224: 'Everything Is Going to Change', with Martyn Day

Welcome to the Troxel podcast. I'm Evan Troxel. And in this episode, I welcome Martin Day back to the podcast for what you might expect. Another marathon conversation. Martin is a co-founder and director of X3D Media, which publishes AEC Magazine and Develop 3D. Two of the most respected trade publications covering architecture, engineering, and construction technology, along with manufacturing. He's also the host of Next Build and Next Dev conferences. And he's been at this for 37 years, building a reputation for saying the things most people in the industry are thinking but won't say out loud. He's sharp, blunt, and genuinely curious about where all of this is heading, which makes him exactly the kind of guest I want to have on right now to talk about the state of the AEC industry and the tech that is shaping it. In this episode, we cover a lot of ground. We get into engineering automation tools that can generate a full MEP layout overnight. AI written software that's already threatening CDEs and clash detection products. And the rise of open SDKs that are dismantling the modes that big vendors have spent decades building. We talk about what happens to Autodesk's business model when we hit peak-named user licensing. Why firms need to take their data seriously before someone else does it for them. And what it means when a company like Palantir decides AEC is worth owning. What we're all feeling that this conversation makes clear is the sheer simultenaity of it all. It's not one wave coming. It's a dozen waves at once and Martin compares it to watching an asteroid approach. The technologies are arriving faster than institutions, firms or software vendors can adapt. The AIAs plan, the architecture firms, billable hour, the traditional software subscription model, they're all in the crosshairs at the same time from different directions. And yet Martin isn't just sounding the alarm. He maps out where the real opportunity is. Firms that own their data, train on it and build spoke tools with open SDKs are going to have advantages. The big vendors can't replicate. The people in a firm who can break down a problem clearly enough for an AI to solve it are going to be worth their weight in gold. Or as Martin says, gold dust. He told me he's excited by the technology and worried about humans and equal measure. I feel the same way. As usual, there's an extensive amount of information in the show notes. So please be sure to check that out. You can find them directly in your podcast app if you're a paid supporter of Troxel Plus. That's just one of the perks. And if you're a free member, you can find them over at the website, which is of course, TRXL.co. So without further ado, I bring you my wide-ranging conversation about the state of AEC tech with Martin Day. I'm being tensed for 37 years. I think now feels like the most incredible time in terms of development. Super interesting idea. Yeah. I think it's very hard. It's very hard to be a software supplier. It's very hard to be a customer. It's really hard to be a journalist because he can't keep up. I write something and then someone releases a new album on Monday and that's great for you, Martin. It's great for you. I have to write quick. Yeah. That's my problem. By writing. Never ending content stream, you just have to pick what to write about. Yeah. I should play. Well, I think that's the intro to the show right there. I mean, that little quick history lesson. So you've been on the show before. I didn't look up when it feels like maybe three years ago. I broke that long conversation up into three parts. I don't think I'm going to do that this time. I'm going to try to keep you on track. I've got six topics that I want to talk about with you today. For this, for this, this is actually rates and notes. Because it's kind of like, I'm so spread out. I feel like I'm so spread out in my little research topics that I'm using track and they all interact with each other. But it's definitely. Interesting. Well, let's just start with a meta conversation about that. How do you keep track of the things that you want to write about and kind of how do those? How do you control or make that connection between them? How do you keep that story straight across those topics? I'm a non-linear thinker. So I'm quite a debt to taking lots of inputs and then suddenly something where might be in my sleep. But then I'll kind of wake up and go, "Whoa, and then I go off and connect two things together." Oh, you sound like an architect. I think I think it's ADHD. I don't know if that actually is. There you go. My sister, most of the schoolteacher, she's giving me a league diagnosis at the age of 50. So I take you lots of signals. I look for signals. In fact, one of the first agents I wrote was to go out and scrape every week forums, industry websites. I don't know how to do it LinkedIn because they're a bit weighed about it. But every week it kind of throws up, it'll throw up some things. So it knows me. So it'll then say, "Hey, you wrote about this, this, and this." This actually impacts that. So I've now got a little friend that kind of helps me. But the forum is kind of just talking to people. I talk to people all the time talk to the bin people. And I go all the time. I talk to, if something is particularly interesting, then I'll trace them down on LinkedIn and start having conversations. So things like consignely getting bought for like a huge amount of money. Oh my god. I get people contacting me, go, "Oh my god, so much money, consignely wow." And then they give me bits of information. And then I'll see Andrew goes and gets 20 million for her startup. And I speak to them. And then so suddenly I've got augmented who are obviously I've been talking to for a long time. Andrew, who are new. I do a search and say, "What else is out there?" Come across them called HVACO. What are the guys? This guy does a crazy video on LinkedIn, like messaging really lovely guys. Have a really good chat with them. I'm doing next build. I say, "When you come over, hey, Andrew, come and speak, hey, augmented come and speak." And then someone else says, "Hey, have you seen that company?" So I'll go and look at Branch 3D and I'll talk to those guys. And then suddenly I invite them to next build and we've got all of them there. So we can have a really interesting conversation about where engineer is going. And once I start looking into that, that then makes me think, all of these plan of works that the AIA has and Reaper has. That's so old compared to what's coming. These guys are going to be nailing a engineering solution either overnight or in real time. Yeah, the pace. The pace of development of those frameworks that you're talking about is, I mean, in the you compare it to what's happening right now in software development, mostly around AI, right? It's like you can't compare them. And so how you're right, like those look like syncing ships, not even just slow ships passing into that. Well, augmented has taken a long time to get to where it is. And like, there's the whole thing, is it AI? Is it generative? I don't really care anymore. It's just, it's it's generative, mainly. But AI is playing an increase in the role, the big role in it. But augmented are only focusing on one country, the USA, they focused on one thing, which was electrical and they spent a load of money and time getting that right. And now they're focusing on MEP, which was always their intention. And I saw a demo of that in December and it blew me away. And they said, you know, this is 12 months before we got shipping product, but they're already throwing a reviv file. You get a MEP built into the reviv file. And that's the way that shipping that shipping product is not going to be a bare bones basic product. It is going to be fully do your building. And it will do electrical and MEP simultaneously. So you'll you'll you'll truck it in for a solve seven hours later, you'll get both done. And then after that, they're going to do structural. And then you look at what Lucas app is doing with structure craft stroke branch 3D. He's got a real time F.E.A. solver. So an architect moves the floor plate. That floor plate will automatically recalculate the F.E.A. analysis. And then it will reposition the structural elements. So you have an immediate response to an architectural movement. And at the moment, that's not really good. That industry is not geared up for that. It's not geared up to connect those people constantly. And that's you in a multi multi discipline firm where you have really tightly integrated folks. So so just looking at this this this engineering conundrum that's going to hit the market. I mean this stuff. I think I think branch 3D is already shipping. But obviously this is going to expand out over the next 12 to 18 months. And you're going to see engineering become a lot more automated, a lot quicker to get to get your feedback. And what the hell is that going to do to the market? You've got your marhabit architect still spending 18 months designing a building. But the engineering is going to be almost like press a button here it comes. You're going to need somebody to suit. So there's going to be an engine. I'm going to go through my through my topics in reverse order based on what you're going to do. about because that's this kind of solver-based stuff that we're seeing, right? It's like the architect, but I don't even know how to frame this anymore because it is changing so drastically and so it really kind of revolves around the business model of architecture, billing money for time. Like, destroys the architecture business model. They have to rethink it. So let's start there. Yeah, so many people say that. We have to rethink it. But like, what do you, so you're painting a picture of why? And so, I mean, there's the why, but there's also the how or the what? Like, what does it look like? And what's really interesting to me is that it actually gets back to the value of architectural design. It doesn't rely on the commoditization of, this is how long it takes to do drawings and drawings are the output. It really points back, I think points back to the idea of where the value lies in architecture. I would love to hear what you think about that. First thing, the industry is being attacked from many different places. It's not being attacked by one thing. It's going, it's being attacked by automation in the traditional workflows that we have. But then you've got a solver, which is above automation. Solver is going to like, here's the MEP for the building that you designed that's really bad and I had to put all this MEP, the the software said you had to put all this MEP in because the heat load on this building's crazy. So you've got that. You've got companies like Palantir coming in. Holy crap. I mean, I'm for me, I can't stand the company. I find them despicable, but they've got an AC division. And that AC division doesn't, doesn't care about them. All of the traditional software, I've got like, lock some people, they're called moats in, is the kind of con thing. But startup language. Yeah. Yeah. RVT is a lock. You know, ACC is a lock. DGN is a lock. All these kind of formats, which just kind of stop you from having ultimate freedom with your data. And Palantir, I'm just going to go to a Skansker, embed themselves in a Skansker. They will own the layer between everything downstream and their business processes. They will learn everything that happens in a Skansker project. Everyone else suddenly just becomes piping to send information to the Palantir system. And Skansker will be incredibly efficient. What happens to construction managers when you have an AI layer that's removing risk, that's automatically removing risk. And it's automatically panning. Yeah. Proactively looking ahead. Oh, this is open claw for construction. Yes. It's like GPS. It'll be looking forward. And so suddenly, well, where does that leave the existing players? None of them have AI development teams or the AI capability or the ontology. Yeah. The Palantir, the one ontology is kind of a, if you don't know about it, it's essentially how everything works, the ontology of a process, how everything works, how everything is connected to one another. And if this changes, then the impact of that change, not just in a kind of like a reddit parametric things, like that geometry must change, it's kind of like this has happened. Therefore, this, this and this are all impacted by that timeline. Therefore, I will act do whatever. So, and it's go based. It's like, it's not going to stop at what you would traditionally even think as solutions for those problems. It will find a way to solve the problem because that's 24/7. I'd be working on your projects. And so we can all, depending on which kind of, it's almost like a, a, it's kind of fractal because we can all concentrate on all being, well, I'm an architect and I've got a rabbit and, you know, I'm using ACC and I've got issues there. But actually, go out, go out, go out, go out. This onion, you can go to the outer onion skin and everything's being attacked by this. And that's why it's so hard to understand the impact of this because everything's changing at once. Or everything, the technologies are changing at once. They haven't really impacted yet. But we're just kind of looking at this asteroid coming towards us and we've got to try and work out what's going to survive in terms of the workflows, the technologies, the products, the, the, the institutions re-bur and AIA have to get their act together and understand how these things are going to impact architectural practices. And I, how do you charge, how do you do build, build by hour when you have automation and you have, you have things being solved overnight that would take three or four months? How do you charge the customer? How do you keep track of that? I think there's a move with automation for drawings, automation for design, even that will, that will make billable hours just, just be useless. So then you have to start thinking about what your value is and how you charge for your value. And by the way, the software companies are thinking exactly the same. They know we're kind of near, we're near peak, uh, name user license because from now on in people will