Earth Observation with AI

Martice Nicks III is the cofounder and CTO of Danti, a search engine for exploring the extensive collections of Earth observation data available today. He was previously at Orbital Insights, Maxar Technologies, and other companies working on geospatial data.

In this episode, we cover a range of topics including:
- What is Earth observation data
- Where does the data come from
- The founding of Danti
- Use cases of Earth Observation data
- How do you index this data and make it searchable
- gSEARCH challenge
- What AI tools are being built for defense purposes

Martice's favorite books:
- The 7 Habits of Highly Effective People (Author: Stephen R. Covey)
- Dresden Files (Author: Jim Butcher)

--------
Where to find Prateek Joshi:

Newsletter: https://prateekjoshi.substack.com 
Website: https://prateekj.com 
LinkedIn: https://www.linkedin.com/in/prateek-joshi-91047b19 
Twitter: https://twitter.com/prateekvjoshi 

Transcript

Prateek Joshi (00:01.046)
Martis, thank you so much for joining me today.

Martice Nicks (00:04.036)
Yeah, pleasure to be here.

Prateek Joshi (00:06.118)
Let's get right into it. You work on Earth observation data, and obviously there's so many different areas and basically sectors where this data is insanely useful. But can you define what it means? Like what type of data does it include?

Martice Nicks (00:32.34)
Yeah, yeah. So the scope of Earth data, how we think about it anyways, is it encompasses anything, any information that's generated about a place on the Earth. So that could be measurement type data, you know, like seismic activity, for example, or wave height data from like NOAA sensors, things like that. But also it could be more like traditional in the sense of geospatial, like what you would...

commonly think of from electro-optical perspective of satellite imagery. It could be synthetic aperture radar type stuff, light detection and ranging stuff, which is LIDAR, hyperspectral, all those types of content. There's a lot of it that's generated. That's more of the traditional sense. Then you've got analytics that produce information about places on Earth, things like, hey, here's a vegetation index of how much forestry is here or not.

So yeah, it's a variety of things and we view it kind of all encompassing as long as we can kind of attach it to a time and place.

Prateek Joshi (01:38.39)
Right, and when you think about products that can get built around this, there's a variety of things you can do. It includes data collection, analysis, how do you present this data, this information, and also where do we put it to use. So let's start with data collection. Obviously there's so many different modes in which you can collect this data and centralize it.

Can you just talk about where this data comes from and where it gets stored?

Martice Nicks (02:13.551)
Sure. Yeah, I mean, so a lot of it, there's, I guess, two main flavors that we think about in the in the public place. There's, there's sensors that are collecting information. So those are things like satellites that are flying around the earth. That's aircraft flying around doing collection. It could be drones. Could even be stuff coming off of like your cell phone, you know, taking pictures or things like that. So the way it gets collected, you know, there's, there's

special sensors that are built. It's either taking a picture or it's measuring some kind of activity. A lot of that goes back to the provider itself. So those could be things like Planet or Maxar, these big companies that actually manage building the hardware. They usually handle bringing the information back from those sensors into their centralized repositories. And then they have catalogs essentially of data.

that you can access. A lot of it is cloud-based today, so most aren't housing this on on-prem type of servers, but they have API interfaces and things like that you can access it from.

Prateek Joshi (03:28.33)
And if you think about this data, one of the obvious use cases that comes to mind is, hey, weather, I wanna know what's happening on a forecast about weather. That's the regular use case. But apart from that, like, can you talk about maybe a couple of interesting, maybe non-obvious places where this data is used? And also, how do they use it? And what decisions do they make using this data?

Martice Nicks (03:56.039)
Sure. So yeah, I can give you a couple of examples. Just off the weather one, one non-obvious thing might be if you're going out and doing fishing and that kind of stuff, if you're going out in certain boat sizes, you can't go into certain wave height areas. And so you might be looking at buoy sensor data to see where can I actually take my boat today to go fishing. Another one would be kind of around

supply chain analysis. So there's a lot of companies that have to monitor the logistics of the movement of cargo essentially from place to place. And it's traveling through a lot of different modes of transportation. But also there's a lot of interesting metrics that you can get by looking at say like a port and kind of understanding how many cargo ships are showing up here or not. And is that very close to a mining operation? And it starts to give you some economic...

view of how goods are moving, and you can actually use this content to do that. Another one that's maybe more common practice would be around your properties and your addresses. There's a ton of information out there at the city level that has information about say, permits that were pulled on a home or for a construction site.

