. You'll never beat DJI. Blah blah blah blah.

There's a million reasons why this won't work. " He wouldn't have been able to get funding. Um, you know, let let's uh let's bring about world peace and then get back to having fun with some drones, making some cinematic videos. That's what I want.

Anyway, uh any other posts you want to close out on or I think we're good for. This was a fantastic show. Thanks for watching everyone. Uh don't forget to leave us five stars on Apple Podcast and Spotify and stay tuned for the next one. Just do it. We will talk to you soon. Looking forward to tomorrow. Bye. Cheers.