going to be using automation. They're going to be using smarts to reduce the number of seats that they need. And when that happens, then these guys, these guys are going well, I bought a desk on a seven billion dollar company. I need to be a 10 billion dollar company. Oh, I might have a problem trying to stay being a seven billion dollar company. If this revolution impacts my software sales, therefore I need to do something a new business model to, to keep bringing that money in and then grow into new areas. There is new opportunity for all software vendors as well. But essentially they're going to have to do it. So what I'm seeing is tokenization. I was going to say tokens. Yeah. Tokens are coming in just like AI, everything running on tokens. And yeah. And this is kind of one of the fundamental things that was software has moved to the cloud or moving to the cloud. Right. Okay. Somewhere else is computer and you've got to pay for it. So not only are companies like Autodesk charging you for a subscription. Let's imagine that Rev it was on the cloud rather than it's on your desktop. But they're charged by Amazon to run that instance and do that processing. They put a mark upon it and then they pass it on to you. So they're making, they're making money not only from selling you the software or subscribing to the software, they're making charging and making money on you actually touching it or using it because it's now hosted. Now if you go out a little bit further and you start looking at APS and the whole idea of a authentic approach is they're now going to charge you tokens based on how many megabytes you access from ACC. So you're paying to subscribe to get the design software. You're getting to pay to have the stuff hosted with the services on ACC which aren't insignificant. And then on top of that you're paid to touch your data to then process it with an AI. That AI might also charge you for processing as well. It might be on your side with the firewall. So pay pay pay pay pay pay. It's it's it's it's got to be a bad thing. So that's why I think the net result of this is that large firms are looking to resure their project data to have it in house. If you have it in house on your server you don't have this kind of constant bleed of money and you could try you could develop an LLM on your data which doesn't have high token usage which you're paying for the nose for. So I kind of think that that that firms need to start thinking about where their data is, how good their data is, can they centralise it, can they protect it and then what is it that they want to extract from that and have they got the in house knowledge to do that or yeah they're going new roles in firms where you're going to have to have someone with some kind of a genetic AI approach. These people have to be structured in mind. A lot of firms don't hire programmers. This episode is made possible with support from a veil. Is your firm managing architectural landscape, interiors and engineering content. Avails all in one content management solutions stretches across all disciplines, file types and content locations. So production teams can find everything they need all in one place. Thanks to avails desktop application and rev it, AutoCAD and civil 3D plugins teams can consistently find and reuse all types of content from specifications and calculations to 3D models and plans. As a result standards compliant models and drawings are easier to create, the documentation process is less chaotic. New hires are easier to onboard and projects are delivered faster with fewer errors. Visit get avail.com to request a demo. That's g-e-t-a-v-a-i-l.com to request a demo. My thanks to avail for supporting this episode of the Truxel podcast. Now let's get back to the conversation. You have to think think about data differently. It's interesting to even think about you are then competing against these companies that you're talking about who are developing their in-house and they're paying way more than architecture firms pay architects, for example. I don't know how much architecture firms pay data scientists, but it can't be anywhere I can't imagine competitive with a planet here. I mean, not that you have to shoot that high, but they're so great. But even in an audience, in any software firm, so some of my friends who've gone off from architecture to join software firms are amazed at the amount of money that software firms pay people. You're an important person because there were a very few of them and they developed something that has a good margin on it. And there's a huge market to sell to. But I, there everything is going to change and it's going to happen a lot quicker than I think people are looking at it in terms of, in 18 months time, well, I'm getting calls from people who are programmers who are saying that AI programming is getting so good that they're going to be at a job within a year. And if that's the case, that level of programmer is going to be available to every single AEC firm on the planet. And all it takes for you to do is to have, you need to be able to ask the right questions of the system to develop the product. This is the big problem, is understanding what your problem is and being able to explain it to the AI so it could give you a a program. And I think a lot of the programs are going to be ones EC. So, right, you have an AI, you have hopefully someone who kind of has a structured mind and says, okay, we need to develop a CDE, we wanted to do Clutch Detection, we wanted to do X, Y and Z. So, you specify it now, every time, if you're just sitting in front of Claude and say, writing your CDE and here's the spec, it's not necessarily going to write the same piece of code with the same stuff in it every single time. It will be unique. There's no framework. It's just off you go, go and, and whatever it does on me, I want to research at that moment in time, it will go and build it differently. This is why a lot of the, I think that some of the most important firms in the next five to 10 years are going to be open SDKs. So, that open company, Creox, ZioKit, those things, they're out there and that open source, they're free. And if you point, or you tell your AI to use that SDK, you get common frameworks, common user interfaces, access to file formats that are standard like IFC, it's all built in. So, when a company develops something, they develop a product that, and then they develop a next product, they will all have the same look and feel. And you kind of end up with creating your own software, but you have a standard. And that standard is this open, these open frameworks. And these open frameworks are starting to crush data modes. I mean, things like CDEs are in real danger because that's, that's going to be easy bread and butter for an AI to program. Clash detection is another one, which is another problem, which will be solved. Things like Cilebree, I kind of have worries about, because model checking is another thing, if you can write rules, put your rules in the AI, then it can then put that in the program. So, I think you're going to, you're going to find certain products. I've already had software firms reach out from that. Just that little thing I put on, then, Dyn saying, I asked, Claude, could you rewrite ACC in Open SDKs? Right. And it just, you know, kind of last year, it said it was difficult. This year, I kind of said, yeah, I can do it, but there are, there are certain things that it couldn't do. It could do 70%. And so software companies are coming to me going, was this mean for me? Was this mean for my product? There were going to be some low hanging fruit that are going to be struggling, because if you really wanted to, you could have an AI write that for you specifically. And that will be from next year onwards. You're still going to have to understand managing code. You might have to have some idea about user interface or how you like it. Code has to be kept up to date to run on various standard software. So there are some, it's not, it's not straightforward. You're still going to need to have some knowledge. But people like foster departments have 30 or 40 computer programmers. And a lot of these other firms have got zero. They're relying on an architect who happens to you, have studied Python because they're big, big, big, big, big, big, big, big, big, big, big, big, yeah, that's like a YouTube. Yeah. That person or those people in your firm who have that mind, they're going to be like, gold dust, because not only are they going to be able to help with the grasshopper scripting, whatever you're doing, but that that idea of breaking down a problem is going to be, and then explaining it is going to be essential in firms to create tools for the for the for the company. And they'll be saving a lot of money. And this is the other thing is that firms will be able to start looking at their software estate and start deciding what's important to them. I don't think offering tools or in any real danger. I think that BIM2.0 and I think Revit and off you can't you're not going to be able to have AI right right you are Revit. It's just not going to happen, but it will do CDEs, model checking, cash detection, those kind of things. But if there's a side plus the side of it that you talked about earlier when you're talking about how how you're synthesizing all of these like it knows you your tool that you use to do research knows you. And that's the other side of the coin that these firms have to develop. It's interesting to me to think about like this idea of like the billing structure, but there's also the phased structure of our projects and how they're linear like seed conceptual design schematic design design development and how compressed pieces of that get for sure depending on who's on the team, right? And who on the team could also mean these kinds of solver apps, right? So it's really interesting to think about like you said everything is going to change. So it's not just the tools that we're talking about. It's the way we do the work. It's how the time it takes to deliver the work. It's like the value of the work. It's so many layers deep. Just thinking about the one that I absolutely know is going to happen first, which is engineering. His hand architects work with the structural engineer and what's the timeline between design and then giving it to the structural engineer and the structural engineer handing it back. And then the way that the way that the core studio did with asterisk years ago, right? And it was kind of too early. It was too a big idea a little too soon. And now they're on the verge of doing it again. And they're doing it to themselves because they know these solvers need to exist. And they need to solve 80% of the problems so that they can work and focus on the reality. Why did they call them? I can sickly for $360 to $90 million. I mean, that is insane. I had VCs ringing me up. I had software vendors ringing me up. No one could believe the price. And then software vendors were worried that suddenly they were going to be competing with their customers to buy AI startups. And that is going to be the case. I think Kaui are on record. Yeah. Kaui, Skanska, A-Con. And I don't blame them because A-Con, if by looking at what was happening, they're going to be in competition with other people with automated systems. So that makes, that's the reason why Endra got a huge VC. They were turning money away. They told me. They were turning money away because they had so many people wanted to put money into Endra. And then you look at Augmento and they're growing. Augmento, the kind of secret giants because they've been working on this for a lot longer than anyone else. And they have, they're a lot more holistic and they're there. Much more press of button and seven hours later we've done something quick. Not only have they generated the electrical, they've done all the conduits, they punched all the holes. The MEP is exactly the same. All the support structures in there to put it's done it from the Revit model. And it's all happening outside of Revit. And this is the thing that's going to happen with BIM. Is that BIM is this kind of thing where you kind of do some Lego modeling and you kind of like building this thing. Well, meanwhile, that data goes out because it can't do it inside of Revit. It needs to be in something more modern and AI friendly. And that's being pumped back into Revit like a puppet just to create some drawings. And this is the feature of BIM 1.0 is just going to become this kind of drawing conduit for geometry. And I think I don't know how long that will last for. I come across firms that are spending it and know it's not about time and effort trying to perfect their BIM 1.0 platforms to really refine it. Yes, it's like you're wasting your money. You were trying to refine something that's already dead. You have to start thinking about these tools that are coming out and how it's going to change the way you work. And I was concerned that a lot of the BIM 2.0 guys started too long ago for AI and the authentic future. I asked them all and I'm kind of okay with a lot of the answers that I got back from people in terms of what they're doing. I do think that they have a handle on it. I think Motif being last to the party are the ones that are going to benefit the most from it because they were still deciding their text stack last year or really for a lot of stuff. So I think they've got a handle on it, a conic have a handle on it. Snacktruth has already been demonstrating some very interesting conceptual design features. I think he's kind of ahead in terms of delivering AI at the concept level. Darkhall has tended to focus on really that one platform that it's got but I know full well that Paul and the gang, their total focus is to replace Reb it at some point in that kind of detailed design and drawing stuff. But the question is going to be, is there really going to be one platform that does that in the future or is it going to be specialist developers like that? It's going to be specialist MEP, Span electrical, it'll be sort of like specialist facade, it'll be auto drawings is coming along. There's now more people involved in that. So are you just going to have your model in whatever you've designed it and just throw it into whatever