Real-Time 3D and AI Visualization — with DGG and Threedy | AI Across The Product Lifecycle | DemystifyingPLM

Episode Summary

The episode titled "Real-Time 3D and AI Visualization — with DGG and Threedy" delves into the integration of artificial intelligence (AI) and real-time 3D visualization in product lifecycle management (PLM). The guests are Christian Stein from Threedy, a platform that enables instant 3D collaboration and visualization, and Max Limper from DGG, creators of Rapid Pipeline software for efficient 3D rendering. Both companies have roots in the Fraunhofer Institute for Computer Graphics Research, where they initially worked on standardizing 3D formats and developing advanced renderers for web and virtual reality applications.

The discussion highlights key insights into AI's role in PLM. Christian Stein emphasizes that while AI can automate certain tasks, it lacks the contextual understanding needed to make strategic decisions. Max Limper agrees, stressing the importance of human engineers who can creatively tweak parameters to optimize processes. Both guests underscore the necessity for developers and product managers to think beyond immediate tasks and consider long-term impacts.

For PLM and engineering professionals, the key takeaway is that while AI tools can enhance efficiency in 3D visualization and rendering, they cannot replace the strategic thinking and creativity of human engineers. Professionals should focus on developing a deep understanding of their processes and the broader context in which they operate to leverage AI effectively while maintaining control over critical decisions.

Full Transcript

Michael Finocchiaro

And we're live. Welcome to the AI Across the Product Life Cycle podcast. I'm very happy to welcome Christian Stein of Threedy and Max Limper of DGG. Thanks guys for showing up today. This is great. I'm looking forward to learning more. ⁓ We're going to talk about, as usual on this podcast, about AI. But I think in this case, more on how Christian and Max are thinking about AI in the context of 3D than actually using AI and 3D together today. And I'm looking forward to it. So why don't you introduce yourself, Christian.

Christian Stein

⁓ Thanks Michael, happy to be here. ⁓ Maybe a little bit of myself and my background. I'm a computer scientist focused on computer graphics software engineering. Initially I worked in the games industry at Electronic Arts for some years before working as a freelancer and doing my masters and then by chance joined the Fraunhofer Institute for Applied Computer Graphics Research in 2012. This is actually when I met Max for the first time. Maybe you can elaborate a little bit on that as well. ⁓ For the next years we have jointly worked on what today is our platform called Instant3D Hub. We went through a long spin-off cycle to get this moved out of Fraunhofer and now for five years we are with 3D on the road as an independent startup ⁓ commercializing the solution and developing it and happy to talk a lot about the learnings and things we're doing there later on.

Michael Finocchiaro

Awesome. You guys are actually both in the same city, aren't you? Darmstadt, yeah. How about you, Max?

Christian Stein

Yes.

Max Limper (DGG RapidPipeline)

Well, yeah, I mean, first of all, thanks, Michael, for having me here. yeah, I'm Max. I am CEO and co-founder at DGG, creators of the software called Rapid Pipeline. And our software has been around for, like the startup has been around for seven years now. We're also spin out of the Fraunhofer Institute for Computer Graphics, which is where I met Christian as well in 2012. And we started working together on things like standardizing 3D formats and

Michael Finocchiaro

here now, we're also spin out of the Brawn Ripper Institute for Computer Graphics, which is where I met with working as well. And we started working together on standardizing formats and

Max Limper (DGG RapidPipeline)

⁓ like working on efficient renderers for the web and so on. My focus back then was more on automating pipelines and that's where I did my PhD also, like specifically

Michael Finocchiaro

working on the renderer for the web and so on. My focus back then was more on automating pipelines and that's where I did my PhD also.

Max Limper (DGG RapidPipeline)

about geometry processing and optimizing data for rendering. So like fast rendering in the browser and VR and so on. So yeah, that's my ⁓ quick background basically.

Michael Finocchiaro

Nice. Yeah, I think that when I've analyzed the history of CAD and PLM, I've focused on the two Cambridges, right? The UK and Cambridge Mass at MIT. But I think actually Darmstadt also has a place, a large place in the history of CAD, right?

Max Limper (DGG RapidPipeline)

Yeah, I mean, I don't know too much about this particular example, but it's just something that comes to my head is like, ⁓ VRED, Autodesk VRED, the roots of that I think are also somewhere like with the front of IGD, connected or the predecessor of that here. There's a long history of computer graphics research. And I think Christian, you also have some contacts, right? And post-app and things like that.

Michael Finocchiaro

Right.

Christian Stein

Yeah, think in general the Fraunhofer Institute has roots for now maybe 50 years. I think there was very recently a celebratory event about computer graphics ⁓ birth in Germany, which is maybe more in the Berlin area, but then... ⁓ important people moving to Darmstadt, building the roots here for a larger team, the Fraunhofer IGD itself and Christ, the university group that it was closely collaborating with and how it extended its reach with different entities across Europe, with the previous head of the institute, Professor Encarnassau, who is also a laureate in computer graphics area. So there's a big history on that, but then I think ⁓ what ⁓

Michael Finocchiaro

So there's a big history on that. then I think what's

Christian Stein

very specific about the Institute and also about the work that Max has been doing there is that for the last 25 years we have closely worked with industry on how to standardize computer graphics and formats for various use cases and

Michael Finocchiaro

very specific about it, even also about the work that we've been doing there, is that for the last 25 years we have been working with the industry on how to standardize the graphics for various use cases.

Christian Stein

⁓ how especially with the automotive where there's a say a hub in Darmstadt about collaborating on 3D formats. ⁓ how this joint efforts over the last 20 years have closely influenced ⁓ how 3D graphics is used in the German automotive industry today.

Michael Finocchiaro

Yeah, it's so juicy because we don't it's not the first thing comes to mind when you think of the history of 3d but actually even for PLM because of first Martin Eigner is one of the grandfather of the PDM so it's cool to think about that's cool for Germany. I wanted to you know our topic is around AI and I know that you guys said that you're not specifically using AI in your software, but obviously you're thinking about it like everybody else. So I was wondering, my first question is always, ⁓ we're now into like year three of the LLM revolution, right? Ever since OpenAI gave us the original chat GPT, which looked like a toy, but some of us saw that there was quite a lot of potential there and now it's taken on a whole nother ⁓ dimension. Were you guys really bullish on it or were you guys a bit skeptical back in two, three years ago?

