Adventures in Automated Design: A Generative-Modeling Case Study

PETER CHAMPNEYS: Hi, everyone, and thanks so much for joining us for today's Autodesk University class. This class is going to be a deep dive into a brand new tool that we have released for Fusion 360 called Automated Modeling. So my name is Peter Champneys, and I work as a Customer Advocacy Manager for Autodesk, with a focus on generative design and automated modeling.

And to start off with, I'm going to give you a bit of a background and insight into why we at Autodesk decided to create this tool, how it works, and a bit about what differentiates it, in particular, compared to tools like Generative Design. And this is a tool that we released in the July release of Fusion 360, but it's actually been available to our insider community for a number of months beforehand.

So joining me today, we've got two of our insiders, who've been using a lot of Automated Modeling who are going to be joining me to give you some insights into how it can be used. So after I've given you the introduction, Daniel Mayer will join us to talk about how he used Automated Modeling to design and make a floor lamp. And then we've also got Marty Jacobson joining us to talk about some best practices and applications for Automated Modeling, so some kind of practical tips.

So before we start, just a safe harbor statement, most of what we show in this presentation is things that are in the products, so in Fusion 360 today. But towards the end, we will be giving you a kind of insight and look forward at some tools that are not yet in development. So this is just to say that although these represent our current plans, we reserve the right to change what actually ends up in the product at any time. And so, you shouldn't use that to drive any purchasing decisions about the software.

So as I mentioned, I'm going to start off by giving you an introduction to Automated Modeling. And Automated Modeling is a design tool that is very much powered by the tools and concepts of Generative Design. So I'm going to start by talking a bit about Generative Design as that gives a bit of background to why we decided to create automated modeling.

So if we think about Generative Design, the key element of Generative Design is that instead of the human coming up with designs, instead, we have a design tool where we describe the design problem and we use intelligent algorithms to create design possibilities for us. And we don't just create one design possibility, but we create lots of different options and ideas for UTs.

So this is a technology which has been used, is really powerful, and has been used successfully across a number of really quite different industries, so things like a Mars Lander, all the way through to fixturing for CNC machining, and from designs of high performing racing drones all the way through to aftermarket components for motorbikes and cars. And if we think about the way that Generative Design in Fusion 360 works, the first stage is that we have some kind of design problem we want to solve.

And instead of modeling something with our capsule, instead we essentially define that problem to the software. And that definition is going to consist of the things you can see here, so things like manufacturing methods. We're going to spend some time describing the manufacturing technology we have available, so that the Generative Design algorithm knows, OK, when I create my design, I need to meet these criteria in order for it to be manufacturable with your CNC machine or 3D printer.

We also give it specific optimization criteria, so things like I want you to come up with the lightest shape possible while not exceeding the yield stress of the material. And again, a key driver of how Generative Design has always worked is that we need to take the time to put in structural loads and forces, and this is going to be a key driver of the shape that comes out the other side. So what we then do of Generative Designs, we take all this data and information that sets up a design problem and we create designs.

So we use a generative design algorithm to create possible CAD models. And we don't just create one design, but we give you lots of different design possibilities that lets you make trade off decisions about aesthetics, performance, and costs. And then lastly, we give you this information and display it in graphs, data, information, that lets you take a look at all the different designs that Generative Design is great.

So as I mentioned earlier, this is a really powerful technology, and it's been used successfully in a wide number of industries. But it also has some challenges, particularly if we compare it to that typical design process I spoke about earlier. So for example, we can see a drone design here, created with Generative Design on the screen at the moment. And if I asked people to model a drone in CAD, then most of you could model something out pretty quickly.

But if I asked you to define, very specifically, the forces and structural loads acting on that drone, that might not be something that you would be able to do as quickly or easily. And so this is what we see with generative design, is that it's an extremely powerful tool, but a lot of these set up information requires specialist knowledge and this takes a lot of time and complexity to set up. So when we were looking to create automated modeling, it really started with a question.

And that question was, what if we could create a design tool that led us in the same way as Generative Design, have an intelligent algorithms to create solutions for us, that explore different design possibilities, but doesn't require all of that complex set up information to do so, and lets me interact with the software with no more knowledge, experience, or time constraints than if I was sketching or modeling in a typical piece of CAD software?