drawing service it is and then you get all your drawings out as opposed to having it in the model. It seems like a lot of, at the moment it's still very file based so that's another kind of problem is all these files being sent everywhere. It's just, they have to be in the cloud. It has to, it has to move to the cloud. But it has to happen in a way which I think the industry needs to recognize they have to protect their data. They have to try and minimize the cost and they don't want other people to train on their data and I think that in house is going to increase. I think people are seriously going to be looking at buying some big hardware, big server hardware in house as opposed to paying and paying and paying and paying for every single thing online. Where ACC and kind of project hosting sites will win out is you're working on a collaborative project project. So you have to put your data somewhere together. So okay, ACC is maybe the place that you do that. But in terms of your archive and other things that you're working on, you really need to have that on your side of the fence with your own intelligence running on it. Because this stuff is really so low cost to develop, it's already free to develop. I wouldn't be surprised if architects start sharing applications that are AI's of written. So why reinvent the wheel? Here's a CDE, here's a crash detection thing we wrote, works really well. Maybe there's a secondary market for that. But the very high end architects are totally rethinking their product mix and the way that workflows go from simple sketching to sketch. You get some nice AI renders from the sketch, then you get a 3D model from the sketch. Is that now viable? And then you in Rhino. Rhino is the one thing that seems to be consistent throughout. Everyone, it's so cheap for what it does and it's so beloved. Is the product could know when they said that they're going to remove from the thing. And then people try to work out how to get drawings. Or if they're going to stick into Reddit, how few rebits they can get away with to produce those drawings as opposed to having a Reddit on every single desktop. So that you are using it primarily for an output of drawings of a model. So yeah, I'm intrigued every time I talk to a company that's doing some internal research as to what they're all seem to be roughly heading the same direction. And it's about hosting their own project information. And then that opens up. But if you've got a server with all of your data in it, then you can run the application on your data. You don't need to put it in the cloud. So you can write your own agents. You can, so at the moment, I've talked to Greg Schluzer, Elijah, OK. He really wants to have real-time class checking. He said that his thing's been on. He's burnt bare as people work on a project or a week and on Friday and do a class check. And the class detection goes, you know, this is screwed up. And that's some wasted to hold out of work. Why isn't that happening in real time? So that it doesn't-- Yeah, why isn't there a class detection recommendation engine that's working alongside you as like an agent would, right? And you can't really do it in Reddit because Reddit is single-threaded. And he told me that every time you need to save something out, then it's up to go pause. And then output that object and then move on. And then it-- I'm hoping that we're going to have these real-time systems demonstrated soon. So you can see the benefit of having your data local and these kinds of agents constantly running over it with no real huge token cost to then give you immediate feedback and flag things up as they go along. But everything at that point, Dan stream, the same kind of agents running, your de-risking the whole process. Right. Because you're checking models. Exactly right. You're checking models. You're checking everything. That's right. And then, yeah, well, and then what happens to insurance? So if you've got these tools and you're not guaranteed, but you've got a high, high feeling that your data is always going to meet the quality. So maybe there's no going to be suing going on, Dan stream. Then, yeah, with that litigation, that's going to be obviously a benefit to them, to the market. I mean, these project hosting sites should be trying to stop future legality. If they're hosting the project, they should be looking at all the data and immediately flagging up if there are kind of anything missing. So when we get to the point of currently being sued, that's been handled before. So it smooths the process. And that's why I know the reason why I think kind of construction managers are very worried, because we're going to have these software applications of loving grace looking after the whole process. Let's take a short break from the conversation to invite you to join the most influential technology leaders in the AEC industry at Confluence composed of in person events and a podcast co-hosted by yours truly. Confluence is designed to foster conversations between AEC firms and technology companies so they can learn, share and engage with each other to support industry innovation. Software company Avail, which creates content management solutions for the AEC industry, started hosting Confluence events in 2019 to understand what firms are needing, wanting and thinking around technology. To learn more about Confluence, explore upcoming events and listen to podcast episodes, go to Confluence.getavail.com. My thanks to Confluence for supporting this episode of The Troxel Podcast. And now let's get back to the conversation. Well, let's talk about the people side of that equation, right? I mean, you're talking then about needing people who have tons of experience on projects, actually getting projects built, right? And so that, to me, is concerning from like the, like obviously there's like the technology layer, but then we've talked for years about those people walking out the door with all of that knowledge inside of their heads and newer generations coming in without any of that because they haven't done it yet, right? And a lack of mentorship and things like that. So, I mean, how do you see firms dealing with that and what are they doing anything to address it? Well, they're trying to capture it. So if you look at products like Victor, where if you've got anyone who writes any code in house, then the idea by using Victor is that you say, they give you all the tools to write the code, but they also store it and then they produce it as a piece of code that the rest of the company can use, hopefully you document it and it actually makes sense. So there's this kind of knowledge capture a little bit going on now. Then AI is going to be able to reverse engineers some of that knowledge or capture some of that knowledge. People like Palantir are definitely going to capture a lot of that project running information once they get really stuck in. So that's, yeah, I don't. It's a lot of that information is exchanged just over conversations or it's real time on the job site Or you know, and it's not captured and nobody does kind of that retrospective Well, not nobody right but but like going back and saying well what happened and why did we decide that and like a lot of that stuff Just falls through the cracks because it's never needed. It's never previously been needed It was just one of those things we we just did in real time Palantir gets enough information to go and kill people I think they can kind of work out What isn't isn't necessarily a project. I think then it's gonna be reading every single email It's gonna be looking every single drawing is gonna be able to be kind of in for What's meant in a red line of a PDF kind of a thing? Eventually it will understand everything and The more you train it the more it's gonna know and I just don't think I Don't think that there is much knowledge that can't be captured over time and the more information you give it the easier is and My my deepest concern is for the next generation because what's happening in in legal Definitely haven't in engineering but not because of AI because of sheer stupidity is if you don't keep recruiting people in and training people up What will happen is what happened in engineering like not build not building engineering like engineering engineering I also run developed 3d magazine and direct 3d live the show. I Go to these events which were all about engineering I remember is just beyond the bus going back to the car in the car park and there's this old guy there We were just like chatting and he's a said the average agent my company is 56 Hmm, I was like wow. They said we're really worried that there's nobody below 35 is actually worked on a project that has got a project done completed understands the whole thing and He put it down to the fact that they Didn't have a apprenticeships for a long time the government introduced some tax breaks. They took on some apprentices Great then they took the tax breaks away. They got rid of the apprentices And essentially what's happened is you you've got a functioning being which is the but the company But everyone's aging out and as they start aging out if they have a recruiters at the bottom Then you've literally annihilated yourself at some point that that business doesn't exist Then you multiply that across all the engineering firms in the UK and the average age is very high So our engineering is in real problems It doesn't matter how much money you shove at them How are they going to find somebody now to want to go and study engineering as a course versus business studies? Which is engineering is a tough you have to actually know your maths and then once you get out someone needs to put time into you to Teach you everything then you have to a crew business knowledge step up and in the company This is like it's a it's a 20 year right process and how many how many firms have that mindset and I think I see AI coming along And in the legal profession it's very hard to get hired now To they're not recruiting anywhere like they were because the easy jobs are where AI is killing Exactly and so you deploy it you don't need to hire anyone. I make more money brilliant So that goes on the number this goes on and the better the AI gets the AI is going to creep its way further and further higher into the capabilities of Law is a solicitors and so you might actually start getting rid of not really knowledge workers who've done stuff But if you if you have recruited at the back end your business is is is is gonna die not that I like Elon Musk and his his views of I don't know what his view is but he keeps taking the white people are killing themselves in if you in businesses If you do not bring be keep continue to bring people in the base your company As a finite life and I that will happen because of Yeah, the same reason why engineering firms handed or offshore and all our engineering to China and so you might have some designs the product in Colorado but it's actually every all of real knowledge is in China for making it and eventually They the Chinese are gonna learn how to make we eat good products like cars and so then We don't have the engineering base and yeah, I mean this is yeah, it There was there was there's not a lot of joined up thinking Going into the next generation and instead of being my it in debt Encourage to go whatever even though if they're studying something is not particularly super useful And that the meanwhile their jobs have been have been off-shored to China or are gonna be AI automated out Yeah, and I don't really wouldn't know what to say to anyone's kids at this moment to go on study I trade I Trade I I do yeah, there are because it's true. It's just true then You don't know if if the next kind of kind of trover Until they've actually gone for the process of studying that and then practiced you can't tell which kid is going to be You know the next big thing and I am horrified to think that we're gonna end up discouraging Greatness or the next greatness that we need An AI in some ways almost some looks like social media. It's just mind numbing because it You can get into the to the part where you're not asking you're not engaging with AI You're just accepting the answers and then that's your all. Yeah, that's your output and I Think as humans as a species were a bit lazy when it comes to that and I think that Yeah, it's more of a danger than Facebook is to people's mentality But I do hope that the people who do persevere and go through the process and think they're good enough Can find a job in the in the industry in the future We're definitely gonna need more civil engineers people who are physical and out there and you know learning how to build projects Right. Oh, that's that's that's pretty the biggest Physical thing we can have you seen this website renta human renta human dot AI I think is what it's called where AI is actually need a human to do something I can go take a photo of a thing or go deliver something to somebody and you get paid a hundred dollars an hour I mean that there's the future for the for the generations will be slightly do the physical things the AI's need Yeah, I yeah, I'm I'm excited by the technology, but I'm depressed by by humans in our inevitability of and some way cap it you know It really confuses me that a lot of people are selling this feature our billionaires and You know, I think musk said that would all be millionaires in the future I think I think so Capitalism for fails basically when when AI and robots are doing everything I think it's been said a number of times It took a hundred years for the industrial revolution to sweep through This thing's gonna hit us and it's gonna be a ten year process and we're just not ready If you I think all half of jobs in America are driving or related to driving So if musk gets his way and you flick a switch and suddenly everything becomes automated That's that's Who where are they going? Well, they're gonna do what how do they retrain? And then you go to Amazon and Amazon's gonna replace the people in their pick and choose things with robots Well, what are they gonna do? There's not there's not a holistic approach by any government to fully understand what we're gonna put ourselves through and I was a technologist I'm kind of excited, but as a human I'm I'm