Christian Stein

Do you want to start? I can start. Actually, I always love to tell how I experienced this on a more personal or social level because let's say my personal network is a lot of computer scientists and a lot of teachers. And when it arrived, it was like huge enthusiasm from the computer scientists about this is going to change everything so fast. And the teachers were shying completely away like, ⁓

Max Limper (DGG RapidPipeline)

Go ahead, yeah.

Christian Stein

We cannot use it in our personal jobs at all. And then one and a half years, two years later, you have to quite the opposite situation because the computer scientists suddenly got a lot of hesitation about, it really reliable? How much can we rely on it in our work?

Michael Finocchiaro

We can go. Right.

Christian Stein

⁓ while the teachers have been the opposite now and they were like, hey, we're doing it every day. We are changing completely how we work with our students with this. So I guess from a company perspective, we have been ⁓ initially a little bit trying a way that we started to... ⁓ see how our internal processes could be optimized, how to use the technology, let's say to accelerate ⁓ daily work more efficient. But at the same time, at our customers, didn't really see them in the engineering and IT space where we are mostly active, really starting to work with it and really getting things into a productive situation. But now I think this has changed over the last maybe ⁓ nine months.

Michael Finocchiaro

same time, our customers, didn't really see them.

Christian Stein

that more regularly you get in touch with AI initiatives and there are more concrete integration topics for us. And now I think over the last month it has really strongly influenced our roadmap and we are happy to talk about the details later on what we're doing there.

Michael Finocchiaro

Awesome. So it's more of a, in your case, it's more of a pull than a push. In words, you're waiting for the industry to pull you in and say, we really need it. That's interesting. How about you, Max? Did you have the same thing with Revit Pipeline?

Max Limper (DGG RapidPipeline)

Yeah, good question. mean, when it first came up, I was, I wouldn't say skeptical. I there are many people who were like extremely hyped. And I think that's even like a small industry in itself. It's like these people who post it every week, like, yeah, I mean, like, it's like this pattern that you can probably write with an AI and ask, like, what happened last week in AI is wild. pointer down, right? It's like,

Michael Finocchiaro

The AI bros.

Max Limper (DGG RapidPipeline)

pure engagement page, right? This is like where I tend to throw up if I read too much of this, ⁓ I don't even click it because like it whatever right? So this like as an engineer, this this really turns me off. But then of course, but I don't know if that's skepticism, right? I mean, if you're not buying, like, overhyped stuff, is that already skepticism? Or is that just being a normal, rational person, right? Right.

Michael Finocchiaro

I think it's more rational, but that's just me.

Max Limper (DGG RapidPipeline)

So that's taken away. ⁓ I I found it quite interesting. There has been deep learning in computer graphics already for much longer time, right? Like 10 years ago, I think around 10 years ago, there were a lot of deep learning papers popping up in SIGGRAPH, like the major conference where people published technical papers in computer graphics. And I think outside the computer graphics community, no one noticed that, right?

Michael Finocchiaro

A long time.

Max Limper (DGG RapidPipeline)

then I think the big AI hype is just, hey, chat GBT, like people can have the feeling that they can talk to something and that something is smart, right? Like ⁓ I had this moment where I was like, that's actually kind of cool where I asked her to write a poem about a character from an animated show when we were sitting around like at ⁓ US Eve somewhere. It was writing it and it was actually fitting that character, right? ⁓

Michael Finocchiaro

I AI is just, HHGBT, like people can have the feeling that they can talk to something and that's something that's smart, right? Like that, I had this moment where, I thought that's actually kind of cool where I asked them to write a poem about a character from an animated show when we're sitting in a round like that. And we'll see if someone was writing it and it was actually fitting that character, right?

Max Limper (DGG RapidPipeline)

That's kind funny. So when you have these moments, then ⁓ you start thinking about what could this do, right? But at the same time, my engineering brain wants to really know what are the limitations? How does it do that? Can I make it? It's like a bit of a troll that is in every engineer also, right? Can I do something stupid and expose? And it's like when you start asking it, I don't know, give me a word that has ⁓

Michael Finocchiaro

Right. ⁓

Max Limper (DGG RapidPipeline)

two times the letter E and is whatever nine letters long or sold for Sudoku or whatever. it just explodes, right? Just cannot do it and can't. Or it says, yeah, like two and two, what is that? And it says, it's four. And said, that's wrong. And it'd say, I apologize, it's five. ⁓ So that's the fun stuff then. But I think it's really interesting how AI is...

Michael Finocchiaro

I apologize.

Max Limper (DGG RapidPipeline)

like what we call AI now mostly, right? It's like LLMs and anything like that is very fuzzy. And that fuzziness can be extremely powerful, right? I mean, we have our algorithms and we like to solve very specific, well-defined problems with an algorithm that gives you an optimal solution. And this is really what we have been also doing with rapid pipeline usually, right? I have an input, I have a desired spec for the output and our software produces that and boom. ⁓ But sometimes I have more fuzzy ⁓ requirements. And that's where AI can really shine, right? Because that's where you usually need a human in the loop. And we can talk about that later, I guess. I don't want to go into rabbit hole. But I think that it's basically, in a nutshell, AI allows us to do things that were previously not possible, but many people don't understand.

Michael Finocchiaro

And that's where AI can really shine, right? Because that's where you usually meet a human in the room. we can talk about that later, I guess. don't want to go into any details. But basically, as a nutshell, AI allows us to do things that were previously not possible, but many people don't understand,

Max Limper (DGG RapidPipeline)

especially in the beginning, what these things are.

Michael Finocchiaro

especially in the beginning, what these things are.

Max Limper (DGG RapidPipeline)

And I it can do everything or nothing, but it's really interesting how we as a community all figure out where AI has its place already now and where it might have its place in the future. And the answer to that is not trivial. that's why, right? You cannot just say all the algorithms are going to replace by AI. That's ⁓ BS.

Michael Finocchiaro

No. Not happening. But so let's talk ⁓ more specifically ⁓ because both of you guys are developers. ⁓ of course, some of the hype of the AI bros has been on vibe coding and, you can build your whole app without actually coding. Well, I think we all agree that's absolutely not the case. You still have a need for programmers for the details and things like that. But how are you using AI from a developer point of view at 3D? to improve or to do the scaffolding or maybe bench stemming? are you using AI on a daily basis as a developer? How are your developers using it at 3D?