There's a ton of risk information that gets put out by a lot of different companies as well as by FEMA too, which is managed under Department of Homeland Security. And they give you all these very interesting risk assessments about these areas that a lot of people don't necessarily know are there. And so one of the things that, you know, we look at doing is kind of pulling all of that together so that you get that single view of, you know, maybe risk on a property, what construction has been performed on it.

and what are maybe some of the images around the home, value, that kind of stuff. So you can start to determine from, say, an underwriting perspective, what's the risk of a home if I were to write this off in an insurance claim or something like that. Or if I was a property investor and I wanted to understand, is this a good investment or not, I can see all of those factors all in one spot.

Prateek Joshi (06:10.03)
Amazing. The boat use case. I always thought if you have a nice big boat, you just set sail into the open water and just do whatever you wanted, but apparently not. You can't just hang out anywhere unless you be hit by a giant wave. That's very interesting actually. And you know, funny thing, ships, they've been around for so long, like sailing on water. Has been like a, for centuries, people have been doing that. And it's fairly...

complicated actually. You can't just like go out and imagine the number of variables you have to deal with when you're out there in the water. It's crazy. So that's a very interesting use case.

Martice Nicks (06:50.355)
Well, and even another one, just thinking about the water stuff. I mean, we have all the goplots out in the Gulf, the big oil rigs and things. And let's say there is an oil spill, as an example. You can use synthetic aperture radar. You could use a lot of the current data and information to kind of look and see how that oil slick would progress in the ocean and where it's going to go.

And you can actually monitor, you know, kind of how that moves. So there's like a ton of like really interesting weird things that you can kind of start to dig into and that stuff.

Prateek Joshi (07:27.794)
Maybe this is a good stopping point to talk about Denty. Can you, for listeners who don't know, can you explain what you do?

Martice Nicks (07:38.647)
Sure. At its root, all that we are really focused on from a product perspective is enabling a user to put in a plain English question, like just a normal question to the computer, and being able to get right context, right time content. And today, you kind of have to have a lot of tradecraft and understanding on how to get all of this data.

And we want to lower that barrier to entry so that you could just ask like, Hey, how old is my roof at, you know, one, two, three main street, or, you know, where were the missile strikes in Gaza or, you know, whatever, and be able to kind of pull content back and get a view of like what's happening on the earth anywhere, but not have to understand things like what's the ground sample distance here and will that actually help me understand it? What's an incidence angle? You know, you don't need to know.

Prateek Joshi (08:28.854)
Right. And the people or the other sectors, obviously there's wide applicability, but have you noticed like a couple of sectors that tend to use it more than others?

Martice Nicks (08:43.791)
Um, so I mean, yes, there are some, some sectors that are emerging earlier than others. I think, um, big ones that I've seen currently, and again, it's just my own, my own sample set, but, but big ones that I've seen are definitely around, um, you know, property investment, uh, and insurance are kind of big. Um, I'm starting to see it pop up in the agriculture side, like in terms of interest, uh, and then.

Definitely in the supply chain analysis. So understanding like anybody that it has a lot of assets under their control that are globally positioned If you're the manager of those assets, you can't be everywhere at once And so there's a lot of questions that come out of companies like that where they're like, hey I got this asset across the other side of the world I need to know like is the part of the factory down like are there anomalies happening there that are causing production slow? Or I need to monitor a construction site

and I can't get there all the time. So people that have those needs where there's a lot of assets under management that are very spread out geospatially, I tend to find that they have a very big need for this and are adopting it a lot faster. But there's a lot of other tangential, you know, personas that are starting to come into play and see the value.

Prateek Joshi (10:05.81)
If you are dealing with a lot of say text data, then people have been building search engines for a while where you take the text data, you index it, you make it searchable. So that's one way to do it. But in this case, it's not just pure text. You're dealing with a specific type of data. So can you explain how you index this data and maybe compare and contrast that with indexing purely.

text data, like what are the additional things you have to do to make this searchable and usable.