Deeply worried. Well, yeah, there's like that day-to-day kind of excitement of development Cycles and how quickly and then there's like you step back and look at the bigger picture And I think like this this this happens on many levels like you're talking about it government-wide you're talking about it Organizational like a I a reba level and then there's and then there's firms and And it's like all of those Organizational levels have to be doing that all the time now, right? It's like a new model comes out You know all the time a new a new function comes out all the time and it's like the companies that are moving at this pace Are constantly at re-evaluating where they're going. They're not just saying this is how we do things This is how we've always done things that that doesn't exist in these companies that we're talking about But that exists in our industry to the deepest root, right and and this idea of Creating a place where people want to be in the future doing this kind of work I don't like you have to step back and you have to look at that holistically like where are we going to be? How are we going to get there who is going to be a part of that and create because architecture firms didn't ever ever have to use they never had to Think about this there were people lined up at the door. There were graduates lined up at the door of architecture firms because they got a degree in architecture. Now that degree in architecture can serve so many different functions. Architecture firms are competing against technology firms and all kinds of other things for those candidates. And they may not be feeling it right this moment, but this is like the near term future that I see, right? Especially on the coasts, especially in the urban areas where that's where the technology companies are to, right? And so you might be a little more protected if you are in the Midwest, but it's going to ripple through everywhere. So I don't know if it's short or state, but over here, a lot of your traditional run of the middle architecture practices will give away the design. So they're going to they want to win the job. So they'll design pretty much right up until detail design. In a year or two, the systems that will be available will be able to leap, literally leap from a concept design to level 350 and a press for fun. So at the moment, people are kind of going up just before the real work starts and just to get the job. And so they're kind of punting time and energy on there. But what happens when you're a company that's using, they don't really care about the, oh, yeah, they want to win the business, but maybe they're charging based on the value of the building and they're not going to charge you billable hours. And they they've refined their process to the point where they produce a pretty nice looking building, all the renders and stuff, but they've already done the level 350 drawings by pressing a button. And so when the customer goes, yeah, do it. Okay, wait two months and there's a year of drawings. He is, he is a business, pay us the money. Oh, we have a thing to manage the site, then pay us the money. But he, Pete, that the, the business models that exist today will come under pressure from some firms that run with this stuff. And the question is, has it dev, will it devalue the designs? I think there is a, I think there is a, a physical limitation as to how quickly a client can understand a design. And so you might have a client can, you win a, you win a, a competitive bid, the client goes, okay, build a building. The next thing is, oh, the guy's changed his mind. He wants to, he wants an atrium. We've already ordered a steal. I mean, this is, this is like the future. We, we, we've designed it. We've given you the drawings you've okayed it. We've started ordering stuff. The client might need six months to a year to really fully seep themselves in your design to go, yeah, this is great. Or we've been thinking we have the meeting and we think that this should be looked like this, this, this. So, maybe there is a physical limitation to how, how quickly clients can fully appreciate what it is they're going to site off on. And if, hey, I, what swung in and within a month and a half, you've got everything to level three 50 and it's required a small team of 10. What, what role does the client have? Hey, everyone, I want to take 30 seconds to talk about something that really matters to me. This podcast exists because I believe this industry deserves better conversations. No agenda, no corporate backing, just honest in depth dialogue with the people actually shaping where AEC is headed more than 25,000 of you listen to this show and less than 50 of you are direct supporters of what I'm doing. And I'm not saying that to make you feel guilty. I'm saying it because if this kind of independent media matters to you, if these conversations have ever changed how you think or introduce you to someone worth knowing, then the most direct way to keep it going is to become a Troxel Plus member, 12 bucks a month, add free episodes, the full leadership edge newsletter with my deep dives and the things I don't say on the podcast and the knowledge that you're one of the people making this possible head over to the website at trxl.co and click the join button. And either way, I'm really glad that you're here. Now let's get back to today's conversation. Well, okay, so yeah, let's talk about that because that small team nimbleness and maybe tech savviness that because you can assemble this team today, right? Yeah. And and compete against these largest firms out there with with an entirely different tech stack mentality way about, you know, approach, billing structure, the way you can kind of spin that up like like a startup gets spun up and to to test a market fit for a product. I mean, that's basically what you're talking about and and why can't it's going to be why shouldn't architecture firms and engineering firms operate. They will happen. There is going to be commoditization of of of random butter design, brand of butter designs, office blogs, maybe not, maybe not a hospital, but can like student buildings. And those kind of things are going to end up in the already got it. So with high arc, I love high arc. I love what they do. And but the idea and what you're seeing there is a expert system. And I truly believe we're going to end up with expert systems for designing very specific building types. And they will be incredibly rapid and incredibly autonomous in generating those designs. The problem is is that even for foster partners, 50 percent, their bread and butter is is probably buildings that no one knows anything about, but they be commissioned to use their foster partners. So I think I think they don't talk about them. Yeah. No, there's no name on them. It's just it's just and those are the kind of buildings that are going to be in danger. And that threatens the size. I don't think it threatens the companies themselves, but I think it threatens the size of those doesn't threaten their brand. Yeah, absolutely. It's it's it's something else. But you're but yeah, it won't continue to need to be that big to do with it. It will be more boutique. And I think you've got the large really large firms, you know, the H.O.K.Z.S.O.N.s KPS. You know, they're going to be in the threat because a lot of what they do is is bread and butter. If it if they've got niches like hospitals or labs or something like that, then I think they're going to they're going to develop their own kind of high arcs for designing those buildings because to get if they know these people know what the problems are with with those styles of building. What a test. I don't know. And they and with the ability now to have the spoke applications written for them, then they're going to and there's enough open SDKs and there are, you know, things like I twin was it I twin studio with Bentley and then it's the potential for them to develop their own their own BIM applications. Well, come on, BIM. But actually, BIM, it's so loaded with this to thinking about what it was, but their own design applications, I think is huge. And then you then yeah, but then you got very unique firms. You've got firms are highly specialized and that's the company you go to hospitals and that's the company you go to for student accommodation. And I think I think niches need to be carved. And it'll be it'll be a bun fight, but also, you know, there's this help capitalism thing hanging over is how profitable they are. There's they're going to be trying to be as profitable as possible. They're going to try and curtail design time and people and software as much as possible. And they need to work out what their rewards going to be if they can't really do bit of hours. What's the accepted way, the new way of working out what that is. And it's going to be the same for the software companies. So there's an article coming out. It should be out by the time this goes out, which is all about AI in the agenda. The agentic feature of BIM. Well, I spoke to Andrew Allen Nelson. He said, you know, we understand things are going to change and we probably need tokenizations one of the things, but then maybe we start charging a fee per project. Maybe there's a slice of that kind of project fee that goes to pay for the technology partner providing the technology as opposed to named user licenses or whatever. It probably could be a combination of those. So everybody's trying to rework what this is. I mean, anecdotes took 10 years to kind of like refine order desk sass. And then suddenly AI comes out. And then everybody's share price gets whacked because that's old software. Sass is dead. It's every single part of the business landscape is being changed by this. It's share price. It's how you pay for software. It's how fast you design, how you charge your clients. Well, what software you do buy, what software you do develop. If you have your if you pay to have your data hosted online or if you pay to have a kid. What are the roles? How do you do your contracts? Yeah, everything. So every single practice should be trying to understand what the dynamic is and start making five year plans, three year plans even. So they're kind of understand. I said five years, I was like, well, that's the old number Martin. So the other thing is that in 2028, 2029, I think all the desks two for one deal runs out. So eight years ago, if it if I was in 2028, so in 2020, they gave people two seats for one if they were moved to subscription for eight years. So that's one thing on the horizon. And then I think also there are some people who are running some very low cost copies of AEC collections because they're historic and they got a special deal for upgrading. That at the same time also gets next. So so there's a kind of a we're heading towards a brick wall in our traditional world. And I wouldn't be surprised if all the desks had some kind of offers coming out. Maybe to try and help help with that because it's it would be a hundred percent price increase. If you've if it's two for one and you've got double amount of software, most people out after eight years are probably using all of those seats. Maybe the manager, the bin manager doesn't know the deal was done eight years ago. So I think it was called M2S. I can't remember what the exact term was. But so on top of all of this stuff that's happening in the next two years, there's a there's a wall ahead where firms I think all the fans I've talked to are saying, yeah, we will have to decide what we cut, what we keep and what we cut. Yeah, but that's yeah, it. Well, my name, the new authoring platforms will be more mature at least. But then you have a then you have an uptake cost of like what we already know versus what we have to learn, right? And I mean, maybe that's where AI plays a part as well. And these agents, right, is just getting people up to just a very clever what people are using now isn't going to be the future because file formats, RVT, all that stuff eventually goes away because it's going to be in the cloud. And the question is it going to be former or are they going to take revenue refactor it? I don't know what the score is with all of this getting more in terms of their plans for that. But you've got Snacks, you'd Connick, R call Motif, to test fit. Yeah. Well, yeah, they can they can still more the front head. I think the factored R call Connick, they're going for a BIM 2.0. They're going to be they're going to most of them were trying to replace the revet stack from concept to drawings. And I think over time and maybe the drawings thing isn't so important anymore. Maybe that's because it's going to go for automations. So maybe that's a automated service in the cloud from your model in the future. So they're not I'm not seeing a lot of effort put into that. I'm seeing a lot of effort now being put into detailed design. But on top of that, there's going to be an agentic layer which is going to be sat inside these platforms running all the time on your data as you enter it. And it'll be a different it'll be a different bowl game to just creating a family of family parts and then creates and geometry. So I want to go back to something you mentioned earlier and you you made the comment that Autodesk doesn't know kind of the nuance of hospital design as an example. But don't you think there's a threat of them learning what those new nuances are by a tool like what you're talking about right now an agent running on top of this stuff all the time or and maybe and not or in the cloud right where all of these models are coming together and it's watching changes and how you know like it knows your docs as well like so it's got your meeting minutes in there and it knows the reasoning behind and what's tied to what and I could see I'm just curious what you think about that kind of potential threat there. And I don't know if they want to compete with architects. I don't know but but it seems like they could learn a lot really quickly with all that data. Given given all of this stance on non-compliance I think that there'd be a bit of miss if they didn't comply with their own terms and use which would they have to have explicit permission from customers to train on the data and it's very hard to prove. No the only thing that this is