So that is, essentially, exactly what automated modeling is all about, is that instead of taking all this set up information, instead, we simply provide a basic CAD-based definition, and that's it. And Automated Modeling will take that information and create automatically a variety of different design possibilities for us in Canvas, in Fusion 360, in the design workspace. And we can then take a look at the different alternatives it's created and decide which one we like. So let's take a look at how that works.

So, here we are inside Fusion 360, and we wanted to create a rocker arm for this bicycle frame. So we're going to use Automated Modeling to do so, and you can see we've got this brand new tab in the Fusion 360 environment called Automate. So when I select Automated Modeling, the first thing I'm going to be asked to do is select faces to connect. And this is pretty key to automated modeling, is essentially doing, is it's connecting, creating designs, which connect things together.

So I've selected some faces, and that's the only information I require, and I can go ahead and press Generate Shapes. Another really key thing about Automated Modeling is the speed with which we can create design. So typically, with Generative Design and Fusion 360, we create many different design possibilities on the cloud in a number of hours. Whereas with automated modeling, we're creating different design possibilities in a matter of minutes.

So this has been slightly sped up, but typically, we would see all our results generated in around two minutes. And you can see, I can preview different alternatives, and I'm going to get a kind of translucent gray preview for what my design is going to look like. And if there's one I like the look up, I can go ahead and press OK. And we're going to create a brand new body in Fusion 360, exactly as if we'd modeled it ourselves.

If I look down at the timeline, you can see that we created an automated modeling feature that if I want to go back in and edit it. I could select a different design, for example. So that lets me now quickly go back, change alternatives. You'll also see in the timeline, we automatically create a fully editable T-spline or body, and this lets me go in and take the starting point from Automated Modeling and adjust it, so modify it using our form environment.

Another really nice thing about this, you can see, because I have set this up with a symmetric setup, we've automatically created a symmetrical design, and we've automatically applied a symmetry plane, so that when I make edits to one side, it will make edits and adjust both. So just to give you a bit of an insight into some of the different alternatives that are created, we actually offer two, three different kind of, currently, three different kind of geometry styles and two different connection points.

So you'll see here, if I change connection point, what you'll notice is that the connection point to the face, and this is what we call sharp connections, has now changed from an organic connection to more of a kind of prismatic shape. So these are two different connection styles that we provide you with. So if I go back into my edit feature, there is actually another option that we have there, and that is called bodies to avoid. So this is basically an optional setting that is often going to be very useful.

And that is basically going to also add an additional body that the created Automated Modeling alternative is going to make sure it stays clear of. So here, we've added a clearance region. These bodies might represent your wider assembly. They can also, as we're going to see later, be used to drive the design maybe in an alternative direction, a design intent, or specific aesthetics. And you can see here how alternatives have automatically adjusted to account for this body.

So that is a look into what Automated Modeling is and why we created it, and just to summarize the differences between Automated Modeling and Generative Design. So Automated Modeling is currently available as a public preview in Fusion 360. And unlike Generative Design, this tool does not require anything on top of the base Fusion 360 subscription in order to be able to access it. So it doesn't require any kind of extension, and it also doesn't cost any cloud credits on a per use basis to run it.

Unlike Generative Design, which is available via the generative design extension. So again, you've seen how Automated Modeling is able to create designs in a couple of minutes, instead of hours, and also doesn't require really any set up information. So it's going to be a really fast towards quickly design, ideate, explore design possibilities.

Whereas, Generative Design gives us the advantage of being able to put in specific optimization criteria, like material properties, load cases, manufacturing manufacturability, which Automated Modeling is not doing. So it's not going to necessarily optimize a design for a specific load case like we could do in Generative Design. So we can see Automated Modeling as being very much a pure design tool, a modeling tool, a design ideation tool.

Whereas Generative Design starts to fulfill the niche of being more of a kind of primarily engineering driven tool. So thanks for listening to that intro to Automated Modeling, and it's now my pleasure to introduce Daniel, who is one of our insiders who's been using this tool on-- well, he's already used this tool on a live project in his role as a product designer. So I'm going to hand it over now to Daniel, who is going to introduce himself and the project that he used Automated Modeling for.

DANIEL MARCOS MAYER: Hello, everyone. My name is Daniel Mayer. I'm 23 years old, and I was born and always lived in Brazil. My background is as a product designer at a Lebrock studio here in Brazil at the moment. So my goal here is to show you a product and the process from ideation to physical production. The product is called Link, and I'll talk about it during this presentation.