Christian Stein

⁓ Yeah, maybe I give a few minutes background on what we're actually building. idea of our software is this like add-on harmonization layer that customers put on top of their data management system, PDM, PLM, on their 3D data formats. And we can harmonize across all of that and provide them with APIs to pull in certain kinds of computation by just pointing towards data using...

Michael Finocchiaro

Yeah, that's fine.

Christian Stein

linked data standards. with that, over the years now in the company, we have grown in certain customers along the product lifecycle or across departments in a way that we kind of ⁓ take care of the 3D part in their digital threat. So ⁓ different departments are being capable of actually talking about the same data because not everyone stores and manages their own truth, but they're kind of agreeing. with a standardized namespace and addressing scheme about how data is addressed and then they can just pull it into any kind of typically web-based application. ⁓ so at the same time for us, I think it's been always a challenge like who's the person we're selling to and who are the people working with the platform. Because ⁓ initially, obviously, you have to solve a certain use case and do like vertical integration so that things are up and running. But afterwards,

Michael Finocchiaro

Initially, obviously, have to solve the service case, you have to do integration so that things are running, but afterwards,

Christian Stein

The most important persona maybe is the product owner or the developer that uses our APIs to integrate the solution in any kind of workflow where 3D data can be beneficial more or less.

Michael Finocchiaro

the most important thing is to use our APIs to integrate the solution. Any kind of workflow where the data can be or an issue. And now if we look internally, it's working with AI. I initially, we...

Christian Stein

And now if we look internally how we've been working with AI, think initially ⁓ we really solved one of the core problems regarding information management in the company that we had with this history of 15 years almost now developing this technology and the people coming in at different points in time, people from the Fraunhofer time, new developers. ⁓ access to the information that was in the head of the people or that was stored in some internal system was already difficult. And there we saw that this can really be helpful to get people up to speed more quickly. And then we had lot of initial initiatives about using it in ⁓ testing, automating certain tasks in the development process, in writing PRDs, for example. So different

Michael Finocchiaro

. All right.

Christian Stein

people, different departments really building their own tooling based on AI to accelerate what they're doing in the daily work. And then what you mentioned, obviously over the last six months, you can just wipe code anything so quickly. And ⁓ for us, this is a huge opportunity on one side to move forward building MVPs very quickly to have something when we approach this initial use case with a new customer, for example, that you can go through all almost finished version of a vibe-coded app with some 3D context integrated, where it seems like the solution is already there and you can buy that. give a really good feeling to the customer about how the solution is going to look at the end, while at the same time you still have to do the hard work afterwards to implement it in a sustainable way. But at the same time, I think one of the major benefits once you have integrated our software is that if you consider the whole technology stack of industrial 3D application that's built with our technology, then we can kind of cover 95 % maybe of the complexity

Michael Finocchiaro

. to implement it in a sustainable way. At the same time, think one of the major benefits once we integrated course on the web is that if you consider the whole technology stack of a just built-in application that's built in is called technology, then we can kind of cover.

Christian Stein

with a standard product and all the systems that run under it. And where the AI is really strong because it's been trained on the internet more or less is exactly this ⁓ HTML5 JavaScript application layer that people then use to build their applications. So for us, we obviously also see a chance to what we call like, yeah, customized. for the individual person, the individual workflow applications that we can streamline the process of developing these tools which in the past have been kind of tedious because you have to find a development team internally or at the customer, you need a product owner, you need to build an infrastructure to serve the application to connect it to all the different data backgrounds. So if you can automate so many things on that and then you use ⁓ LLMs on top to really quickly ⁓ generate and accelerate the implementation of just the user front ends. ⁓ think it's going to, yeah, in the next years we will see that people will be able to pull out a user interface on demand for the individual tasks they are doing and vision maybe that we already had 10 years ago that it's going to be about ⁓

Michael Finocchiaro

I think it's for the next years we will see that people will be able to put out the user interface on demand for the

Christian Stein

lots about a swarm of minimal applications that just provide you with the

Michael Finocchiaro

individual task. Mm-hmm.

Christian Stein

individual interactions instead of a monolithic check of all trades application that is rolled out to many users. I think there will be a shift and this is where we see a real opportunity with our approach.

Michael Finocchiaro

Interesting. Max, do have a different perspective?

Max Limper (DGG RapidPipeline)

⁓ I mean, from our company's point of view, maybe slightly different focus here and there, but in general, I would echo ⁓ everything that Christian said. think especially this idea of basically shortening that cycle from an idea to a mock-up or even a functional UI for something very specific, I think that already works very well. ⁓ mean, it's funny, like you could say AI is basically helping you to search and put together pieces what developers would do manually before by just browsing Stack Overflow. Now we all know AI has probably scraped Stack Overflow and a zillion other sites. And yeah, no matter if that was all legal or not, but what they offer now is just, yeah, it's not going to go away anymore, right?

Michael Finocchiaro

Yeah. Obviously, yeah, straight, about it and do it in other fights. And yeah, no matter if it's good or a bad not.

Max Limper (DGG RapidPipeline)

people are just using it to basically have a shortcut there and it works ⁓ like amazingly well, right? It's like when I use a library, like usually developers use a lot of libraries, right? I think the times where you like, I don't see this at least a lot that people are like, let's say I'm programming UI with Qt and I'm the number one Qt expert in my company and I'm only doing Qt every day. Maybe, maybe I will be

Michael Finocchiaro

Of course, yeah. I'm the number one QTH person in my company and I'm only using QT everyday. Maybe, maybe I will be fast enough to type things by hand and they will magically do whatever. a very, good developer and I have to only do this all the time.

Max Limper (DGG RapidPipeline)

fast enough to type things by hand and they will magically appear. But I have to be a very, very good developer and I have to only do this all the time. And there's no use in me knowing as a developer, like all their APIs and all the functions and their signatures. But if I just say, okay, I want an application that looks roundabout like this and I can give it a picture and it creates me a UI, and then I can just take it from there. mean, AI even adds.

Michael Finocchiaro

and there's no use in me knowing as a developer like all their APIs and all the functions and their signature. But if I just say, I want an application that looks roundabout like this and I can give it a picture and it creates me a UI and then I can just take it from there. mean, AI even has

Max Limper (DGG RapidPipeline)

a lot of comments usually, right? And like it gives a lot of context. It's not even like dirty code or anything. It's usually quite readable. And then you can just go over it and, tweak it. And I think it's just changing

Michael Finocchiaro

a lot of comments usually, right? And like it gives a lot of context. It's not even like dirty code or anything. It's quite easy. And then you can just go over it and tweak it. And I think it's just changing.