Martice Nicks (10:39.031)
So I guess I'll start with the latter. So like in traditional information retrieval, you're really looking at like token frequencies inside documents. And so you're at its root level, you might do like a strict match and say, say you were looking for flooding, right? As a keyword, it's keyword based. So when you do that, you type in your keyword flooding and then it looks up all of the documents and says,

Here's all the documents that have an exact match of flooding in it. And then based on term frequencies, things get ranked higher and that kind of stuff. Um, but also like, then you expand it a little bit further, you get into fuzzy, fuzzy matching and even like synonyms around flooding. Like there's alternative words that might be close in the semantic meaning of the space. Um, and so you're trying to like, maybe upgrade documents or expand your retrieval on that. So that's like usually the scope of what you're kind of dealing with in a traditional just text.

text-based search. On the geospatial side, like when you're dealing with Earth data, it's not just about the terms, but it's also about where was this information collected. It could also be where was it observed from, where was it collected. There's a lot of other positional elements that can come into that can change decision making.

And then also, what's the time element of this? Because there's a lot of activities or things that happen and they're fleeting, it's not fixed. A boat that's in one position at one hour is maybe not in the same spot in the next, and vice versa for cars and other things. So there's a lot of movement and time-bound context when you're observing these places, and there's also fixed ones.

adding in the spatial and the time component to the retrieval of documents as kind of like the expanded context of that and then also adding the semantic hydration. So like what did you actually intend to ask about? And that's a lot of what we're trying to make in on our product is kind of connecting the dots between traditional information retrieval, also making sure that we have the spatial, you know, the space and the time piece.

Martice Nicks (13:00.275)
correlated in there, and then the semantic relevance.

Prateek Joshi (13:06.59)
That's very interesting. Let's shift gears a little bit. Earlier this year, you won the G-Search challenge. Can you explain what it was?

Martice Nicks (13:19.475)
Sure. So G-Search was a challenge put out to industry by NSIN, which is the National Security Innovation Network. And they were jointly collaborating with NGA, which is the National Geospatial Agency. So NGA is responsible for doing analytics and kind of disseminating that information out to military components.

And then this Ensign network, they're really a group that's focused on engaging with industry and trying to bring industry in and get that, like make it easier for industry to sell into the government kind of thing. So we competed in this challenge. The scope of the challenge was they wanted to know what technologies were being built today to do information retrieval on geospatial data.

They've got a ton of this information being generated at any given time. It's very hard to sift through all of the content that we have. And you typically, if you're taking a look at some of this high resolution imagery, sometimes as an analyst, you're in there and you're looking meter by meter or kilometer by kilometer, right? And you're scanning this whole image and then there could be thousands of these things coming in. So there's just not enough.

people to necessarily prosecute, like looking at all this stuff. So one of the things is like, how do you quickly find the right images based on the need that you have in the moment? So the challenge was around being able to bring a search engine to the table that allowed their analysts to find content for remote vendor archives. So they can search their own content, okay. But it's harder for them to keep up with like...

all of the industry advancements. I mean, you've got a ton of the space industry going crazy right now with putting satellites up. There's all these new instrumentations. You've got really cool stuff going on with very low Earth orbit satellites coming into play that are imaging at airborne level resolution, which is like 10 centimeter or better, but from space, which is crazy because your revisit rate is much higher than an aircraft. And anyways, so...

Martice Nicks (15:40.839)
putting all of that together, it's an explosion of content. And the way that you access that content is very siloed and segmented. I mean, like literally vendor by vendor, you would have to set up all these connection points. So they were looking for a technology that would allow them to kind of bridge that gap and be able to see, well, what are the commercial holdings for this stuff versus what we have? And how do we more easily

find this content. So that was the challenge was that and then also allow them to like search their own internal content. So we did that. And then the and part of that was also to the government actually has a mandate where a lot of the imagery that they buy depending on like under what contract ceiling they buy it, a lot of that can get shared out to like anything that's commercially bought, that gets shared back out to US citizens. So

So they wanted to better improve the partnership with academic and US citizens on putting that content back into their hands because it's the people's content.

Prateek Joshi (16:56.718)
You know, it's amazing. First of all, congrats on that. And it opens up so many doors. Maybe let's talk about working with the government. Obviously, you won the challenge, led to a nice contract. Hopefully in the future, you'll work more with different government agencies. So what should people, or specifically, what should founders know about working with the government?

Martice Nicks (17:01.651)
Thank you.

Martice Nicks (17:26.035)
It is definitely a different beast than working with commercial. I've been through a couple of companies now where I've seen the bifurcated, both commercial and federal, and the business models are definitely different. One, when you're dealing with the government, you have to be prepared for very long sales cycles, sometimes up to 18 months before something really closes. There's other things that you can do that'll fast track.

