# Gaussian splats aren't just visuals. They're photogrammetry – Ronnie Yaron, Skyline Software Systems - Show: 2D Is Against My Religion (2DIAMR), host Michal Gula - Guests: Ronnie Yaron, Founder, Skyline Software Systems - Published: 2 Nov 2026 · 23 min · Expo & conference podcast · recorded LIVE - Topics: Gaussian Splatting, Photogrammetry, Spatial AI, 3D Visualization - Web page: https://www.2diamr.com/watch/skyline-ronnie-yaron/ - YouTube: https://www.youtube.com/watch?v=EgG4DrAZxz8 - Spotify: https://open.spotify.com/episode/66Fz9yvIBcwh3lHrPlXUxY - Apple Podcasts: https://podcasts.apple.com/us/podcast/gaussian-splats-arent-just-visuals-theyre-photogrammetry/id1878399775?i=1000799413746 ## Summary Recorded live at INTERGEO 2026, Ronnie Yaron, who started Skyline Software Systems more than three decades ago and still manages it, tells Michal Gula and Tomáš Barnáš why scale means the amount of data processed, why Skyline started on Gaussian splats the day the first paper came out, why Ronnie argues Gaussians are photogrammetry you can measure on, how 3D models help AI know where everything is, and how Skyline bought out its VCs. ## Key ideas - According to Ronnie Yaron, scale means the amount of data processed, not the size of the area: Skyline cuts its pipeline into sub-problems and runs single projects on thousands of computers, on premise, in the cloud or both. - For defense customers working without reliable GPS metadata, Skyline developed automated referencing that ties drone images with no usable information to existing maps of the area, according to Ronnie. - Ronnie decided on Gaussian splats the day the first paper came out; according to Ronnie, Skyline extracts a point cloud from the Gaussians that beats regular multi-view stereo on thin wires, poles and vegetation, and users can measure and run line of sight or viewshed analysis on them. - According to Ronnie, AI needs to know where everything is in the real world; Skyline uses its 3D models to locate a video camera on the ground from aerial data and is adding embeddings so users can ask the 3D model questions with logic behind them. - Skyline has about 70 people, is profitable and has no VCs: according to Ronnie, it was funded initially, then bought out the VCs and went fully private again, which keeps the startup mentality. ## Questions answered ### Can you measure on Gaussian splats? Yes, according to Ronnie Yaron of Skyline Software Systems: measurements, line of sight and viewshed analysis all work on Gaussians within Skyline's software. (discussed at 12:02) ### Are Gaussian splats just for visualization? No, according to Ronnie Yaron: Gaussians are basically a point cloud, and Skyline extracts a point cloud from them that is better than regular multi-view stereo on thin wires and vegetation. (discussed at 10:58) ### What does scale mean in 3D geospatial processing? According to Ronnie Yaron, scale is the amount of data processed, not the size of the area: millions of images of a single cell tower count as much as a whole city. (discussed at 5:04) ### How can drone images without GPS metadata be georeferenced? According to Ronnie Yaron, Skyline developed automated referencing that matches such images to existing maps of the area to get accurate products. (discussed at 4:17) ### Is Skyline Software Systems VC funded? No, according to Ronnie Yaron: Skyline was funded initially, then bought out its VCs and went fully private again; it is profitable with about 70 people. (discussed at 21:42) ## Quotes > “Whoever is saying that Gaussians are just for visuals, I don't think that's correct.” – Ronnie Yaron (11:23) > “Basically the Gaussians are photogrammetry.” – Ronnie Yaron (11:31) > “Scale, when I say scale, I refer to the amount of data processed.” – Ronnie Yaron (5:06) > “We have projects that are running on thousands of computers simultaneously.” – Ronnie Yaron (5:41) > “We bought out the VCs that owned the company.” – Ronnie Yaron (22:01) ## Chapters - 0:00 Meeting Skyline at INTERGEO - 0:55 3D geospatial over 48k modems - 2:04 One pipeline from capture to analysis - 3:23 Hours, not weeks, and GPS-denied data - 4:32 Scale means data, not area - 6:05 From voxel games and Snow Crash to Skyline - 6:58 Gaussian splats from day one - 8:29 Splats shaped as surfaces, sharper edges - 9:33 Why Gaussians are not voxels - 10:58 Gaussians are a point cloud you can measure - 12:20 Navigating from video alone - 14:10 AI needs to know where everything is - 15:36 What does it mean for AI to understand? - 17:57 Asking 3D models smarter questions - 19:01 A startup mentality after 30 years - 19:50 The OGC standard that died - 20:37 70 people, profitable, no VCs - 22:38 Wrap-up ## Transcript _Automatic transcription (ElevenLabs Scribe), lightly