Max Limper (DGG RapidPipeline)

the, the role of the developer a bit. It's not replacing the developer, right? I mean, that's also stupid. ⁓ It's more like facilitating it. And it's the same for like, if you're very average or below average, then maybe you're in trouble, but that's for everyone. Like the below average writers. are going to have a hard time. But if you're a good writer, then AI is not going to replace what you do. But maybe it helps you be faster and get to an initial state faster. And I think that's the same for every developer. And for us, we have this deep ⁓ SDK-like software. It's a command line tool, usually, what people use. anyway, it's very technical. ⁓ We're building interfaces for that.

Christian Stein

you

Michael Finocchiaro

Yeah, it

Max Limper (DGG RapidPipeline)

And there AI can help us to iterate quicker and can also help our clients to iterate quicker on that. So I think that's very similar to what Christian said.

Michael Finocchiaro

is. ⁓

Christian Stein

Yeah, I think for us it's a lot of the work to get things up and running is often mapping functionality that has to be coded. there are fixed structures from systems and somehow you have to map them to each other so that data can flow, that connectivity between data sources is established, for example. And this is exactly where we see a huge acceleration because you can, I think, get rid of a lot of the hallucination problems if the guardrails are fixed enough. And if you have very hard standards specified on each side, AI can do a really good job in just mapping them dynamically. And for example, an important integration for us often is with knowledge graph systems that are at the customer already.

Michael Finocchiaro

you

Christian Stein

spanning as an umbrella across different systems, pulling out information, generating higher level information mappings, now enabling 3D data in this environment often comes down to being able dynamically to build this linkage and pull in information.

Michael Finocchiaro

And this is

Christian Stein

And this is, the accelerator that we see at customers where the knowledge graphs are becoming like standardized interface to build LLM-based AIs

Michael Finocchiaro

again the accelerator that we see at customers where the knowledge

Christian Stein

on top. And where we can, again, then deliver the integration piece to enable 3D data in these applications and enable the LLMs to control. information flows or pull in computation on the 3D data. this is for us again a very, very interesting ⁓ topic for current ⁓ co-innovation topics with customers. On one side... ⁓ ⁓ pulling information that they feed into AI models in general, but then also for this idea of live interaction with your 3D data. And there you mentioned this from push to pull also a little bit. I think the overall... way how functionality is made available to applications will change in the way that before you had to kind of push it through to the end user and you needed to get his verification that what you build actually works for him. That's like you expected. And it's a very tedious and slow cycle. And with the AI based integration approaches like MCP now, there's a pulling mechanism now that kind of enables application layer logic to pull in functionality as required. And this is where we also think that the world's completely going to change over the next years and how all this like agentic application frameworks might in the future be a lot less coupled with specific functionality that is in the back end, but instead like pull them in as required and connect to other systems as required.

Michael Finocchiaro

Usually at this point I ask like how you guys are actually using AI in your programs and you told me ahead of time you're not specifically today. So why don't we turn the question around and say, like, you know, when we first saw Google, it was just a line on a screen where you type something, you were like, where's the application, right? And there was all this stuff going in the background. When we got ChatGPT, we just got a line that says, you know, ask me a question. And, and it's all very 2D and you and I, all three of us are, we're in the 3D world. We've been, there seems to be a disconnect and how, how are we going to get AI into 3D world? We were talking beforehand also of like, There's no large CAD model of ⁓ LLM that understands geometry intimately. mean, Leo AI created a mechanical model, right? So they understand mechanical engineering, but we still have got geometry as an LLM. So how do you guys see that evolving? Is that even useful or, you do you guys think that that's maybe the next phase for us in the 3D world that we'll have some kind of a model that'll... natively understand geometries, you can say, actually move that fillet over by two millimeters and make sure it doesn't intersect this other one. And it'll really be able to do that without me having to move a mouse or type anything. Just what do you guys think? Max, you haven't talked a while. We want to take a stab at it.

Max Limper (DGG RapidPipeline)

Yeah, can always talk. Yeah, I don't know. I think I'm just from a product perspective. ⁓ Like, I'm not sure if that's a satisfying answer, but my answer is it would really depend on the use case. And what you just said is very interesting, just picking that up, that random example, Move that, it over. I thought about that as well. I mean, for our software, you can do things like, OK, you can teach an AI to understand our JSON schema for the 3D processor. And you can say, remove all invisible parts of this model. And then it would work, right? ⁓ That's probably useful, especially for first-time users. if you want to quickly configure a workflow, I can see that on that level. I want to remove all invisible parts. I want to create a model that has a maximum of 50K polygons. I want to create UVs. And I want to kick out all parts that are a little detail smaller than one millimeter ⁓ and close holes that are smaller than whatever, five millimeters. And ⁓ you can write that in one query, and it's much more efficient.

Michael Finocchiaro

I want to remove all the good parts. I want to create a model that has a maximum of 50K polygons. I want to create a needle and I want to figure out all parts that are a little bit smaller than one millimeter and close holes that are smaller than whatever. Yeah

Max Limper (DGG RapidPipeline)

than ⁓ typing it in by hand and understanding where all these buttons and switches are and whatever. So I can totally see that. But then when it comes to detailed editing, I don't know. Yeah. And the mouse is also a powerful tool, right? So it's like, is speech always better? Is text always better than a mouse? I guess it depends. When you have a high level operation that is easy to describe in five sentences and usually would have a person then finding the right knobs and sliders, I think that is a very obvious thing that could be facilitated with text-based interface. If it comes to the details, and maybe the fillets would be already on that level, I don't know, Depending on what it is, right? ⁓

Michael Finocchiaro

Well, you have to train the model on parametric modeling in that case, right? move things around to be able to move them around.

Max Limper (DGG RapidPipeline)

Yeah, mean, I can totally see how you could say, I don't know, I have ⁓ something that has five holes in the design and I say, can you make the same thing that looks stylish and symmetrical with three holes or whatever. But that's again, one of these fuzzy queries, very precise things. I even read this at some point that someone said like, hey, math is actually a very powerful language, right? ⁓ And it's very nice. So if you want to say something very precise, ⁓

Michael Finocchiaro

Yeah. Yeah. ⁓

Max Limper (DGG RapidPipeline)

Right? And sometimes the interface that we have are probably the most powerful ones. But then if it's fuzzy or if it's a very general top-down task, then maybe the text will win. Right? That's what I'm thinking at least.