So being in Brazil, I developed many partnerships, the main one being Lubbock studio, a very prominent company, design studio, here in Brazil with a profound connection, because it is a company with a focus on caring for the nature and the sustainability of wood commercial operations. We go by a wood selection process carved on a term we created in the studio called testimonial wood. This is a term that means wood that has history and past use, like rescue wood, wood collected after natural disasters, and adverse effects of nature, and even accidents.

We render a service to the local community and, therefore, we receive their assistance to do it. So all of Lebrock's work starts from forms found in nature, like animals and even the human body. All the products you see here were created entirely using Fusion 360, based on sketches and concepts, but without Generative Design and even before the existence of the Automated Modeling tool until now. Due to limited production, Lebrock pieces are sold at high prices here in Brazil and the surrounding South American market.

So putting together a daily demand for the company, innovating, and creating new designs, with the fact that I'm an inside user of Fusion 360, where there was a competition for the early use of the Automated Modeling tool going on, I decided to serve both purposes and to design a piece of furniture. And to inspiration, I chose to do something that I admire a lot, which is the Madagascar ecosystem, and nothing to represent it better than the ancient baobab trees found in one of the regions is fair.

So I was not familiar with the Automated Modeling tool, and then I started exploring how shapes and connections interacted and what the results of the choices I made were. So I started to create shapes of what I would do, and I started creating various pieces along the way, that you can see here, as a preliminary drawing of some lines that I would like to have. One of the parameters I set was the proportion, which gave me the idea of making something cylindrical and tall, just like the baobab tree.

And that's what I did, observing several photos from different angles of the tree. I'd find proportions through primitive shapes based on circles, and divided it into stages. It was at this point that I noticed the resemblance to a lamp. So with that in mind, that I proceeded. So here, you can see that I started with stage one. Here, I select the two sides you see. In here, you can click Generate, and even iterate between the forms while they are being generated.

So here, you choose one of your preference, and you click OK. And here, I disabled the construction plains that he created too. And here, you can see that we generated the fruit of the lamp. So here we go to the second stage that I showed you before. Here you select the top phase and the bottom phase of this upper object. And here, you can select a body to avoid which generate something that have no intersecting parts with the already existing parts.

So here, we can iterate between the generating parts. You chose one. And here, what I did after clicking OK was some modification, some polishing in the form, some rotation, some copy and pasting. And here, well, they use mall work, amount of work. You can see that we almost get it done virtually. So, what you see here is a timeline of what we created using these two stages. Here you can see stage one and stage two, and the final form in the end.

So here, we go to the manufacturing planning part. So for that, you had to adjust the model knowing the machinery parameters I went and limitations, And the model turned out as you see here. I like part as several strategic points to accommodate the height of the [INAUDIBLE] drill. And I chose to use a simple lathe to make the most significant part of the pedestal by slicing it in half and extracting the profile to control the dimensions.

So here, you can see some videos of the-- photos and videos of the production phase of the model. And after that, the piece goes through numerous manual finishing processes to take the quality expected for a premium product. So that's it and here we go to the conclusion where we look at our past use with the tools, and we look back on my history with Fusion. I realize the potential it has with the tool palette developed by the Fusion Team, which grows day by day. It is excellent to know you are using industrial leading technology solutions daily. In Lebrock we run all our operations entirely in Fusion, from ideation to running on milling machines and miniature prototyping equipment like 3D printers. So that is what I had to show you today. It was a pleasure. I hope you enjoyed it. And now stay with Marty for more.

MARTY JACOBSON: Very cool stuff, Daniel. My name is Marty Jacobson, I'm also an insider and participated in the same exploratory process with this tool that Peter invited us to. Yeah, it's a really fun tool. And I'm going to be talking about some of the things that I discovered and kind of distill it down to some recommendations and best practices that I think you all will benefit from if you want to try this tool.

I'm a lecturer at Georgia Tech. I run a machine shop, which is where our biomedical engineering students and researchers make prototypes for all kinds of different stuff for science and startups. And I also am a CAD, CAM enthusiast. I built this machine that you see in this picture here using Fusion 360 and machined it with code posted out of Fusion 360. And I also in my free time build guitars and mandolins like this one in Fusion 360.