edited; times follow the audio._ **Michal Gula** (0:00): It's funny how the world works, because, we have been trying to get in touch with Skyline for quite some time. I, I mean, not only like from the business, position, but more like some marketing collaboration or our 3DISE conference and so on. But luckily, we were able to get in touch directly with the men... from the Skyline, at the opening event at, INTERGEO. So we're happy to have you here in the podcast, and we are looking forward to discuss what you guys do at, at Skyline. **Ronnie Yaron** (0:36): Fantastic. I've been- **Michal Gula** (0:37): Ronnie Yaron, welcome here. **Ronnie Yaron** (0:39): Thank you very much. I've been personally following you for a long time, so-... it's great to finally, uh- **Michal Gula** (0:44): That's nice **Ronnie Yaron** (0:45): meet and talk. **Michal Gula** (0:46): Thank you. Thank you very much. can you like briefly tell us what Skyline is trying to solve? What problem is, is it trying to solve? **Ronnie Yaron** (0:55): So Skyline is not a young company. We've been around for more than three decades. We started when the, internet was not really there. We were working on modems that were 48 kilobits to stream 3D data- **Michal Gula** (1:10): Nice **Ronnie Yaron** (1:10): before people even knew what 3D was, you know- **Michal Gula** (1:13): Yeah **Ronnie Yaron** (1:13): except for the big silicon graphics, machines-... and Cray computers. so Skyline brings 3D to ge- geospatial from the beginning- **Michal Gula** (1:25): Mm **Ronnie Yaron** (1:25): when before 3D was cool and everyone was doing it. our mission is to solve real-life problems, from anywhere from like defense customers to commercial markets, from the data collection, production, dissemination, down to visualization, analysis, and implementation into applications. So that's what we're all about. **Michal Gula** (1:50): Does it mean that, I can stay with Skyline like from the moment I capture pictures to the moment I need to do the final decision on, on some projects? **Ronnie Yaron** (2:02): Absolutely. **Michal Gula** (2:03): Sure. **Ronnie Yaron** (2:04): That's... Part of our ability is to actually solve the problem end to end, from the collection down to the analysis of whatever you need with the, data, whether it's a 3D mesh, point cloud, Gaussian splats, and anything in between. **Tomáš Barnáš** (2:20): I saw first time Skyline solutions, I would say 10 years back at INTERGEO as well. And I must say, like that time I was extremely s- like impressed by everything what you can do because... And that's 10 years, not 30 years back- **Michal Gula** (2:34): Yeah **Tomáš Barnáš** (2:35): right? Because today, like there are a lot of companies which are tr- trying and doing like viewers and simulations, but even 10 years back, I would say like you were one of those which had l- had already like super robust solutions. And I was, at that time talking with Tomas Orlik, one of- your directors. **Ronnie Yaron** (2:54): Yeah. **Tomáš Barnáš** (2:54): he was also from Slovakia, as we too are, so that was kind of a nice, talk because we didn't know in the first moment. We talked like half an hour in English. Then we figured out-... that maybe we can talk the same language. And, yeah, I just wanted to say that it was like... I was very impressed by seeing like all the functions, options, simulations, a lot of also for military, and, and, visual views of, how you can move through the city. So there are a lot of, yeah. **Ronnie Yaron** (3:23): Yeah. So part of what we do very well is scalability. you can see a lot of people doing 3D, but to do it in a large scale and at speeds that are counted in hours rather than weeks or months for data production, that's where we excel, and that's why we're focusing on customers that have very specific requirements for doing things, not just on a massive scale, but also very, very quickly. and there are some unique capabilities that we develop, by catering to our defense customers, like working in GPS-denied environments where you can't count that you'll have metadata along with the, drone images every time. In fact, you can count that you probably won't, and even if you do, it's not accurate and you can't count on it. So we develop referencing capabilities that are automated, and allow you to bring in data that doesn't have any information, but still get very accurate, products by referencing it to existing maps of the area, for instance. **Michal Gula** (4:32): When people are speaking about the scale, they are usually referring to whole cities or whole states or even continents and, and so on, right? But in the, in the core it might actually not be about the scale in terms of the size- **Ronnie Yaron** (4:50): Mm **Michal Gula** (4:50): but in terms of the detail, right? Because in theory I, I can digitize this tree- **Ronnie Yaron** (4:55): Right **Michal Gula** (4:55): with more polygons- **Ronnie Yaron** (4:57): Could be sub-millimeter **Michal Gula** (4:58): than, than the country. Yeah, exactly **Ronnie Yaron** (4:59): it could be sub-millimeter. **Michal Gula** (5:00): Exactly. **Ronnie Yaron** (5:00): It could be... Yeah. **Michal Gula** (5:01): So is this the real, definition of the scale? **Ronnie Yaron** (5:04): So that's a very good point. Scale, when I say scale, I refer to the amount of data processed. So if you have millions of images, it doesn't matter if you're, at sub-millimeter resolutions and you're just doing a cell tower- **Michal Gula** (5:17): Yeah **Ronnie Yaron** (5:17): very high precise, engineering style, or if you're doing an entire country or city at very fine resolutions. But dealing with vast amount of, information, develop- we developed a pipeline that actually cuts all the, pipeline into sub-problems that we can disseminate and distribute across multiple, machines. we have projects that are running on thousands of computers simultaneously. so a single project could be distributed almost endlessly, whether it's on premise or in the cloud or a hybrid approach that we developed that allows to divide between on-premise and cloud resources, to maximize, the speed and the scalability. **Tomáš Barnáš** (6:05): Was your target customer base changing over those years? It's pretty unique for me to speak with someone who, who has a company or a software company which lasts over thir- three decades. So, h- how is that like the customer base? **Ronnie Yaron** (6:19): So when I started, I was actually, developing computer games and the engines for the computer games. before NVIDIA or any kind of 3D cards, we were doing it voxel-based 3D, and that's how we developed first Skyline. My idea came from a book called Snow Crash, where the protagonist had an Earth that they could zoom in and actually look from space down to very high details, and I said, "Well, we- we're doing that in the games. Why don't we do it for real, with the Earth?" And that's where we got some old Landsat data, and we developed an Earth that you could actually zoom in from space down to the highest detail that we had back then. so... **Michal Gula** (6:58): You- Yeah... y- you already mentioned the, the 3D Gaussian splats, being implemented in your workflows. **Ronnie Yaron** (7:04): Yes. **Michal Gula** (7:05): we just had a panel discussion about that, and I was asking one question, and I'm wondering also about your answer. was it a tough decision? When, when you found out that you saw the first papers, the first video, you, you certainly got interested because it looks nice. what was the decision-making? When did you say that like, "Okay, guys, we have to do this"? **Ronnie Yaron** (7:27): The day I saw it first. So the-- as, as the first article came out, I knew that this was the, direction. We started developing splats back then. This is, more than a couple of years ago already. Mm. it took a while to implement it throughout the entire pipeline. It's not just the, processing of the data in a scalable way, but it's also, the visualization and the ability to do more than visualization, also analysis within our tools on the Gaussians. But there's no competition, for some, some kind of, models with, the thin structures and vegetation and, the many advantages for certain, types of, data. Today, we're actually combining the best of both worlds. So for flat surfaces, using mesh-like techniques, and for the, thin structures, vegetation, the, the splats, and, doing it in a way that basically provides a product that is the best of, both worlds. **Michal Gula** (8:29): But are you also displaying it, in a some kind of a hybrid mode, like both at the same time? **Ronnie Yaron** (8:35): No, actually the display is, uh-- But yeah, we, we can combine meshes and splats at the same time. but we use different techniques to shape the splats into the form that we want. So for a surface, they're shaped as a surface. It looks indistinguishable from a mesh, basically. **Michal Gula** (8:52): do you use it mainly for surfaces or more like for edges? Because we, we know that with the splats always edges are the problem to-- Especially when you're zooming in. **Ronnie Yaron** (9:02): The, the, there are multiple problems that, yes, we're doing a combination of techniques to get the best of both, worlds. It's,.. So even with the meshes, we have special techniques that actually, sharpen the edges of any kind of like buildings or structures. so even in the old techniques of, doing the, 3D mesh, we still add our secret sauce to make it, look, better than, some of the, competition. **Tomáš Barnáš** (9:33): Could you say that, using-- Because you started with vox, voxels, you mentioned- Yes... like back then. Could-- Was that kind of like a good base also for Gaussians? Because on the end, we're like, if we want, we can say like voxels are also like point turned into squares, flats, something, and Gaussians are also like, splats pro- like projected over points. 