Michael Finocchiaro

That's sort of what NTOP is doing with implicit design, right? mean, everything comes down to the math. can just do your mathematical stuff. You're nodding there, Christian. You must be in agreement with us.

Christian Stein

⁓ I think we are not so much about, let's say the actual ⁓ work of engineer in the CAD, but more about how the data that's generated can be used for other people. So looking at that one, think on one side there are like all these different initiatives, if there might be some better representation to do CAD modeling at all.

Michael Finocchiaro

course.

Christian Stein

And I think it's going to be very interesting to see if it's rather someone like we talked a little bit about the Techsoft approach before about using data as is and generating better fitted representations to feed into language models, or if it's going to be about a different kind of representation in general that might be more suited to that. But it's definitely not where our work is or our experience lies for now.

Michael Finocchiaro

where our work is about our experience in ice hockey.

Christian Stein

At the same time, think if we look at this higher level usage of data, I had a very interesting talk yesterday with someone from Bosch, especially if you want to teach AI information about the physical world,

Michael Finocchiaro

At the same time, I think if you look about this higher level of research of data, I had a very interesting talk yesterday with someone from Bosch, where especially if you want to teach AI information about physical.

Christian Stein

think at least in data research there's a very strong understanding that... it's going to boil down to a lot of very different and very specific models to do individual tasks. And then maybe the LLMs can be something like the umbrella that translates between them that like generates how the different sub-modules are going to be used. ⁓ But there are like so many questions still to be solved. then I think at the...

Michael Finocchiaro

Okay.

Christian Stein

big player level, people might move towards building new world models, et cetera. And those probably are then still a few years away before we see the first incarnations. I think for the very specific task at customers right now, it's going to boil down to, there a combination of existing AI capabilities that can be combined to do a very specific task? And then in our experience, this maybe will cover the next maybe one to two years to really get those up and running and reliable in production. But people are starting to build those now. And ⁓ there's still a huge way to go, I guess, to build something.

Michael Finocchiaro

Well, that's good news for us as developers, right? That means we still have work for a couple of years before we get obsoleted by an agent. ⁓ Actually, we do have a question in the chat. That's a good question. ⁓ Dick K Digital Twins. How did he get that name changed? ⁓ Dick K said, what tools or frameworks do you recommend for developers?

Christian Stein

Not so far.

Michael Finocchiaro

wanting to start integrating AI with real-time 3D projects? I mean, is it just cursor and Windsor for their, you know, is it hoops? Maybe this new AI hoops thing? Well, you guys even, I mean, maybe you haven't even thought of it, but I think it's a good question, right?

Christian Stein

I guess it's always boiling down to what do you want to achieve? mean, the question is very generic. And I think, ⁓ yeah, if you want to start like training something on cat data, you should always come from the question, what do you want to train? Where's that information? How can I get it out of my data in a reproducible, reliable way? And what's the representation I feed to my model to learn something? If it's about maybe decide where we are more active. if you want to develop an application that kind of interfaces with real-time 3D. I always recommend our software obviously, but it's going to come down to how much of the work is already available in things that people have built that are reliable and how much do you have to do yourself.

Michael Finocchiaro

Hahaha Makes me wonder if the native language would be closer to STEP or closer to Hoops or closer to the actual parametric data. Anyway, Max, do you have an answer for that one or some thoughts on it?

Max Limper (DGG RapidPipeline)

Yeah, I I think there is no specific solution for general real-time 3D projects, a way we can do that. But yeah, as Christian said, I guess it depends. So if you are building, let's say, a real-time app based on Babylon.js in the browser, that's something that's very popular. And this is just coming from this more lightweight real-time site right now. ⁓ And you know exactly what you want to visualize. And you want to, I don't know, have some clickable hotspots or whatever, and it's very static and it's this one thing that you want to do for, I don't know, a trade show or whatever. You can probably do that with any ⁓ generic AI that can do coding and has basically seen the Babylon.js docs and can put something together. And then if you want to do that with a CAD model and you want to have more meaningful interactions, then think what Christian and Tim are building in Instant3 hub has these web components. the AI can read that, the docs for that, it can also put those together and can say, okay, can I, I don't know, have something where I can have like clickable cat lines and like distance measurements and all these cool things that instant 3D Hub allows. And ⁓ what we can do is basically the data prep part if you want to generate LODs or whatever. But I guess it's really depending on what you want to do, like, ⁓ And what building blocks do you want to drive? And the driver of that is probably a generic. AI, suppose.

Christian Stein

So I wasn't sure if it just the name of the guy, Digital Twin, or was it about Digital Twin. ⁓

Michael Finocchiaro

I think he just added it after his last initial.

Christian Stein

Because ⁓ I think ⁓ in our world, obviously, we first need 20 minutes to discuss what a digital twin is, but there's often this, in our opinion, understanding that whatever 3D data set you kind of clamp together in a game engine somewhere is now a digital twin. And ⁓ our opinion is that the digital twin is all about You're managing, understanding your data, building a layer where you have interconnected information that represents your digital twin and then the instantiation of a product that you have built, obviously, and all the data that's connected to it along the life cycle. That at the end, that's the, so it's all about the data that you're managing and how easily you can make it available and connected with each other. And then... Viewing, visualizing it, it's just a form of consuming this data layer that you have to build.

Michael Finocchiaro

I wonder if the question could also be oriented towards the industrial metaverse though. Like I'm doing this series of articles right now and I was last week I talked about co-pilots and this week I talked about agentic AI. Next week I'm going to talk about industrial metaverse. And if you think about real time 3D, I'm thinking more about on the Nvidia omniverse platform. I take my 3D model and I put it in this virtual reality where I can manipulate it and figure out where the weak areas are and then I can put that data that I learned back into the PDM and to the CAD system and then iterate to improve my model without actually having to build any kind of prototyping or anything. I think that's an interesting place where developers should work. If you learn OpenUSD and you start working on that, that could be an interesting area. I don't know if that coincides with what you guys are working on or.

Christian Stein

I can maybe just state one of the customers we are working with, are using Omniverse currently as a, they call it simulation tool for experts. And I think there's a huge complexity in bringing your data into Omniverse in general and making sure that the information that you need can be reflected in the OpenUSD standard. mean,

Michael Finocchiaro

Okay.