me off is that Autodesk keeps saying we've got all the data we've got all the data. They do they say that and I believe when they announced Autodesk AI they even talked about it being like opt out like it was automatically opt in I could be wrong about that but but it's like one of those things where it's like oh it just seems like a gray area and so they're going to capitalize on it as long as they can. It is a gray area and I but the thing is that a lot of the a lot of the data that they've got is is not great because those models aren't super accurate you really need to have a read a really really really good rather user to make sure that you've got some great data to train off if you pop up and crap in then you can end up with a a user's one so. So, hey, hey they've got all the data. There's an argument for keep making crap models architect. Yeah just throw a whole lot of rubbish in just get it over the finish line people. I think that Autodesk has to comply with its customers. I think that they do desperately want to train on that data. They're turning the markets that they've got all the data because that's part of the AI play. They're also saying that most of their sales are based on the fact of all the great AI they've delivered which they haven't. So, I kind of think that the Wall Street gets one version of it but we have to look at the reality and they haven't really developed a lot of AI yet. There's no huge changes for people at the moment. It's coming but they definitely not not delivered it. This is another way to do it. Yeah, oh my god yeah Siri, I go from claw to Siri and all I can do is set a timer for cooking on Siri. I see the only thing I use it. Send my alarm. Send my alarm. I just realized I set Siri, send an alarm to go to. So, yeah I think there's a whole lot of stuff that needs to go worked out. I think customers are better off training on their own data and cleaning up their data. That's the important thing is not only bringing it in house but also cleaning it and realizing that their data has great value and it will have great value for the future. Also, start running agents over them to find the problems that were in past projects. That's really useful information to know. Historic failures as opposed to to help hide in the man in the future. But I think there is a big problem in the whole data world. I don't think many practices fully understand what they can extract from the data. It's too easy to rely on a large vendor to do these sort of things when in fact they will capture all of the information. It is possible that in the future, we have 20 years from now, what a desk has fully understood every single aspect of the industry and they actually say I want to build this is the land parcel, give me a building, give me the spec and they'll design it. Why not? Because they have everything. That's my question. Do they actually want to compete with architects? I kind of don't think so. I think about the AEC market of Autodesk which is a pretty isn't it a pretty small world within the grander? It's the number one. Their MCAT is so how the seven billion. It's really hard to work out because there's AutocadlT and they kind of weirdly allocate those to different divisions but I think over 50% is AEC and then you've got MCAT. How many is kind of dwindling? 300, 400 million MCAT is over a billion. I would imagine. I scratched my head on what the future of them is and they were yeah sell it. I don't know what's going to happen there. This is the thing with the whole panning thing is you're on a desk, you're bending your name a check. Do you do a deal with a panning to you or an open AI or a Google to superpower your products, your business offerings because that seems to be like a quick way to get in but if you don't then panning to you're going to come in anyway they'll once they're kind of king of the castle because they come above you then suddenly everyone co-tows to that system but then there's this whole ownership thing, the whole proprietaryness of and movement. road structure, which we might be swapping, we might be demolishing all of these lower down modes, but then suddenly panning to turn up and they own the entire process, the entire business model of a company, they become such an important substrate for running that business that they can't be removed and anything else that's underneath, what's stopping them panning and all today is going to be a Bentley and having a vertical, direct vertical channel. In the south it will, there's huge questions about even down to the kernel level, because knowledge and when you create an object in Reb, it doesn't really know what it is, it's just geometry with some metadata. What happens if the wall when you created the wall had some AI to the point that it knew what it was and it knew the limitations of what could be built and physics and stuff. So maybe there's an argument that we need to go right down to the bottom level and say you know what, we have to go to the kernel level and start having AI right from the get-go, so everything that gets made has some level of knowledge, has some intelligence and then you have a totally different outcome from the bin systems that we've used, even Bin-T point out, doesn't really go to sub-level. There's a great guy called Blake Cauter who I knew him once first time at PTC, he kind of helped write the granite 3D modeling kernel and I've had lots of chats with him and he thinks that even geometric kernels need to be rewritten in the day of AI, so that the kernels understand what it is that they're making so they can help, as opposed to just be done geometry. So that's quite a deep thing for a lot of people because they've pretty known some of kernel's work and kernels in AEC tend to be very done because it's like faceted geometry. They're very rarely solid models, sort of benty-do's, parasolids and vector works to parasolids. What did the Revit has base maker, I think, Neil? So they got their own thing, which was AECs, which isn't really used for much inside of Revit, I think it's STI out, but maybe someone will come in at that lower level and start thinking about how we can make software tools that are intelligent from the atom up as opposed to us formulating what a door is, what a window is. Yeah, it's interesting to think about that though from a design process standpoint because a lot of times design starts ambiguously, it's like this represents space and there's no space object in these tools that you're talking about. So it's containers of space and walls and ceilings and softlets and floors and like a wall could also be a something else and it's really ambiguous. And then at some point, I mean that's why people love Rhino, right? It's because you're dealing with form, like you're not dealing with those level of specifics, but at some point you have to transition to that. But this is whole leap from conceptual massing to level free 50. Let's say you start off with some blocks and you convert those blocks into walls. Okay, now it has an inherent knowledge of physics, it has an inherent knowledge of construction of walls. It will then realize which walls are supporting loads. So it'll then automatically detail the right, the right place structures to support that wall, that the windows and the doors and stuff in that feature. And as you leap forward by pressing a button, all that's all that detail appears. I mean, you're going to have, it's not impossible to have construction level information from the press of the button because it will have inherent knowledge of structures. And I think kind of schema, do it, but it's not really AI, it's a, they sit down with the architect and they say, okay, what kind of things do you build? Okay, they'll capture how they, how they design their kind of like multi-occurrency housing projects. And so every time that you've done a massing model, you press a button, it'll leap forward and do all the detailed design for you. Yeah, it takes like those, that catalog of spatial components and plugs them in like the geolocubes, right? And that, that, again, that's kind of almost like a expert system because they have to go in and program their expert system with how their customer is working. And you're going to see a lot of that. You're going to see a lot of people trying to make all the software companies I speak to are trying to compress parts of the, the, the monolithic BIM. They're trying to speed up drawing, they're trying to speed up massing to detail, they're trying to speed up a detailed modeling to fabrication. And they'll come in with all these different ideas, I don't know if anything, I'm going to stick, but they all have the potential to save huge amounts of time. And, you know, I think architects need to maybe be a little bit more about off-site construction or just construction. Actually, you know, the actual physical and about how that is represented within their model. And maybe they need to start thinking a little bit about workflow better, where I'm saying, well, if I see the great thing about, architects, architects weren't super great at high quality drawings in CAD. They would, they always have like lines that didn't meet and that he was, it was, quality wasn't necessarily there. You jump up to 3D and if you haven't got good quality in 2D, then that can get even worse. And then, if you try and get auto drawings out of a not-great model or a model that you'd just say, I'll fix it in the drawings because I didn't know I'd stop rev it from doing that automatic connection and I don't read out the way that those walls join. That's a problem because auto drawings really relies on your model being great. If your, if your model's great, then your auto drawings are going to be great. And when I, when I wrote about auto drawings, Michael from high-up contacted me and said, we did great motor drawings. And I was like, well, yeah, you, you designed the model. There's no modeling involved in high-arch. It really is. Yeah, right. Incredibly automated. You didn't get the model built by an intern, right? Yeah, there's, it's, and it's level, it's one to one. The whole thing's model that one to one. So, yeah, you get great auto drawings out, but everyone else, they're going to have to have trouble. But I think now with with agentec layers, maybe constantly patrolling your model, making sure that it's correct, then you're going to end up with really good models. And yeah, I was going to say that. Then the next thing is to get that helps auto drawings. So a lot of these things are kind of stacked benefits. And I think that all the stuff we're seeing is some compelling. It's so exciting. And that's one that, well, yeah, to start where we started off is I see, but I see people trying to perfect their rivets by, you know, just I am just like, this is, this is really, I know it's current technology. I know it's going to be probably for the next five years. It's going to be incredibly popular, the most popular way of probably getting binned. But the leading firms are not stopping there. They think that that the rivets are drawing tool. It's a thing to get drawings out of. And if automation comes into it, then that's gone. And they're they're more keen on brino, they're more keen on some of these AI tools that are coming out. The rendering tools are all over. And they're the ones that are looking at hosting their own project data in house, doing their own AI, trading their own AI's. I think a lot of them are building their own workflows. And so they'll build a workflow where they're routing stuff to go into AI applications are in the cloud as they're modeling. So they're modeling, they might be modeling in rhino, but automatically there is there are renderers ready to go because the modeling is being fed directly into these things. So productivity wise, it has huge potential, but you have to watch your tokens because that's going to be the new bleed point is. And mitigates is and fetted access to AI tools, which require a payment for a large team is going to seriously start adding up. I was talking to Greg H. Okay, and he said, you know, they're spending like $10,000 a month on tokens. And he could quite easily see that doubling or tripling based on the. Well, that gets more useful. The more you use it, it actually gets more useful. And more, I hate to say addictive, but you know what I mean, right? It's like, well, yeah, I'm going to try, I'm going to see. And there's a lot of experimentation that needs to be done to figure out how to work and the best way to use it. And you're going to get stuff that didn't work right. And you're going going to have to try again and again and again. And the real skill is having somebody who is pay eye cognizant to say that, okay, when you're doing this the first time around, your first iteration of this process, you're going to bleed tokens. But once you refine it, you're going to, the most expensive thing you can have is like 4.6 and that's, that's really expensive to run. But once you've worked that out. Well, you're going to come up with an instruction set and you swap it over to Sonic 4.6, right? You start, yeah. You start for farming it out. So that's, that's, so that's the new thing is kind of like spending a lot up front and then over time working our ways to minimize your cost. And some of it, yeah, reduces it almost not to zero, but really low. I'm working our watch out LMS best to do each task because they're changing all the time. Which is another task just managing that, yeah, for somebody. I've, I've figuring that out. So I, yeah, I've spent 30,000 tokens this year just learning and I've had some real fun and real pain. I mean, I've been up, I've spent five hours nurturing this, trying to create this one thing. And if, yeah, my account can sound to me saying, I need all your invoices. I go, oh, God, there's a pain. I just got to keep going to all these different sites every month. I was like, so I'm going to write an AI agent. So I sat down with an AI agent and we went, okay, right, we're going to go to this site, this site, this site, this site. I don't really want to give you my long-in details. So I'm going to, and all of them were two part of the organizations. So I had a sit there with my phone when it wanted to get in there. So I was like, okay, here we go. And out