3D object in a space, basically, right? So- **Ronnie Yaron** (9:56): It's... Yeah, but the, the, the Gaussians uses a deception technique that artists have used for, you know, the hundreds of years already by, you know, having something look photorealistic, where actually it's made out of a lot of, small dots that, when you look close up, you see that it's, it's nothing close to reality. but it's the merging or mixing of these, alpha of the Gaussians that creates the, illusion. That's why it's different from voxels, which are strictly boxes of a unit. So that doesn't look right. The voxels will always look kind of, not dirty, but I would say not the smooth that you get with the Gaussians and the photorealistic. **Michal Gula** (10:42): So- Because the, the-- still, like voxel is a different representation of the space compared to point cloud or, or a 3D mesh. **Ronnie Yaron** (10:49): Yeah. **Michal Gula** (10:49): But the, the, still, like main goal is to represent geometry. With the Gaussian splats, the goal is to represent the visual side of the scene. **Ronnie Yaron** (10:58): So what we find is Gaussians basically are a point cloud. And today we're using techniques to extract a fine detailed point cloud from the Gaussians that is way better than what we get from the, regular multi-view stereo techniques, especially for thin wires or, very delicate structures. so whoever is saying that Gaussians are just for visuals, I don't think that's correct. I think the technique that we're using, basically the Gaussians are photogrammetry. at the end of the day, we're mixing the two disciplines. So we're starting with the, photogrammetry correlated point cloud, and we're doing a special technique that then takes that point cloud and uses the Gaussian technique to, A, pop out all the thin structures and poles and the wires and things like that, and vegetation, but also at the same time preserve the accuracy of the photogrammetry, from the correlation. So **Michal Gula** (12:02): when you represent in, Gaussian splats, like wires or, or leaves on, on, on trees, are you actually able to measure on them? **Ronnie Yaron** (12:11): Yes. Can measure on them, you can do line of sight or view shed analysis. All of those tools, work within our software. **Michal Gula** (12:20): Okay. we, we have-- Like in last few years, we have two, hype words. one we already solved, the, the Gaussian splats, and the, the other is AI. How, how are you working with AI? **Ronnie Yaron** (12:35): Well, well, let, let me tell you what we're working on, next. This is, uh- So visualization is something that a lot of people are solving these days. we're taking it to the next level where we're using the structure of the 3D that we're creating for... not just for humans to visualize or do the analysis, but also for the machines. And what I mean by that is that now you can actually create a 3D model from a drone or an aerial photography. We can allow you to navigate on the ground just based on video. So point a video camera, we'll tell you where you are exactly, what you're looking at exactly. and this has multiple uses for, you can imagine, for many types of, applications. a lot of them are in defense at the moment. **Michal Gula** (13:23): You started, you started with us speaking about the drone scan, but this is actually useful for especially indoor navigation, right? **Ronnie Yaron** (13:30): That too. But the challenge is having different modalities, for instance, taking imagery from the air and align, doing the alignment from the ground. This is a very difficult task for regular matching. It doesn't work. what we are doing is we're generating basically features from the 3D, model that allows you to do things from different angles that were not in the original, images or how it was created. So this allows for something where you can collect data from the air and do matching from the ground, which is a hard problem to solve without the 3D. **Michal Gula** (14:10): But my question was about AI. **Ronnie Yaron** (14:12): So AI, in order to function in a real world, needs to know where everything is. And by allowing the AI that analyzes the images to correlate different images that are looking at the same object and know that they're looking at the same object, this is where we come in. so we are... yes, we