Christian Stein

It's a great initiative. A lot of things are being harmonized and it enables people really to get this information from different sources. But at the end, it's a format-based standardization. so the complexity is in automating that. So considering the automotives again, if you're talking about getting your factory, your robots, your production data of your car for a given point in time in a specific facility,

Michael Finocchiaro

Mm.

Christian Stein

⁓ together dynamically so you can really run any kind of simulations in Omniverse on that. think people are still having big challenges of achieving that truth of data at all in that environment. And then once you have run your simulation and the results you generated, you have to transfer it back to the other systems again. So this closed loop of information is really something still to be solved. And while I think there's obviously a lot of capabilities and so many interesting work done, at this point, I don't see that many real productive systems built on omniverse yet. There's maybe just as with AI, there's more and more touch points and people are starting to maybe generate value from it, but it's also very early in the beginning.

Michael Finocchiaro

I agree. That's why I called it the stage three AI. You know what mean? After copier that's in agentic, even if the AR VR stuff we've been doing it for a while, but really integrating that. with the 3D and not just as a funny way of telling people instructions over the internet. No, no, no, actually use the third screw there on the left. No, no, move your hand. know, where you're actually using the 3D in real time. Do you have any thoughts on that, Max?

Max Limper (DGG RapidPipeline)

Yeah, mean, would. I think Christian is already like ⁓ thinking about this probably. I'll just say it for him. Like they had an amazing project ⁓ that is publicly available as a case study with BMW where it has this thing called 3D Concept. If you haven't heard about it, check it out. So this is actually doing this already for some years. It doesn't require omniverse or anything. It's just instant 3D Hub and people at BMW, I think.

Michael Finocchiaro

Nice.

Max Limper (DGG RapidPipeline)

Christian, how many ⁓ were there at the time of the case study?

Christian Stein

⁓ some several thousand users on a daily basis.

Max Limper (DGG RapidPipeline)

Yeah,

Michael Finocchiaro

Wow.

Max Limper (DGG RapidPipeline)

that's what I got, right? So we have several thousand users interacting on real cat data and communicating about it and using that as a source of truth and doing that interactively. like Christian and team have an integration with MS teams and things like that. So they're really where the users are. And with Omniverse, ⁓ yeah, I mean, I could talk more about USD and how that could be connected with AI. ⁓

Michael Finocchiaro

Yeah, I mean, I could talk more about you and how that could be. Sure. It's interesting, I'd like to know about that.

Max Limper (DGG RapidPipeline)

Yeah, but I think maybe just to add a side note here on Omniverse, like we are working with like USD conversions for Omniverse, for clients and so on. And like it's a platform that can do certain things, but NVIDIA is not acting as a solution provider there, right? I mean, they're not doing that. ⁓ You can work with Accenture or someone who then does the actual solution building. But yeah, I think you have to put in a lot of work to achieve something similar for what, like, comparable to, let's say, what instance 3D Hub does with the 3D Concepts app. If you want to have really a domain-specific problem solved, I think there's no way around, like, building something specific. And then ⁓ for the USB side of things, I think just wanted to add, one thing that people often just see it as a file format, but it's

Michael Finocchiaro

I think there's no way around like building something specific. And then for the USB side of things, I think just want to add like something that people often just see it as a file format, but it's actually more, right? It's a runtime also. And...

Max Limper (DGG RapidPipeline)

Actually more, right? It's a runtime also. And if you embrace it as a runtime, then you can do a bit more and also programming it that way, It's basically like a scene graph with many cool additional features and kind of like scenes interlinking and you have all these concepts for like measures and primitives and stages and whatnot. ⁓ Yeah, and I think that's also.

Michael Finocchiaro

Interesting.

Max Limper (DGG RapidPipeline)

interesting potential, but it's kind of disconnected from Omniverse per se. It's like Omniverse is USD based, but you can also program USD outside Omniverse.

Michael Finocchiaro

Right, I was just in the article that is a spoiler for next week, you know you basically the three big vendors have already committed to being able to export that format and then you know Siemens and PTC have signed deals with ⁓ Nvidia to integrate to the universe, but that's really cool.

Christian Stein

Yeah, yeah, but I mean, Siemens is embracing this for maybe two and a half years now. And at the end, they give you technology on how to pull data from Team Center and generate USDs from it. And you can automate your flow. But the question then is again, how much time does it take? Because I mean, you mentioned the metaverse initially. ⁓ for us, while it's probably a burn term still,

Michael Finocchiaro

Mm-hmm. Right.

Christian Stein

At the end, it's a lot about these XR spatial computing use cases that we see, for example, in the production at customers, where it's all about having the data as you need it because you want to place manuals, you want to project information on your physical objects so you can support working steps where before you had a lot of plastic tooling and stuff.

Michael Finocchiaro

Of course, yeah. you ⁓

Christian Stein

⁓ I'm not sure if these two giants just by giving you a way on how to convert between systems are actually solving the problem of making data accessible.

Michael Finocchiaro

no, no, no, think that you're right. It's a different, a separate issue. And I think we'll talk about that in the last question. But before we get to that one, ⁓ I wanted to know. So we've talked about, ⁓ you know, 22, 23 when the LLMs came out and how you guys, it was more of a pull than a push, but you guys were both enthusiastic with a little bit of skepticism, right? If I characterized. Now we're in almost 2026, right? We've been through. ⁓ the LLMs and then the agents and now we're looking at MCPs and everything is just going a bit nuts. You've got the crazy financial juggling between OpenAI and Oracle and nobody really knows where that's going to go. Are you guys still bullish on it? Do you guys think that there could be a bubble problem, you know, like we had back in 2007 with the subprimes and it's just going to... I think AI is here to stay and it's going to be of massive influence to us as engineers for a long time. Do you guys also feel that? what's your opinion on it going forward as of now in late 2025? Max?

Max Limper (DGG RapidPipeline)

Alright, yeah, I'll try to keep it short. I mean, I think there's clearly a bubble, right, in terms of valuations. And I don't know, like the circular deals between Nvidia and Oracle and that's all crazy, right? And the whole Oak May, I promise. I mean, I don't know if many care what I think, but like I dare to put my word out and say like that.

Michael Finocchiaro

Hahaha You it.

Max Limper (DGG RapidPipeline)

this is not going to fly like their projections with the because many people and you can see smart people blogging about this already like many, many people that use OpenAI are using it for free relationship advice and things like that. Are they ever going to pay? Right. And I saw someone posting this recently about like, if you look around you and including people like I pled guilty here, like ⁓ how they use YouTube.