But you just have to be prepared for that. And a lot of it is also around being able to interact with the people that are putting the contracts out, which they always make the contact information available. But getting in there and asking them questions about, well, what challenge are you really trying to solve? And then thinking about how your product or solution actually supports them. There's also kind of a, this is a personal philosophy, but mindset wise,

I prefer to always put myself in the shoes of the war fighters and think about, hey, if my life is on the line, for example, because we got a lot of men and women that are, is this solution really helping them or is it not? I think going in with a very serious mindset around that, hey, look, we're trying to do our patriotic duty and provide solutions that are solving real problems for you, I think is important.

tends to resonate very well. And so getting plugged in with the different innovation units, there's a ton of works like Afworks, Spaceworks. Socom has, I forget what the, Softworks is another one, which is down in Tampa. There's a ton of these innovation units that are specifically being designed to engage with industry. And those are really important to get plugged in.

with and starting to build those relationships and that will really help you in that path. But I'd say those are probably like the big top things. Get plugged in, have the conversations and make sure that you're like really putting their problems first and how you're solving for them.

Prateek Joshi (19:38.142)
If you're selling software to, let's say a company, then you would use a traditional like a SaaS model where, hey, use the software and proceed per month, or, hey, we'll give you, can process X amount of data per year. So there's like a recurring component to it. But government doesn't really, I mean, they've been doing business for a very, very long time, and they're gonna do business the way they want to. So can you talk about...

Martice Nicks (19:55.376)
Heh.

Prateek Joshi (20:06.85)
the business model that different agencies use to engage with vendors, especially the ones that are selling software.

Martice Nicks (20:16.059)
Yeah. So you can, we're still a little bit earlier in our phase, but I think about it in kind of like two major phases. So phase one is if you're in more of the research and development side of the house where you're still finding the product customer fit type of thing or product market fit, those contracts are usually a professional services type of

the product sales that you're thinking of, right? That we would see traditionally. However, later when you meet certain technology readiness levels, part one, and then this is actually gonna go to your earlier question as well. Security is another big component, so you have to build against a risk management framework a lot of time for the government, so.

Once you're meeting those credentials and you have TRL levels, then you move to different contract type, which is an O&M, like operations and maintenance type of contract, where they're actually buying licenses. They like to buy at year levels and that kind of stuff. They like prices that don't skyrocket for no apparent reason kind of thing. They do have the traditional.

buy but it's usually a bulk buy and once you actually meet certain technology readiness levels and Accreditation standards and the accreditation. I mean like the risk management framework type stuff

Prateek Joshi (21:51.51)
That's actually a good point, you know, because as you said, selling to industry versus selling to government, they're fundamentally, they're just two different go to market motions and they just had to be prepared for that. And obviously if you do a good job, there's a lot of business that can come your way. Right. Let's talk about the role of AI, especially for defense purposes. And earlier you made a very good point that, hey, if you had to put yourself.

in their shoes, you want to ask, hey, is this a useful tool? Will this help make them faster or better, keep them safer, give them more intelligence? So if you look at all the big areas that can be impactful, obviously, Earth observation data, hugely important, very strategic for many government agencies. And AI can play a big role in taking that data, converting that to useful Intel. What other AI tools?

maybe your top two to three favorites that can be built for defense purposes, looking forward.

Martice Nicks (22:58.517)
Um.

Well, so I mean, I guess it really kind of depends, but I think a lot of the areas that I think AI can impact currently would be around the human to machine interface is really big, right? And bridging that gap in a way where you don't need experts to go and get the information that you need as, you know, someone that doing the execution, I think is a really, really big deal. That's like a huge area.

And there's a lot of gradiation, you know, and how that would play out depending on where you're looking, you know, inside the government. So I think that's like a really big area, big opportunity, especially with like what we're seeing with current AI trends. I think another one would be, you know, around just how do you characterize and classify information to...

uprank it in a way where it's important for you to look at. So because we get inundated, and you deal with this a lot too, right? I mean, think about how many emails you get a day or text messages or phone calls or any of those things. We're inundated with information so much, and it's the same problem out there for them. So anything that helps cut that noise down and improve the trust in what is coming up above that line, that threshold, and saying, like, hey, you should pay attention to this.

Anything that's going to help with that I think is really important for us to build because there's just so much content it's really hard to deal with right now.