are doing a lot of, research and development in that, area. but the important thing is to allow not just, taking, AI that, is doing it over separate images, but an AI that actually understands what it's looking at and knows that it's looking at the same object from multiple, angles, which- I was- **Michal Gula** (14:59): Th- this is something what I've been saying, like, already for, for quite some time, that because people are, are saying that they are afraid of AI, and I'm always saying like, "I'm not afraid that AI will know everyth- everything, because we already have Google." It knows everything, right? what I'm afraid, quote, unquote, is that AI will understand everything, and that's, that's a very different thing compared to knowing everything. we are going to, to times when AI will see the video or image or anything el- and it will understand what's happening. **Ronnie Yaron** (15:36): Yes. We're already there, but understand is a word that you need to define. By understand, the way I see it, is the AI essentially forms a hierarchy of clusters of information, much like we do in our, brain. and this is the understanding. It- It's classification in a hyper scale, multidimensional way, that allows it to infer and to, generate, I would say, more intelligent, information, on the, data it's looking at. **Michal Gula** (16:15): For me, understanding is, knowing that what's actually happening on the image or the video, and what consequences it will have. **Ronnie Yaron** (16:26): Yes. **Michal Gula** (16:26): So, for example, and I don't remember who I was speaking about this to, but, when you have a picture of a person touching the glass towards the edge, the AI will know what will happen, that the glass at the end will fall. **Ronnie Yaron** (16:40): That it will fall. Yeah. **Michal Gula** (16:41): This is understanding, and this is, bringing us to like,.. You have seen the movie Minority Report? **Ronnie Yaron** (16:48): Of course. **Michal Gula** (16:48): Where it, where it was predicting the, the future. Are we getting there? **Ronnie Yaron** (16:52): Predicting the future is a bit, more difficult because of- **Michal Gula** (16:55): I- I- I know, like **Ronnie Yaron** (16:56): but to a degree that the glass will fall, that's an easy prediction. This is physics. beyond a certain point, chaos reigns in, and it's very hard to, to predict because a- **Michal Gula** (17:07): But AI- **Ronnie Yaron** (17:08): the butterfly, flaps it- its wings, and a, a- **Michal Gula** (17:11): Exactly **Ronnie Yaron** (17:11): thunderstorm forms, but- **Michal Gula** (17:11): But you have another variable. You have the intention also. **Ronnie Yaron** (17:14): Yeah. **Michal Gula** (17:14): So, m- maybe for the future, AI will not only know what will happen if the glass will go to the edge, but also from the face expression or other movements, it will be able to, to, to predict the intention of the person, whether- **Ronnie Yaron** (17:29): Yes **Michal Gula** (17:29): he is or not going to do. **Ronnie Yaron** (17:31): That's the lip reading in, HAL, HAL 9000-... in, Space Odyssey, where the AI lip reads the, the intentions and it understands. **Michal Gula** (17:40): Yeah. **Ronnie Yaron** (17:41): Yes, and the answer is of course. This is what AI is excellent at, is, having almost like a human or superhuman intuition about things just by detecting patterns and analyzing how they will, influence other things. So the answer is definitely yes, and we are looking at augmenting the 3D with embeddings that allows you to ask intelligent questions and get answers, not just "Find me the red cars," but, "Find me something that, there's a red car parked red to a-- next to a restaurant," or something which is- with more, I would say, logic behind it. so the answer is yes, this will get very sophisticated very quickly, and it is already. So the, the question is whether all of us will have a, a job in a few years... or we'll just keep doing it for fun, which is what I think will happen. **Tomáš Barnáš** (18:37): We can see that you are still deep in development of, of the software and everything because, you're loaded with all of this information. It's great to listen. **Ronnie Yaron** (18:45): My, my passion is technology, and, you know, I, I try to stay, on board with everything that's happening in that, domain, and a lot of it to implement where it applies, with what we're doing at, Skyline. **Michal Gula** (19:01): What is the biggest challenge for you for the future with Skyline? **Ronnie Yaron** (19:05): It's keeping our, company at the forefront of, the technology and at what we do, to be the best at what we do, and to keep making, making money from it. **Michal Gula** (19:17): When, when I asked you