Michael Finocchiaro

Yeah. Yeah.

Max Limper (DGG RapidPipeline)

Like I put on some background music on YouTube and then I'm annoyed by the ads, but I'm not buying YouTube premium This tells you and someone wrote like this tells you everything you need to know about B2B Yeah, I don't know like what Sam Altman thinks about his business projections there Yeah, I wouldn't be so enthusiastic. I think they're still going to stay around right and it's like with the dot-com bubble like we still use eBay and Amazon and like they grew so big and like of course everyone uses the internet and shops there but that doesn't mean that all the domains that were bought for ⁓ so many millions back then and then were a complete flop that that isn't a part of history. So we're just going to see a big cleanse where basically overvalued, overhyped stuff is going to disappear and like where actual value is created that's going to last and sure yeah it's going to stay around.

Michael Finocchiaro

You mentioned Amazon. makes me think the impact that Amazon had though is it did destroy a lot of brick and mortar stores, right? Bookstores, music stores, it just decimated them. ⁓ Downtowns are emptied of all those stores. Well, in Europe we still have a handful, but in a lot of places, I mean, it just decimated downtown and the old concept of downtown, you go downtown and go to three or four shops, but those shops don't exist. You just have a mega shop or you have Amazon, right? I think there are societal impacts. I don't know, Christian, if you want to talk about how is there sort of a trickle down effect there too, in terms of AI and this hype and where we're going to go with it from 2025 on.

Christian Stein

I guess I'm totally aligned with Max in the general opinion. ⁓ I very recently read an article that I thought was really great about this not being a bubble but a wildfire and it's going to be the cleansing burn that enables the new generation to grow from it and the big trees, the sequoias will survive and I really like that one and I think...

Michael Finocchiaro

⁓ you That is a good one.

Christian Stein

If we take the AI bubble and maybe also economic situation in general, especially here in Germany right now, it's going to be tough for the next years, but whoever survives that is going to be a lot stronger. the AI products that will have survived, they will really provide a huge value. And then we'll see what the next generation of products will be on top of that.

Michael Finocchiaro

So I want to just switch gears for the last 10 minutes of the discussion and talk about the real world in terms of customers and AI. So when I created this podcast, my friend Rob Froney said you should do a thing on AI. And I said, nobody's going to listen to me talk about AI. And no customer is going to tell me what they're doing with AI because it's all confidential and all that kind of stuff. However, vendors like you guys who are on the bleeding edge, who are doing awesome stuff that people that didn't even imagine several years ago are certainly coming into contact. You just mentioned Bob, but we don't have to mention any questions by name. when I think about this problem of data access that you mentioned several times, Christian, I think about a spectrum of digital maturity from one to five. And I think one is like, I'm still using email and Excel. And five is sort of that industrial metaverse on steroids with adaptive digital twins that are completely agentic and autonomous. Basically like a Kim Stanley Robinson's red Mars and the train that builds itself, like eating the regular, then building the tracks and audit robots repairing themselves and all that stuff. Well, I don't think anybody's at five. I don't think anybody's even four. I think most people are between one and three. Maybe I'm wrong and maybe you guys have a different opinion. So first of all, like first part of the question is, know, where do you see your customers on that spectrum? And the second part of the question, question is when people start using AI, when, when the, customers start implementing. different solutions, do you think there's a ripple effect or do you think it stays in the bubble? Like people are still skeptical and they're like, well, no, I don't need to change and fix my data silo problem because you know, it's, if it's not broke, don't fix it. Right. So I just wondering what you guys have felt from a customer perspective. I'll let you go first, ⁓ Max.

Max Limper (DGG RapidPipeline)

Yeah, I mean, I just want to share one thought. Because I was under this illusion earlier when we started our company and we had worked with a couple of customers, but not so many. And ⁓ I wasn't aware how important company politics are and ⁓ siloing when it goes to enterprise.

Michael Finocchiaro

I believe it.

Max Limper (DGG RapidPipeline)

If you have a marketing department that has no idea what a 3D file is or CAD file, right, and you could do amazing things with that, ⁓ but they don't get that data and the data is not in the right form or shape, is that going to change, right? it politically, is it going to change? ⁓ Unless there is a transformation on another level that breaks up these silos, but that's a different story. And as long as that doesn't happen, ⁓ the end to end dream, that's always something that like,

Michael Finocchiaro

the end to end dream, that's always something that's

Max Limper (DGG RapidPipeline)

I think when you start off with a startup or like investors also like are very quick to just make up these scenarios. But what are the requirements not only on the technical side, but also on the organizational side, political side incentives, right? are all incentives aligned? Are really people going to push towards that end to end? And that's,

Michael Finocchiaro

like, think when you start off with startup or investors also like are very quick to just make up these scenarios. But what are the requirements, not only on the technical side but also on organizational side, political side, incentives, are all incentives aligned? really people going to push for that? And...

Max Limper (DGG RapidPipeline)

I think the far bigger problem. than anything technical. And I don't have a definitive answer to that for like for sure. So I'm just pointing out, I think this is probably an equally big roadblock, if not the bigger one.

Michael Finocchiaro

The OCM problem, the organizational change management issues.

Christian Stein

Yeah, I see it the same way. think from the numbers somewhere between two and three, maybe. ⁓ I think it's an organizational topic because what we have seen a lot now is that ⁓ each department has its budget and does its innovation independently, which seems totally crazy. ⁓ With these ideas, we had a ⁓ case where ⁓ people even told us that for the initial things they're doing, they would ⁓ keep IT and engineering out of it because those people make it complex and slow. ⁓ the question is... Can that be a foundation for success then at all? ⁓ I think if you have these first initiatives implemented successfully, there definitely can be a ripple effect and it can maybe be the drop that helps people to really decide to break up silos, etc. But we're not seeing it yet.

Michael Finocchiaro

And do you think that's a fact? Well, it could it also, there's the fact that the company is using like one of the huge monoliths, right? The two, well, let's say the three, the big three, is that, is that what's holding them back? Because you've got SAP data over here and I've got either Siemens or Teamcenter or Teamcenter or 3DX or Windchill. And these don't really communicate. And then I've got CRM and all the other data. Is it more a fact that I've got this monolith? Whereas if I had gone If I have more open interfaces and I'm using smaller, more agile companies, it might be more easier to coalesce the data into something.