of the 12 sites I wanted to go to, I could only get into two. Ten of them immediately recognized it was a bomb. And said, oh, your IP address is here from Germany, well, and it's a well-known bot IP, goodbye. Then your cursor's moving too quickly in my browser. You are a bot, goodbye. And I was just running, I was running into this every time. And you kind of realize, well, all this great thing and all this stuff, these current agents are doing, they are bots. And we have spent 20 years developing internet protocols to farm these, to figure out, you know, to pack factor authorization, whatever, to try and stop this happening. And I was going to get to the brain. It's going to be the main way you actually do things. And now it's, if you haven't got a good API. And so this, this, this agent said, I was trying to get it to scrape Google Analytics. That shouldn't be too hard. The summary is the API key that this thing kept getting would only allow it to look at YouTube videos. It wouldn't allow it to go into our account. And then several times it was like delete, refresh the screen, get rid of all of the detritus from history, or okay, if I did that. And then it got, oh, hang on. The problems of my end, I can't flush my own cash and it's stuck in there and I can't do anything. However, I cannot write anything cryptographic in my sample. Have you got Python? And I've got Python on my computer. Right. Okay. Let's write some code to get around my, my sample. And I was at this point. And this point, yeah, it's the goal oriented thing. It's like, it will figure out work around. It will not give up. I had to go make coffee because I was like sat there with it then and we created a script. Then it said, okay, create a Google Cloud account, put the script in the Google Cloud account. I can then poke it to then generate a key and that will get me into after five hours, it could still completely failed. But it was relentless. It was, it tried every which way it could turn hack into Google Analytics. I told it to which is kind of worrying because he, this is out there. This is what everyone's got. This thing, well, it was, it was writing code going to some other site, putting it in a debugger. I was watching it debug its own Python scripts. And this is a, it's a very, if it couldn't do it for APIs, it's actually showing you the interface of the site it wants to go into. If it can't get in there, it asks you to put in the password or the two-factor author and it goes in and it just does whatever it wants to do. And I'm blown away by the in trepidness of it. It just didn't give up. It was hardly fallible. But at the same point, these things are now out there and people are going to be using them. And it kind of gave me a bit of some security feelings that perhaps we're automating something which security-wise were also not really prepared for. All it takes is one person in your company to, to what's, be scripting agents and give it security access, password protection to emails or something because it thinks it's, because a couple of the person doing that thinks it's a great idea to automate the solution and you've got an immediate security problem. Right. And the humans as usual are kind of the fallible part of this. And you're going to pay for it, right? With token usage. Yeah, I emailed the company in the morning and said, I've been up, I've been up all night with your computer trying to hack into Google, which I thought you might find interesting. And we also spent 4,000 tokens in the process that gave me the cat tokens back, apologised and said, try again. How's it going? Let me say, okay, nothing about this conversation. They didn't quite understand. But I, I, I, I'm in this really weird place of being in all of it all, but at the same point, every now and again, something comes up which makes me flag a problem worries. And I want every design IT director, every big manager needs to read up on this. They need to, I wouldn't say use open claw, try and find something else, but just learn about creating agents and a genetic feature because it's coming. And there will be a C agents to do exactly the same. Someone contacted me today saying that they were working on something like that. So it's got a chat interface, it will write programs, it'll go in and scrape information from AEC, destinations. Yeah, use with caution. But, yeah, your legal department needs to know about what, what's involved in this. And, and I just shut it down, right? Like, I think that's kind of an important distinction is, yeah, they need to know. And it needs to be a dialogue about because to your point, like this is coming. And it's not like lock it down. We're never doing that. Like I think about this kind of internet thing, right? And I kind of see it splitting into two. I mean, this is not an original idea, but it's like, there's going to be like the internet for computers and the internet for people. And those are going to be kind of two different things. And I think that you want to apply that thinking to these workflows and these processes and the access to data and the learning. The way that these agents can talk to each other behind the scenes and argue with each other and do all of this kind of thing while you're designing a project. We haven't been discussed that. I mean, we're talking about individual agents of now, but there has to be rules and governance for multiple agents. So in the future, when you have automation, you're going to have, when you've got an agentic system. So these things are running constantly on your data. You'll have an MEP expert agent. You'll have a electrical expert. You'll have a AC architectural expert and all of them will need to work together to solve problems because everything impacts anything. It's got comes down to that ontology argument about understanding not just the geometry in the, it's actually the process and how different things impact other parts of the design. And no one's come up with that. Yeah. I mean, right. What we're seeing at the moment is very early agentic scrapers and going to get some information from here and some information from there and then we'll try and mash it together and offer you some kind of insight. That's very hit and miss at the moment with Claude and with the tools I'm using in terms of what you get out at the end of the day. You can get some very deep reports and it's all based on assumptions. So you have to go to question as well. You know, it's it's overwhelming. Some of the reports you can get back are thinking, wow, that's so in depth and so. But yeah, there were some assumptions made on there. So it's all to do with how good your data is underneath and how clean it is that if you've got some information there that's not quite correct or not broken up. Very well, then it makes assumptions based on what it sees and you'll you'll have errors based on that. It's yeah, I would do just now right like people are doing this now with like hiring they can create like a virtual high. team of different roles within a company who look at a candidate or candidates and then argue about who they should hire and they can do this fully autonomously. So where you got people using AI to write CVs to be emailed in to be read by AI is to read the CV. Yes. And eventually the other end it's filtered out the people it doesn't think. So you don't even stand a chance. If you haven't matched that in a criteria then you're just not going to get a response. So this is another thing I feel incredibly sorry for young people trying to get jobs is that they are literally feeding into an AI and they're not given a chance by a human who might have more intuition and more emotional understanding of what this person can bring their company but that filtered out already by an AI system that isn't you it doesn't understand how many how many good people you're going to miss out on because they haven't hit your criteria. I don't know in the future if people are going to do degrees. So if you're going to recruit a human being they might not have done a degree which means they might not get in because automatically the AI has gone nope that person might have more business knowledge and experience or whatever it is that you need but because they haven't matched the identity. They won't get in. Right. And I'm yeah the more the more I'm I want more people in in human resources than anything because the enathemer of having AI do human resources is just so ironic it's stupid. I mean it's the one in tight sweat humans shouldn't be replaced. Well this kind of goes back to your your idea about like intentionally recruiting the next generations for the future of your company. It's like traditional HR departments have been just playing catcher right like they're just sitting back and waiting for inbound to come to them and and what you're talking about is a flip of that and go and actually like doubling down on the human side of it and going out and seeing and talking to people and what are they doing and why are they doing it and and that it takes a different kind of effort right. Look look while banks and a canton stood over here when I was going to college the banks and the accountants would turn up the open days in the university try to recruit people and there'd be human beings talking to human beings try to get people to come to their think because they needed the manpower. Didn't really see many architecture films rocking up a oxipolli to try and recruit people. I think I've heard stories of Wenzah has have needed they've got a huge project in they've got a huge deadline they're going to the AA and just like say who well who's got a laptop now and what's the job and they'll just send them out to Zahar's where we're in corridors doing bits of work for people you know that I love that idea but I think I think that there has to be outreach if you first you have to realize that you're going to you're going to you need to recruit you still need to recruit even though it's a cost and AI everyone thinks is a way to save cost increase productivity save time please recruit but go out and find the best people you can and go and grab them from university now because how many people are going to be put off from studying in the future I don't know but the people who are in uni now are going to be like gold dust because if they know grasshopper if they understand the concepts of BIM that they understand AI those people now are going to be very very very beneficial to you as you as you make a transition towards a tech stack that we don't know what it's going to be a workflow that we don't know what it's going to be a bidding system that we don't know what it's going to be but having the smartest people in your company normally is a good sign for success as opposed to not having or not having that many people in rely on AI and in some ways AI is really aimed at automation yes there's some stuff in here for looking at your legal side of things which make me quite useful for your contracts looking at your bits analyzing stuff like that yeah I thought I'd train I tried to train it was chat GPT on things I've written so I literally loaded 300 articles into this thing and then I said okay write me write me something on BIMTy.0 and it was so negative it was I kind of read through it went oh my god this is that really well what it's kind of come up with is that everything had a caveat everything was slightly crap and I was just like okay this is this is like I'm learning something about what things I write about like it was it made up stuff it hallucinated it was yeah this is not reliable this is not a reliable thing so my use of it now is I write my article but I'll I write I tend to I do such big topics I can't write them all in one go so I'll write about one part of it and then I'll write another part a bit of different day and I've got a different tone so the first day I'll be using I in the next day I'll be using we and it's the same as the BIMMS so then I'll put I'll put one part into I know you's Claude and now we'll we have a discussion about what it is like I said and we have a debate and it kind of knows it might history or more I've written and so then I might change my tone but the one thing I use it for the most is to blend these one bits with another bit that has a different tone because I have written on a different day and every day I wake up on a different kind of person or something something's taken me off at a different angle multiple personalities yeah well it's it's I've had a bit of information so was just contacting me therefore yeah I've slightly changed the nuance of what it is I'm trying to say but it yeah it's all right to write in PR it can't really do a proper article it's very good at taking an article and dismantling it and giving me five what is it I'm saying it look at me the five points and I'm going okay this is this isn't that important so I was my business partner Greg always is horrified when I send him an article because we kind of we kind of agree what's going to go in I kind of give him a vague word count and then I get carried away because something is major so a 2000 would article some hands are being six and he's like saying yeah right God needs to try and fit this in the magazine somehow I know I know Greg I can imagine this this reaction yeah yeah I just don't I'll give it to him I don't if I don't hear from back from within a day I know it's it's not necessarily a positive reaction so but yeah that that's why there's already so the article I sent you the the agentic feature of BIM that's an issue that is now available on issue I think it's going to be on the website soon I've already got the follow-up is already written about governor I was going to ask you about I was going to ask you about that because you you you asked some questions in there and it's like well it's just amazing how fast things are changing because you're answering your questions you know in the next article the next I rate it gave it to Greg and then the Monday a Google paper dropped about agents working together and how first it was how do you how do you get an agentic system to be precise and and part of the solution is to break a problem down into infotestinantly small parts and break down the processing into that so that when it goes back up that you end