about the Gaussian Splats and when did you actually decide to, to do it, you, you said, like, right away. **Ronnie Yaron** (19:23): Yeah. **Michal Gula** (19:24): Does it mean that you are not afraid to, to take risks? Because after th- 30 years you are, you are not a startup. you're not, you're- **Ronnie Yaron** (19:32): No, but we have a startup mentality. **Michal Gula** (19:34): That's what I was- **Ronnie Yaron** (19:34): We, we have a startup mentality- **Michal Gula** (19:36): asking **Ronnie Yaron** (19:36): as far as staying ahead of the curve in the technologies, the bleeding edge of the technologies. We try a lot of things. We don't always succeed, but in many cases, we do manage to be the first ones, to do things and to win the business because of that. **Michal Gula** (19:50): Please do tell us, some, some, sorry, s- some mistake that you did. Because when you are taking risks, there is always some mistake or some failure. So can you share with us some? **Ronnie Yaron** (20:00): There are many of these, along the way, of course. in the, in the early days, we developed a service that allows, through an API to get a 3D image on the server. So you would have the 3D model on the server, and we wanted to have it browser-based. This was early 2000. Uh- **Tomáš Barnáš** (20:20): And clever guys were stealing data **Ronnie Yaron** (20:21): We, we actually, made an OGC standard from this. **Tomáš Barnáš** (20:25): Okay. **Ronnie Yaron** (20:26): But it never caught on, and it, it kind of died. So- it's a, it's an example. There are a few of these along the way that you invest time and energy, but it doesn't pan out, of course. **Tomáš Barnáš** (20:37): How big is the company right now from a employee? **Ronnie Yaron** (20:42): We're about 70 people in the company. **Tomáš Barnáš** (20:44): Okay. I just wanted to, mention this with like, where you can still, handle and say like, "We have the startup, thinking," because- **Michal Gula** (20:53): Mentality, yeah **Tomáš Barnáš** (20:54): startup mentality. **Ronnie Yaron** (20:55): Yeah. **Tomáš Barnáš** (20:55): Because sometimes you just see, companies with 20, 30 people, and they already say it's like, "We have to have structures." So- **Michal Gula** (21:02): But it's coming from the top. Like, it's not about the mentality of the company, but in reality it's about the mentality of the management, right? **Ronnie Yaron** (21:10): I think so. yeah. I mean, I'm- **Michal Gula** (21:12): Because if, if you didn't have this mentality- **Ronnie Yaron** (21:15): Yeah **Michal Gula** (21:15): you wouldn't expect it from employees. **Ronnie Yaron** (21:18): No. I mean, I'm, I'm managing the company, and it's, basically setting the, direction. I have, two partners in the, the company that are not active in the, management. but, you know, I, I enjoy very much what I do. It's my, my passion, and I get to play around with these new technologies and having fun at it and making it into a business. so far, very successfully over the last few decades. So a profitable company. We don't have any VCs or anyone to report to. **Tomáš Barnáš** (21:47): Okay. **Ronnie Yaron** (21:48): we're having fun in our little corner. **Tomáš Barnáš** (21:50): Nice. **Ronnie Yaron** (21:50): And, one day maybe we'll be, uh- **Tomáš Barnáš** (21:54): This is- **Ronnie Yaron** (21:54): a lot bigger than what we are now **Tomáš Barnáš** (21:55): this is quite unique on the, on the market, right? **Ronnie Yaron** (21:57): It is very unique. **Tomáš Barnáš** (21:57): Not to be funded. **Ronnie Yaron** (21:58): I think, I think so. We were funded initially. We bought out the, the VCs that owned the company. **Tomáš Barnáš** (22:04): Oh, nice. **Ronnie Yaron** (22:05): And went fully private again. **Tomáš Barnáš** (22:08): Nice. **Ronnie Yaron** (22:08): Yeah. **Tomáš Barnáš** (22:09): Congratulations to that. I mean, it sounds amazing. **Michal Gula** (22:11): That, that's also a thing that allows you to, to have a startup mentality because if you had like 30, 40, 50% of investor there- **Ronnie Yaron** (22:20): Yeah **Michal Gula** (22:20): they wouldn't allow you to do- **Ronnie Yaron** (22:21): They will set the goals, and they will- **Michal Gula** (22:23): Yeah. **Ronnie Yaron** (22:23): Yeah. **Michal Gula** (22:23): Yeah. **Ronnie Yaron** (22:23): Yeah. So it's, but that, that's me. You know, I'm, kind of an innovator at heart. I always try to optimize processes and to do things that no one has done before. and I enjoy it, so that's what I try to do. **Tomáš Barnáš** (22:38): Great. **Michal Gula** (22:38): Perfect. Thank you very much for the conversation. **Ronnie Yaron** (22:40): Thank you. Thank you for hosting me. It's fun. Yeah. **Tomáš Barnáš** (22:43): Good luck.