Prateek Joshi (24:39.702)
Right? Earlier you talked about satellites. So I wanna touch upon that for a second. If you were to do a quick 101 on, hey, here's where satellite technology is today. How do you describe it quickly in terms of what they can do today and where it's going maybe in the next three years?

Martice Nicks (25:07.763)
So I guess the in the nutshell version of it would be, there's kind of like two major parts to it. So part one is how do you get your satellite into space? And obviously advancements with SpaceX and some of these other companies that are doing regular launches and bringing the cost down to launch is really important. And part of why that impacts things is size and weight matter.

as these satellites are going up. So in a nutshell, they're trying to build satellites that are cost effective, that can measure, that can have different payloads or different sensors on them. And so a payload would be like, you know, hey, I've got a camera mounted on here, it's electro-optical, and I can take a picture of the earth, right? Or I've got a sensor on there where it can measure things like maybe, you know, humidity levels or things like that.

from a remote sensing perspective. So it really just boils down to like, what's the vehicle? Is it compact enough to be transported up? And then is it performant enough to hold an array of sensors on it where I can get these different phenomenologies of measurement? And then the next piece is like a communication network. So how do you actually get the data down? Like once you...

taking this picture, maybe one of these things is like a couple of gigs potentially, how do you actually downlink that information in an effective way? So there's a whole communications component that goes in to that. There's two different forms of that today. There's like satellite to satellite, so it's like a communication network mesh. And that just helps relay information quickly to a location where there's a downlink.

So what happens is, you know, as these satellites are moving around in their particular orbit, there's a particular location on Earth where there's an opportunity to download the information, and there's usually a lot of these. And so that's a big part of it as well, is like, how do you set up the right locations for the downlink opportunities so you can, once you collect it, you don't have to wait too long to get it. And then there's the information processing and dissemination, which is like, how do you actually get it into the hands of the consumers?

Martice Nicks (27:29.031)
So that's kind of like the life cycle of the satellite and how we get information collected and passed down.

Prateek Joshi (27:36.398)
Amazing. That's wonderful. It's funny how everyone, a simple example, we all use phones every day. Phones have been around for a while. We look at it, we quickly Google, oh, how is the weather tomorrow? Like we use them every day and yet they're, I mean, they're fairly complicated systems to build. I mean, this boggles my mind. So it's always fun to kind of do a quick 101 to know where it's at. All right, one final question before we go to the wrap.

fire round. And it's around the utilization of satellite imagery. Now, if you look at the images that we have access to today, some are, there's public information, there's private information, and people use it for many different reasons. Like hedge funds use it to make decisions. Like if you are a government, you'll use it for something else. Or maybe you are an agriculture company and you want to know, hey, what's going to happen to the

part of the world. So can you just quickly talk about what data is available publicly and what is available via private channels where they're to pay more money, but it's high quality data. So just can you bifurcate between that and kind of explain what kind of quality jump can we expect as we go from public to private?

Martice Nicks (28:55.869)
Mm-hmm.

Yeah, so I definitely don't know all of them off the top of my head, but some big ones for visualization. So it's put out by ESA, which is the European Space Agency. There's a portal there for Landsat and Sentinel. That's joint effort with NASA as well. So there's like a ton of stuff there. There is the stackindex.org site. So if you go to there...

There's a bunch of catalogs actually. So the stack is a spatial temporal asset catalog. It's like a standard for how they're characterizing information. When you go to that index site, it'll show you a bunch of free sources that are there that you can take a look at. But the resolution on, say, Landsat or Sentinel is very high, like several meters worth. And what does that mean? That means if you think about pixels on your phone, for example,

I, you know, those tiny little pixels. Well, it's the same concept, but they're much larger when it comes to a satellite. So if we said like one meter resolution, for example, that would mean that one pixel is a one meter, like one by one meter square type thing. Um, so, uh, so the Landsat, Sentinel, and some of those other ones that are free in public, the resolution is very high, like some of them up to like 30 meters or something like that. Uh,

When you get into the private sector, things like Umbra for synthetic aperture radar, they have an extremely high resolution product. I think it's down to about 15 centimeters, which is very small. I mean, you can make out details of cars at that point. Maxar has very good resolution imagery, so does Planet. All these provide black sky as well.

Prateek Joshi (30:38.742)
Wow.