up with much more likely specific answer as opposed to hallucinating and then the other thing was that these agents need to work together so you need to have a governance structure so that they all have the same playing field so that they're not they're not acting independently so that's like that that was one and then three days later I think on X this guy from the finance industry finance and legal wrote this huge home about which modes in business are at risk and it was obviously from a kind of like a finance and legal point of view and then I was trying to work out taking those they were very generic abstract issues that you were talking about not necessarily about within the industry but one of them was kind of like payment systems payment systems are very hard to replace with AI so things like pro court were actually quite strong well things like ACC were a little bit there's some elements in there that are kind of easy to replace and so I've got this matrix now whenever I take a product I slap it into this and AI will go through it and say this can be replaced by this plus at that open SDK the very generic this This can be replaced by this and this SDK. This can't be replaced. So these big systems have been developed in our industry over time. You can see which ones are going to be kind of like the massive game of Yenga. Some of these applications will be ubiquitous and you can create loads for free. Don't have to buy the whole thing. So then what happens? Some applications, some 25 application collections, 70% of those applications were replaceable with SDKs. Wow. So there are the stuff sounding. One of the things you mentioned earlier was this idea of like we were talking about Rhino and like the conceptual to the more level of detail. And then there's kind of the other way now which is you can ideate at a high level of detail through the image making process, right? With nano banana, nano banana 2, nano banana pro, whatever. So that just kind of flips the process around potentially, right? You don't have to start with lightweight geometry. You can start with literally lines on a page and it can imagine just to kind of take the mid-Jerny parlance. And I wanted to just talk about kind of how this goes spatial because Autodesk recently announced their acquisition of, is it world labs, right? Where it's like this. They invested 2 million and the total raise for the company was a billion. So Autodesk didn't open. It wasn't an acquired, correct, right? It's a, they'll open the Camono or something at some point so you can see some of the technology. Okay. I, I, yeah, it's, so the one thing that happened in February I think was that Google DeepMind showed a 2D sketch from, it was a 2D sketch of a complex shape. The complex shape had gyroids. It's a very mathematical shape which in engineering and 3D printing is kind of an interesting structure because it's kind of self-supporting and it's got a lot of things that could be used in 3D printed solutions, air cooling and stuff. So DeepMind took a 2D drawing from the 2D drawing it created a 3D model and it was kind of like a laptop stand made up of all these holes which were, which had gyro geometry. So it made that leap from a 2D sort of asymmetric drawing to a 3D model, a meshed model. Then they said, okay, make this into a laptop stand. It must support 3kg and it needs to be able to cool XYZ, has to be free standing, whatever. So it gave you some physical, physical constraints of a product. It then somehow, somehow made sense of the mesh. It then started playing with the size of the gyroids to let the air, air flow because the laptop was going on it. Then it had to create a structure that was strong enough to see more 3kg, then it produced an STL file and then it 3D printed it. That's right. I can get powers right there. That was perfect. Yeah. That's like, whoa. And we don't really know what STK is it was using or whatever, but just as a demonstration it was like a bomb landing in our little part of the world. It also goes back to Blake Cauter with his idea of having a kernel that actually works well with AI and ML and we were talking about that because he was blown away by it. That STK is available and he's already all over it. Because this is, we're living, everyone's living in fear in our industry, the anthropic will just drop another one in our space one morning. Oh, we wiped out this job. Maybe it's automatic drawings, Google mines, or get you to the article about Cobal in IBM stock, Canked. Every day these little bombs are going off and they're only going to get better. AI is the worst it'll ever be today. It's always going to get better. And so within that realm, the possibilities of creating a sketch and having a model is coming a reality. We've been able to do it for a while with certain things, not particularly brilliantly, but if you could get from a 2D drawing, you could kind of make a crappy mesh. The crappy mesh you could try and make it into nerves, clean up a bit, very manual, but those little leaps were there. Now deep minds just went, "Oh, I'll go from a drawing and I'll deliver you an STL. It'll be a functional object for that intended usage." So yes, I think I'm pretty sure. Large significant architects are developing those kind of tools in-house for their teams. So they can ideate from 2D, 3D, go back to 2D, AI render, 3D model, 3D print. That's going to happen. The question, I don't think it kind of removes the need for Rhino and there is something to be said. You always have a translation from a 2D drawing to a 3D model. It's always going to be slightly inaccurate. There is still a strength in being able to model well and the foster empanners and for Zahs, the form is everything. Is that a roof or is that a wall? I don't know. It's glass. It just flows. So I still think that Rhino and those skillsets are going to be needed. I think the sketch can be taken further and maybe it's enough to win the competition. The idea of entering competitions with this kind of technology behind you, it just crushes the cost that you'd spend on developing all these different models and all these different ideas. Ideation is going to be very fluid. You'll be able to deliver a lot of different ideas very quickly and you'll be able to sketch and get your 3D renders and then your sketch and you'll move instead of doing another sketch. Maybe you flip the model around and go, "Oh, okay. I don't need to make the back of the buildings symmetrical." Or maybe the back of the building isn't symmetrical and there's made some stuff up. You've got no idea, but this is happening. So you'll be able to scan a building, a existing building. So one of the guys you've got coming to next build, Floral Poo PO UX. He's a professor of Point Clouds and he'll take a Point Cloud and go with AI straight away. He'll recognize what's in the scene. The intelligence that you can apply to dumbness is just going to make everything incredibly fluid as to levels of detail. This is probably what it boils down to. We're so used to building up this level of detail from kind of. to very precise. You can start at any point. You can start with a scan. You can start with a hand sketch. You can start with a massing model. That's all free. Those are all versions of sketching. I mean, coding is sketching. Nothing will stay dumb. Nothing stays dumb. I think that's. And this technology filters all the way down to this, right? And you've got it in your pocket or it's in your glasses. It's in your newest wearable device or all of those and they're all talking to each other. And then do you go. both drones, drones are kind of like a common thing now. And to grow my trees really good. Gaiasean's spawning technology is amazing. Just a few photographs you can kind of go ahead and a really nice model now. It's. I was talking to Gensro and they were talking about the. before they even get started. If they've won an airport job, they're not doing the old airport. Maybe they're doing a refurb on an airport. It's really important for them to get the survey data in so they can start getting designing. And normally they'd have to go and commission a. surveyors, surveyors would go out. And then it could be weeks or month if there's a big site before you get the data stopped. Security clear answers, all kinds of things. Now you can sense some office duties that have some mobile phones and they can capture the inside. And you can get going. You might want to. You get going. You can have the surveyor do the proper job, but you can get going based off some phones and or just a video. This even just go around making the video and the video is enough to capture the current state. Gatsbyn's brain is just coming along amazingly well. So yeah, we'll be doing a lot of on reality modeling. This is yeah, there's so many workflows that could exist and I think a lot of phones have a workflow and that's our workflow and I think that that is inhibiting of taking on some of these new technologies and having a team look at other technology and think, you know, well, maybe we could improve. It doesn't have to start here or we can start over here. It's all about pipelines and getting data and having the data in the right format at the right time to do the thing you want to do. And I think architects especially need to spend some time thinking about their processes to adopt some of these technologies to really rapidly advance. I guess the biggest fear is risk is getting the wrong kind of data in an era that can be a big problem. So I guess risk is not the architect's best friend but it also stops them from exploring some of these avenues. And you have to hope that they've got employees who are slightly the architect that goes off to learn by conscripting because they're interested. And so a lot of firms will have people who are going off looking at this technology because they're interested. They're not like an HOK where they've got dedicated people or the foster partners. Well so much of business has been established on locking down the way we do things because that's repeatable, right? I mean, there's efficiency built into that. And you're talking about everything is a moving part now and you don't have to pay attention to all those. You absolutely don't have to. But I mean to your points like you're going to be competing against firms that are paying attention to all those things. Yeah. And had the very high ends, they were very aware of that. And that is one of the reasons of striving them to look in the medium to lower ends probably the impetus isn't quite there yet because everybody is quite stuck in their ways. And everyone keeps telling me how AEC don't spend less of agriculture as a fishing which really gets boring and is crap. Not the firms I see on lakodes. And I kind of think they must take the entire construction market and then they just like tell architects you're really slow. Well, yeah, if you're sharing some guys out with a spade and a pickaxe, then yeah, OK, if you take the whole technology funds, then maybe they're not spending that much. But architects today are spending a lot of money on their reverts, spending a lot of money on real-time rendering, spending a hell of a lot of money on workstations. These people are trying. I think it's a laziness in the software industry to try and label the industry as lakodes to in some way get them to buy more. And it really annoys me. I think there's a lazy thing to say. But changing workflows is difficult. It undoubtedly requires risk. But if other people are doing it, you know, from just this year, just a January to now, the stuff that's been released is mind-blowing. I've got software companies that are worried about having their motor destroyed by AI next year. I've got construction managers who think that software, AI software layers are going to come in to remove their entire roles. This is panic. There's a lot of panic in the people I speak to. And yet I can see architecture firms with foster parties of the week and talking to Greg H. OK, then I'm absolutely happy. I want to heal these things out, try to work out a new tech stack to make use of these benefits to streamline their processes, to use everything that's out there. From real-time AI rendering to automatic drawings. You know, that's a big shock last year. And that's what I started producing. Like you understood what a drawing was. Now how accurate that was is another kind of fear of mine. Fitchfedeer doing a learn of testing with producing drawings with Nana Banana as a kind of like a DWG output. I'd love to see what the accuracy is that versus a real drawing. I mean, I always had that thing when they got the video AI generated video of that actor eating spaghetti. And then now it looks like it's a, it's a, it's a, it's a strict one movie. All the four is like a disaster area. All of these things are just improving and the chances of having an, I think I send you the thing where was somebody was just designing a, a internal building layout. And it started, it started with a floor plan, right? That was it. And even the floor plan was, was generated, wasn't it, by AI? So, these are all canned. So there will be some messing around to get that. But if that's doing it now in two years time, you can imagine what that's going to be. The wrong thing to do is to look at those things and look for the mistakes. Like that, that, of course, there's mistakes. Like, there was a, a, a, like an editor movie producer who posted a, a six minute movie, a short film on LinkedIn. And it was entirely AI generated. And there was lots of mistakes in it. But you seriously just have to step back and be like, can you see what's coming? Because that, and that's why it's important to not just sit back and watch what, how this is unfolding, but to actually get in there and use it. And I, I hate to say like you've got to use this stuff. But if you don't get in there and use this stuff, you will not be able to use the next version that comes out as effectively as you would if you're in there doing it now, right? If you wait. So I think it's, and just to elaborate on, on your example, so, so it started with a, I think it was an AI generated floor plan. And then it generated like top down perspective floor plan views, fully rendered. And then it was renderings of every room in this