So the things I want to talk about here are some themes that emerged. First of all, we're going to look at it the way I think most people will probably try to use it at first, which is not necessarily the best way. But it definitely has its place as a simplified form of generative design.

Secondly, I think a really useful way that a lot of people will start to use it, especially people who are either constrained by time or maybe Fusion 360 is a very novice friendly tool. It's very discoverable and explorable.

And it's being used widely in education now right. So I think a lot of people will turn to this tool as a way to solve complex CAD modeling tasks if they may not be aware of how to solve a problem using surface modeling or something like that. And they're going to use this tool to solve those problems.

And then finally, because of the nature of this tool-- as Peter mentioned, it makes assumptions for you based on the proportions and the scale. And sometimes those assumptions, they are based on physics. And so sometimes they give you an idea that you otherwise may not have considered and might point you in a good direction.

At the very least, the forms tend to be visually interesting. So I think that's the third way to use it is as a concept generation tool, maybe not even with the intent of using it to create the actual final shape but just as a tool to explore what the design could be. So the first case here is using it as a simplified or a shorter ramp to a type of generative design.

And so let's look at the classic GE case study. This is a bracket that was used widely as generative design, traditional generative design was entering the market a few years ago. I'm sure you all have seen this.

And basically with generative design, you have to define your obstacles. You have to define your preserve regions. You have to define your manufacturing process and material properties and your structural loads and everything.

It's a fair amount-- it's a fair amount of research you have to do. It's a fair amount of modeling you have to do just to even run the process and even start the solver. How many features are here? Maybe 20 different features in the setup.

And the result is worth it because you get this validated range of options. You can get hundreds of options to choose from. And you have the integrated cost estimation, all that cool stuff. While the results from these things-- you've defined all these different regions and you have built in FEA, validating the results to make sure your safety factor is right. So all those things are worth it.

But now look at how simple it is. Using the exact same geometry, if you were to just run automated modeling, it is between one and three minutes, something like that, to start seeing your results and then another few seconds to generate the outcome. And look at the result. It's not that different. Some of them are a little leaner, a little bit more lattice like. But some of them are virtually identical to the classic GE model.

Now, there is no assurance. So some, there is no validation going on for the safety factor and those kinds of things. So it's going to be up to the designer and engineer to validate. But as a quick study, just to see what this might look like-- or imagine you want to use generative design, but you're doing a design concept and you want to show a rendering of the design to a customer and just say like, oh, well, would this style be visually appropriate. You could use it at that stage just as a placeholder.

But anyway, yeah, so it gets you a very similar result. And if you go take it a step further, give the algorithm room to play, and in fact just delete all the obstacles-- the way the team has set this up, this actually in some ways is-- I almost feel like it's more intelligent than the generative design in some ways, which is that it makes appropriate assumptions about where you're likely to have obstacles. The way it leans out the structure, it's a very fluid way of using the software.

So check this out. The result without the obstacles is almost identical to the result that we saw with the obstacles at least for some of the alternatives. So, yeah, it's just kind of a easy way to try out these ideas and get some results that make sense and with that caveat that it's up to you to make sure that it's really up to the task.

But imagine you're making a piece that's going to go on a lightweight drone or something like that. And nobody could get hurt by it because the whole thing weighs like 20 grams or something. Just 3D print it and try it. And if it doesn't work, then you can choose one of the other outcomes. And print it that way.

So I think students will like this a lot-- because a lot of times-- like a high school student who's designing something for a robot, they may not have the engineering background to use generative design, which is a garbage in garbage out kind of thing anyway. So it may not really be that different for some use cases. And it's just so much faster and easier, and it doesn't cost anything.

But I think once you get into it, the real value comes from these next two use cases, which is basically solving problems that would be tricky otherwise. So think about this, what if I told you, you could take any number of things in any orientation, any shape, and just click on them and then get a connection and not just a connection but actually six different alternatives, three main design concepts with two variations each? I would pay somebody to do that for me. And this takes two minutes. And you see the result.

And maybe you don't need something structural. Maybe you need something that can connect tubing together to allow for fluid flow. We do this all the time in biomedical engineering. We need to connect two things together. It's awkward. And if somebody isn't a really expert user of CAD, the way they do this is usually very clunky. It's usually like huge chunk of material that takes forever to 3D print, whereas this, you generate the outcome, apply a shell, and you have a tubing connector manifold thing that works.