Christian Stein

Yeah, I'm with you there. think if you now have to decide on which monolithic platform you're going to put your AI initiative, it's going to be siloed to that domain. So as long as you're not able to kind of mature from that, build something unique on top of it with other technologies, it's going to be really difficult.

Michael Finocchiaro

I don't know if you guys have heard that expression, but I really like this. I think he's actually a German professor, Yousef Hussmann. He said this ⁓ nearest source of truth and single source of change. I really like that because it's true we're never going to get agreement between an ERP and a PLM on the truth because they both have two different unique views in the world. ⁓

Christian Stein

Yep. Yeah, yeah. We actually worked with him some years ago at Mercedes, where he was defining this federated data layer before he switched to Neo, I think. He's really strong there.

Michael Finocchiaro

okay. Yep. Brilliant guy. Yeah, I got to see him at the SIM data show. Very, very good. It's been really great. I really appreciate ⁓ your guys' availability today. If you have any closing thoughts about DGG or about 3D or about AI in general, before we go.

Max Limper (DGG RapidPipeline)

I mean, just for my end, we didn't talk much about the conversions of workflows with AI for marketing imagery and CGI, which is a completely different topic. But that's my closing thought, that there are some interesting things happening there. And yeah, I'll be happy to talk about that another time. So.

Michael Finocchiaro

Okay, which is a DGG thing, right? Absolutely. I'm quite optimistic that the other things were, there was a very interesting direction, which we got more and more crowd-vibes with the help of AI and things.

Max Limper (DGG RapidPipeline)

I'm quite optimistic that things will evolve in very interesting directions and we'll see a lot more momentum around 3D even with the help of AI in other areas ⁓ that can even help to break silos. ⁓ But ⁓ yeah, it's going to be department by department effort basically, not cross-functional from day one. That was my last thought then.

Michael Finocchiaro

I actually want to ask another question, a quick one, but more of one that actually has come up a lot of the time. I still get, I get calls now because I've been on LinkedIn a lot and a lot of people, younger people call me if you say, if know. I'm in engineering school and I don't want to be written out of my job. What should I work on in order to become a better engineer that's relevant? Would you guys have any advice? Like one quick little advice for these guys that are in school or grad school and they're looking at AI and trying to figure out what they need to focus on to be more, to remain relevant.

Max Limper (DGG RapidPipeline)

You wanna go first, Christian?

Christian Stein

I mean, if it's about ⁓ becoming a developer or let's say an AI enabled product manager or however these roles are to look like in the future, I guess it's going to boil down to really do the hard work and understand what's underneath still so you can become the human in the loop in the future world of software development. And in general, I guess for us regarding the last statement there, ⁓

Michael Finocchiaro

the real engineering stuff.

Christian Stein

We think that five, 10 years from now, using 3D data anywhere in your work life at our customers will be an absolute commodity. we really think that the next generation also of hardware will enable these kind of mixed reality interactions where we now have, I think, the next wave of companies with a lot of funding, maybe building for the US military, but then at some point, enabling their hardware to the world. And there AI is going to be the glue that enables that. think, as you said, it's going to stay and it's going to play a very important role, but maybe not in way it does today.

Michael Finocchiaro

Thank you, Christian. Max?

Max Limper (DGG RapidPipeline)

Yeah, I would just echo again what Christian said. And he said a word like a product manager. I just think like, if you can become at least a small product manager and think beyond the task that you get as an exercise when you're learning to be a developer, for example, right? Like, oh yeah, here's a program that we already started now at a function that for each user in the database does this and this. You should learn this. I think it's really important to know.

Michael Finocchiaro

things like, if you can become at least a small product manager and think beyond the task that you get as an exercise when you're to be a developer, for example, right? Oh yeah, here's a program that we already started now as a function for each user that are based up to the distance.

Max Limper (DGG RapidPipeline)

like have the low level understanding of everything that's going on. But you should also understand that many of these things that are very kind of quote unquote stupid exercises of like here's an exact description what needs to be done. Now write a program that's not going to be written by you possibly in the future. So you want to have that understanding ⁓ of the context, right? And that's what a product manager has. It's like, why are we doing this in the first place?

Michael Finocchiaro

like why are we doing this on the first day?

Max Limper (DGG RapidPipeline)

Are there other solutions to this? Can I tweak the parameters a little bit of this task so that it becomes 10 times more efficient but still gets 95 % of the job done for the user? This kind of thinking cannot be done by an AI in that way, ⁓ at least not that easily right now. And I think, yeah, that context knowledge is just becoming

Michael Finocchiaro

Are there other solutions to this? Can I tweak the parameters a little bit of this path so that it becomes a kind of for me, but still get 95 % of the job done for the user? This kind of thinking cannot be done by an AI in that way, at least not in that interview right now. And I think the other context on that should be just.

Max Limper (DGG RapidPipeline)

more more important. so, yeah, don't be just an implementation machine. Don't do it to learn that, right?

Michael Finocchiaro

Yeah.

Max Limper (DGG RapidPipeline)

understand context and creativity.

Michael Finocchiaro

Be an engineer, be a human engineer.

Max Limper (DGG RapidPipeline)

Yeah, exactly. We're still needed.

Michael Finocchiaro

I think we lost Christian, but ⁓ I guess we can ⁓ wrap up and thank you for attending this podcast for the people that were on live. I think we had a better crowd than usual. And ⁓ next week I'll be talking to Lambda Function and another German startup in OptiParts, as well as ⁓ OpsMate AI from ⁓ XPTC guys and ⁓ TD Engine, which is a really exciting ⁓ industrial AI tool. Thank you. Christian's back for the goodbye, think. We lost you there for a second, Christian. ⁓ I just was saying ⁓ that we've got two more webcasts next week. ⁓ I was extremely happy to, I really enjoyed what you guys had to say. I think I learned a lot. I learned more about the industrial metaphors. Thank you very much. That was helpful. I'll probably inform my articles next week. I'm gonna have to rewrite them now a little bit. ⁓ Thank you very much. Did you guys have a good time?

Christian Stein

Yeah, thanks a lot.

Max Limper (DGG RapidPipeline)

Yeah, that was great. Thanks, Michael. Thanks, everyone in the audience. was a pleasure.

Michael Finocchiaro

Okay, thank you and talk to you guys soon. Thank you.

Max Limper (DGG RapidPipeline)

Vielen Dank!

Christian Stein

Thanks. Have a nice weekend.

Max Limper (DGG RapidPipeline)

Thanks, bye!