Martice Nicks (30:56.247)
30 centimeters and two to three meters. So it depends on their vehicle and all that kind of stuff, but they usually have a range of products for that. So that's on the paid side. So anytime you're getting down to like sub one meter, you're probably gonna be in a paid tier. And then anything kind of above that, there's probably some data that's available there. And that's on the visualization side. There's also a lot of...

content put out by say like NOAA space agency where like the WayPy data I was talking about before, that's actually open data to the public. Also all of the temperature data and things like that. There's vegetation index data. There's a ton of stuff and you can find a lot of that from that stackindex.org. But there's a lot of that free content and then there's a ton of special investments with other companies where they, not special investments, but...

targeted investments where they specialize in building that content and making that available. So they'll build the analytics to extract that information. Sometimes they might even be building the sensors as well. But yeah, there's a lot of those companies as well. So you can kind of start to find them if you have a specific niche need.

Prateek Joshi (32:13.382)
That's a wonderful level of detail on this question. The way you started, I thought you're gonna just give like a very high level like hand-wavy answer. You went way deep in terms of sources and availability and in terms of resolution. So yeah, that was brilliant. All right, with that, we're at the rapid fire round. I'll ask a series of questions and I would love to hear your answers in 15 seconds or less. You ready?

Martice Nicks (32:23.387)
Hahaha

Martice Nicks (32:39.631)
Mm-hmm. All right.

Prateek Joshi (32:41.186)
Question number one, what's your favorite book?

Martice Nicks (32:45.147)
A realistic, useful book, probably Seven Habits of Highly Effective People. My favorite creative book is probably Dresden Files.

Prateek Joshi (32:55.446)
All right. That's actually good. I think I'm glad you gave two here. All right. Next one. What has been an important but overlooked AI trend in the last 12 months?

Martice Nicks (33:08.283)
Well, I don't really like trends so much because they lose focus on things. But I think the main thing that's being overlooked is picking the right tool for the problem at hand. So with the advent of LLMs and the impact of that, it's great and it takes you that last mile, if you will, in things. But there's a ton of great tools that really target things and do things very well in that path. So I think a lot of people can lose sight.

of what tool to pick for which part of the problem they're on.

Prateek Joshi (33:41.906)
What's the one thing about Earth observation data and technologies that most people don't get?

Martice Nicks (33:51.419)
I mean, there's a ton of things. I guess the number one thing is just how that information presents itself, like how you actually interpret the information because of the huge variety that can happen when you collect it. If you're looking at something from the side versus the top, you actually see different things and it's not always obvious when you're looking at these images.

Prateek Joshi (34:19.158)
Maybe I should have asked you like what are the top 23 things because that would all those will all be like equally near. Yeah, it's like one of those things where whatever you tell people it's going to be like new and interesting to them. All right. Next question. What separates great AI products from the good ones?

Martice Nicks (34:24.314)
Yeah.

Martice Nicks (34:31.868)
Yeah.

Martice Nicks (34:38.119)
Yeah, I mean, I think from my perspective, it's AI that's not obtrusive, seamless, and intuitive for a user to use. I don't want to have to really think about it. I want it to be just kind of this natural companion by my side as I do things that makes me better at what I'm doing. That's what I think makes it great.

Prateek Joshi (34:57.82)
Amazing. As a founder, what have you changed your mind on recently?

Martice Nicks (35:05.207)
Uh, how difficult it is to build a business from scratch, I guess.

Prateek Joshi (35:10.144)
All right, next question. What's your biggest AI prediction for the next 12 months?

Martice Nicks (35:21.514)
I think that with the advancements that have been made today in AI, there's going to be a lot of companies that are seeing really good productivity increases without really increasing their headcount. So I think that's going to be pretty big as they unlock that.

Prateek Joshi (35:38.934)
All right, final question. What's your number one advice to founders who are starting out today?

Martice Nicks (35:47.643)
Do your customer due diligence and be willing to throw away your ideas. Starting out with one idea and then getting the customer feedback, be willing to adapt and change if you don't like this problem.

Prateek Joshi (36:00.994)
I think that's a brilliant point here. I think most people just hold on to their ideas for a little bit of their life. And there's no, I think it just misses the point of the whole reason you start with an idea is so that you can iterate based on what the market is telling you. And I think that's wonderful. And the great ones always figure out how to inculcate and bring customer feedback into their ideas. So, Martis, this has been such a wonderful discussion. I love the topic.

first of all, so I'm glad we were able to do this and thank you for sharing all your insights here.

Martice Nicks (36:35.912)
Yeah, it was awesome. Thank you for having me.

Prateek Joshi (36:38.71)
Perfect.