So I think this is going to get used a lot for those kinds of things. And also, it's parametric. So you can move these things around now and regenerate the outcomes in two minutes, and you have a new permutation that updates to accommodate the new configuration, so very cool tool. Looking at this classic drone example from the Fusion example data set, look at this gold part here.

Let's say there's maybe five features present in this design. And it took somebody a few minutes, maybe an hour or so to design that part. I'm thinking about how it goes together. But in two minutes, you click the parts that need to go together, and you have a 3D printable part that at the very least will hold the things together.

So here's an option. This option here is pretty similar in terms of its size and structure to the original part. And I like the fact that you can do things like click on an edge and choose obstacles. And you can easily create connections that snap together or could be captured by two 3-D printed clamshell halves, things like that, using those obstacles to create your connections, really easy way to do it.

And don't forget, these outcomes, as Daniel was showing, they infer symmetry. So they stay symmetrical. And if you want to edit it later, you can edit this form that results and maintain symmetry with your edits too. So if you need to bulk something up selectively or add a detail or punch some holes in it, whatever you need to do, that's just as easy as it is if you had modeled it from scratch, so pretty cool.

When you start to get a little bit more sophisticated with this, you can use some obstacles to add some kind of stochastichness, some randomness to the algorithm. Because of the assumptions that the algorithm has made-- we're not really told what those assumptions are. That's the secret sauce, I guess.

But the beauty of that is that by playing with it, it creates this very playful approach, where you can do very small things, like this example. I'm going to show you this tripod. I'm just going to show you these three concepts here-- were the result of just taking the tripod and connecting it.

So these are three valid concepts. These are totally useful concepts. But then check out what happens when you just throw a simple sphere in there. And what I'm going to do on these concepts is I'm going to-- for each iteration I show you here, I'm going to move the sphere slightly. So that's the general sphere of the size and scale of it in the design.

And just leaving it exactly where I showed it there, the outcome goes from being this Eiffel tower shape catenary arch thing, to this interesting diamond shape with a triangular base. And, yeah, it's a little bulky. But, again, all these forms are editable easily. You could even go into the T spline workspace and cylindrify elements if you want them to be a little bit crisper and things like that. So that's one option.

By moving the sphere down a little bit, you kind of get this serpentine shape. This would be great if you were designing like a trophy or something like that and you want to engulf something to hold it. You can make like a cell phone holder or a GoPro holder or something like that. Just let it engulf the object.

And then if you move the-- if you move the sphere down still further, you get a little bit more of a pseudo tripod kind of thing that has some of the characteristics of the other concepts as well. So that's just such a simple example but a lot of variations really quickly. That probably took 10 minutes total to do that entire form study.

So, yeah, great way to leverage this stuff for inspiration. If you go a little bit further and you start giving the algorithm not only room to play but a little bit more information about a little bit more a richer context for it to play with, then the results get really interesting. So I'm going to look at this-- imagine we're creating a piece of furniture or a tent or something that has these tubes that come together.

And normally you just mock this up with simple cylinder primitives or tubes or something. But when you want to leverage this for aesthetic purposes, the best way to do it is to give the algorithm room to play. So shorten those up. Don't have them actually intersect, but allow it room to build a form on its own.

And when you do this, you get not only visually interesting and generally quite practical, from a structural standpoint, solutions, but you also-- then you still have the opportunity later to add obstacles and things like that to get that kind of quirky look going on. So that's how you start doing this. But when you start adding a little bit more detail and then like I said, a little bit more rich context for the algorithm to play around-- I just added these simple notches to these same end conditions.

I was thinking maybe I had some bamboo stakes or something like that. And you throw them on a milling machine and just cut a quick notch. And then you could 3D print these connections. Well, once you start doing this, you have more faces that you can select, so you can create more iterations quicker.

But it also is going to flow organically around those. And you get some really interesting visual-- I kind of think this thing looks something like one of MC Escher's creatures or like a spirit, the spirit of elegance thing on the Rolls-Royce hood ornament or something like that. And that's not something-- I didn't go out trying to create this form.

I just wanted to connect these tubes together and have something visually interesting there. So, yeah, that's basically the three main use cases that I think have a lot of value for this tool. And, yeah, I will turn it back over to Peter to talk about where the team is going next.

PETER: Thank you, Marty, and Daniel as well. I think a really interesting look from two of our insider group who've been using this tool for quite a few months now and a real wealth of tips and expertise in where this tool might be used and also some practices for getting the most out of it. So I'm going to talk a bit now about-- I've got a few extra slides there.

I'm going to talk a little bit now about some of the things that we as a development team have planned next, automated modeling, again, a reminder that none of what I'm about to share now is-- as of the time of recording in the software, this represents our plans, our kind of intended roadmap for what's coming. But of course, any of this might change. But this is what we're hoping to do.

And to give you a kind of a flavor for some of the stuff we have planned-- so the first thing that I wanted to show you was-- and this is probably one of the most requested things from people we've seen using automated modeling so far is you get an outcome from automated modeling.

And you can see, hey, this is an interesting concept and idea, and I like where that's going. I just wish it was a little bit more x or y. I wish it was a little bit more in a bit thicker, a bit thinner for example or in a slightly different direction.

So you can see here that one of the things we have planned is essentially a thickness slider. So here I would get on my alternative a slider bar. And that would let me just drag and drop that. I think Marty showed some examples at the beginning where, hey, the shape is what I kind of hope it would be, but it doesn't look strong enough, let's say. So maybe we could just drag that out make it a bit thicker and quickly move forward to get a different design.

The second thing that I wanted to share in terms of things that we're planning for the future is different alternative styles. So currently the outcomes that we get from automated modeling we can describe as very organic, free form style shapes that we model and edit with TS points. But we're actually working on a tool that would let us get outcomes that are more plate style designs, maybe representing, for example a design we could manufacture with welded plates quickly and easily rather than with something like 3D printing.

Again, a key thing here would be that the outcomes would be sketches that we could come in and edit. And we can also imagine how the thickness slider might also be able to interact with these plate style designs to quickly, again, iterate through different design possibilities, again, as I mentioned, the key thing being these designs we want to keep as fluid as editable as possible to come in, make adjustments, or even just start from scratch and use as inspiration.

So the last thing that we're thinking quite a lot about is that automated modeling can potentially serve as a kickoff point for deeper exploration through tools like generative design. So, for example, if I want to quickly design a component using automated modeling, I might want to then take that as a starting shape into generative design where I can then ask deeper questions about things like manufacturability or cost or put in detailed performance requirements that automated modeling is not going to take into account.

So as I mentioned, those are three exciting things that we are actively working on and hoping to bring to automated modeling soon. Now, if you are interested in getting early access to features like the ones I've just shown or to things like automated modeling as they come around, you can-- anybody who's interested is allowed to sign up to become a Fusion 360 insider like Daniel and Marty are. And if you sign up as an insider, that gives you often early access to new technology that's being developed as well as an active voice and say in how you like things so that we can-- and we really use this a lot to help us shape the direction that things develop in.

So, again, a lot of those features that you saw were things that might well have been mentioned or brought up by insiders as we first gave them early access to automated modeling. So if you are interested, there's a small link in the slides here that you can follow. But also you should just be able to search for Fusion 360 Insider program to learn more. So that is the last for me. And I'm now just going to pass back over to Marty for a final time just to share his final takeaways for automated modeling.

MARTY JACOBSON: Yeah, so just to summarize, I think what I'd encourage everybody to do is use this as a playful and exploratory tool. Basically any time you're not sure how to proceed with a design maybe as a modeling challenge or even you're just not sure exactly what the structure should look like, you could use this for anything at any scale as a way to generate form inspiration maquette scale, all the way up to a final product.

But just try it out. Worst case scenario, you have fun for 5 minutes. And you can kind of use it as a creative exercise. You might do a chipboard model or clay model or something like that for form inspiration as well.

I also think it has advantages over some traditional form inspiration exercises that you might do as a designer because it's based on this DNA of generative design, it encourages you to actually abstract your design to its essentials. It encourages you to think about what the function is. It encourages you to strip away stuff that you might have thought was necessary but may not actually be.

That's one of the main things that we see students and inventors struggling with is getting locked into their first idea. Well, use this as an exercise to blow the doors off it and just say, well, what might it be? And then finally, take it a step further and really, really build in that space for the algorithm to play.

When you do that, you take it a step further. And you can think about the mechanical considerations and really help your design process to be a little bit more focused on what the design really needs to be and then maybe get some cool form inspiration as a result. So it's a really fun tool. I hope everybody gets a chance to try it out. And, yeah, thanks for listening.