apartment. And then it was movies taking you through it. But it was all of that. And it was like, it was kind of build as like interior design or heaven, right? Like that, this is who they're aiming it at. And it's absolutely incredible. The kind of work that's coming in and what does it take? It takes, you've got to get in there and play with it to actually be able to do that. And the way you think is a lot of it is down to being able to explain it to a chat box, which isn't very intuitive for a lot of people to try and explain. I mean, I've tried generating images just straight forward fly images and boy, boy, that's hilarious. You know, just just trying to get the image that you have in your mind, but trying to put it into words and what you get into your point. Like that is a muscle that you build over time. It's not something you just go in there and you start getting good stuff immediately. You just, I mean, it's just like learning a tool like right now. It's just like you're not going to build a great model the first time you do it. How do you learn how to build great models? You build a lot of models, right? Like that's how you do it. Or you go online and say help. And then there's loads and loads of, you know, thank God for Reddit and all those kind of things. But there's a great self-help community. In fact, so I was having trouble writing an agent and I asked the agent that I was having trouble writing an agent and it went, oh, okay, let me watch. And then it was deep hugged itself. It's all I was doing wrong or the clarity, the lack of clarity that I was giving it. And then I was like, okay, so now agents are kind of extracted from itself from the agent to look at everything. It could look at my repository of other agents and say, oh, I can see I've already done this bit before. I will instruct the agent to do it this way. Oh my God, I'm going on. You've already done that. I found it over here. Do you just want to modify it a little bit and update it so that it's the latest version of what you're going to hear? Shall I run? And then go run. And it just, now after a month, I've deleted 50% at least of all the agents I created. because they weren't giving me what I needed, all the value. All right, yeah, I've done dumb things. But I've kept half of them, and they're the ones that don't have security issues. They are, you know, I've got one of my nicest ones. It just goes out, and it looks at all the major BIM vendors forums. And every Monday morning, I log in, and I can see what the top five talk to about things are on the Revit forum, the Hockeycat forum. And I've never had the time to do that. Someone was taking me home, so it's all blowing up on the Revit forum. I'm not going to have a look. Yeah, the pulse check every morning. It's kind of like, okay, there's nothing major's gone off. If it release comes up, the causes some crap, I've been getting this way away, and see what it is. So it's trying to be omnipresent. It's a godlike feeling of being able to scrape in what's happening that month, that week. One of the worst things I did was I created this app to go and monitor the share prices of all the major BIM vendors, of which there aren't really that many. And then I said, okay, I'm going to give you a fictitious $1,000. You can go off and invest it, and we'll see where you are at the end of the month. And then anthropic launched its bomb. And because I don't know you're giving it $1,000, it can only afford four auto-deshires. Or it bought two auto-deshires and it bought something else. And by the end of the month, everything was underwater. The whole thing was just like, yeah, this isn't quite where our photo would be. You need more money, more money to give your AI. I should have given it more diversity, what are you giving this to do? I think it's a good thing to sit with a blank sheet and say, what is it that I've always wanted to know? But I haven't been able to, it's just been too complicated. It seems too broad that I can create an agent to go and try and get me this information. And there's practically no cost in coming up with it and scripting it. The more complex it is, the more likely it is to go wrong. That's the first national I could tell any money. One of the more interesting things that I find about it is that you can get to a point where it'll make recommendations about how to do the next steps. And then you can literally just say, do that. You thought it was telling you how to do it and then you just flip it around on it and say, yeah, just do that for me. And it's like, okay, done. Are you sure you want this or do you want them both? They'll always give me like, I'll go ahead and do both. And then, so yeah, I, and there was something to be said. I think someone said that AI doesn't necessarily give you a productivity benefit. It actually just gives you more work because I could do that. It's like BIMs. Yeah. Yeah, we get absolutely enamored with the fact that we can get all this data. And I'm just, I'm reading more reports that it's generated on things that I'm interested in that I've ever read in the past. And that's made me think more, which has made me want to write more, which has made my Greg's life hell. And I really don't fit in with our schedule of right, which has been kind of regular. And now I'm sorry, I'm sorry Greg. So I think there's a, there's this, it definitely gives you more, but I, yeah, we're human. The more information you give us, and it's not, it's always just not conflicting, but it kind of builds your, your knowledge base on something. I think, I think it's absolutely possible to not know anything about, let's say, 3D printed concrete. And within two days know everything that back to really pretty good. Okay, that's going on because you can, it can go off and find it. It can prep it. It can press it. You can give you information. You can ask your questions. It's like a private tutor and research agent. And what wouldn't in the old days of comp you serve and telephones? You'd never get access to that information. And now you can even find out who the leading lights are, who are the people who are most popular and respected the most. And then you can go and track that person down on LinkedIn and then have a real conversation with somebody. So I think there's a, it's, it's for all the bad things. It's got huge upsides. So it's very conflicting for me to kind of sit there, worry about the next generation, worry about lots of jobs. And then I think people who, in your company, who run with this thing and will build up a level of knowledge and competence, there will be so valuable to firms. You need, definitely need to keep people who are there. You also have to have this legal oversight because it's so easy to stray into this world of giving away access to stuff security. I mean, everything you upload to these sites, I'm not, I think chat GPT kind of, I think it has a right to and stuff. I'm not too sure on the legal ins and outs, but I know that if you go, how do you bury, how do you hide a murder and bury a body? If you get arrested, they can go and find your chats and use it in evidence against you. So it's, it's a, it's an official record, which you have to be careful. Not that I murdered anybody, but you have me careful. So it was email, right, and how many people have just put stuff in email that they should have never put an email, right? People think it's a safe, secure, private system and it's not. It's the opposite of that. Yeah, I, I, so may Winfield from Bureau Apple, she's a regular at next bill because she's just all over this stuff. Very good. And that's the great things that she's like a super, super, super lawyer. I am not, but when I was looking at all the ULUS stuff and try to work out what was going on, I could actually have a conversation with May on her level, because I, I, I, I used AI to go through the legal documentation and I had questions from that that I could ask her about. And I might not have understood exactly what she's on about, but I could take that conversation and get the AI to translate it into more human terms. So, um, non lawyer terms here. Yeah. So I could, I could have, it's enabled me to have conversations with people that, um, that I, on a level that I couldn't do before. And then sometimes I get responses from people I don't really understand. So I've, so I'll put the chat or that portion of the chat in and a number of times it's kind of said, oh, what they're saying is this and, okay, this stuff starts me from misreading. One of the journalistic skills you get is interviewing somebody historically, I'm on a tape or on a, even as digital. And as there was no voice, voice to character recognition, I sent my headphones on and I would be literally typing out exactly what they said. I don't know, you ignore me. But within that process, I hear things in what they've said, like they were trying to tell me something here that they felt they couldn't say. So there was a really weird kind of like nuance that you would pick up on the reading between times. Yeah. Right. So at the time, you know, the hot fire of debate, you wouldn't necessarily pick up on it. And that's the, how do you think I think it's probably missing is the nuance of the, the human trying to be language. Yeah. But I think that will also be a I'd, I think it's only a matter of time before it picks up on this person's putting towards that point to this. If you put this and this and this together, then this sounds like what's happening. They've already got AI that they can tell if you're lying or not allegedly. So CEOs are really worried about reporting to Wall Street and seeing their face because the AI might be able to kind of identify which things are the big fibs. I don't know what the world's going to be like. You might as well be able to, if everyone's got glasses on and you got an AI and you're you're talking to somebody saying, this person isn't telling you the truth. My wife. There's been movies about there's been movies like this, like a liar liar. And yeah, these things have played out for sure. I think I think we look at it differently. Yeah. So, so let's make the case for the anti AI slash like a lie. bloodite AE firm. Do you think that there's going to be. I could see, you know, this is my identity as a firm and I align with people who are not going to go down that road. I'm just curious if you've thought about that. If you played that scenario out in your mind at all. I think there will be a boutique service where you're dealing with people, but you still want them to have AI to remove the risk and stuff. They will have to use technology, but you will. Sorry, look at banking. I used to have a bank manager a year ago, maybe 15 years ago, 20 years ago. Now, I have not a bank manager since. The people who have bank managers are very wealthy people. The one-to-one service is for you, where you know, this yes or three begs fall. Everyone else just deals with their webpage. And I think that human content, human interaction is going to be incredibly important for very expensive high-end services. And I think that rest of us will be dealing with machines. And the machines are going to get better and we probably will forget we're dealing with machines. But ultimately, the machines are going to be right most of the time. And humans are fallible. And I don't know if we'll be too happy. When they saw cars, they used to say they were hand-built by robots giving you the idea that they're very precise and that means they're really good engineering. That didn't mean they rust didn't rust or it was a bad design or whatever. But the idea of something being automated. I think, yeah, in architecture, I think there's always going to be the savant. There's going to be somebody who comes up with something that no one's come up with before. The problem is how that person gets educated and into the business goes off to do their own. I mean, you see practices like foster and partners, recruit all these top guys that come in and then eventually they leave and they start up their own practices. And that will be still the case. But I think it's going to be fewer because there will be fewer people in the system. You're not going to need a legion of people doing drawings. You're not going to need a legion of people doing renderings which you used to be the case. They're saying that right. Off-strand partners still was like 2000 plus people in it. It's huge. They're busy as hell. So there is plenty of work. Maybe it's either sees probably, it probably is. But it's. I just don't know how big the films are going to be. What's the ratio that AI is going to compress increased productivity and compressed team size? No idea. You're always going to have to have somebody. You're always going to have to have an engineer sign off on anything that gets produced by engineering software. I think the architect will always want to speak to a structural engineer and not just. Kapita says yes. Kapita says no. I think there's still. that has to be in there. I think the software industry is going to be a very different place. If you're a software developer, you're going to have to offer more than just a. well, it could be created with AI plus an SDK. You have to offer knowledge. You have to offer a higher level of value. We're coming to the end of the low-hanging fruit software developer. You have to add real value, whether in breadth or in depth. I think that people are going to be rethinking their product strategies. So you're going to have very different kinds of products coming out. I unfortunately think a lot of people are kind of thinking in this tone of creating an AI feature. What feature does AI? A couple of features that say AI. So it's in a standard bit of software, but they've got AI on this one bit, which will give you checkout. That's a feature. That's not really an agentex system. That's not really where this thing goes next. So I kind of feel the next two years, you're probably going to see a whole raft of people doing single features, which aren't working together. We'll take something, automate, spit out. But what we really want is something that all together works together. I don't know who's going to be developing that or where that's going to come from. You'd kind of assume it was going to be like the order desks and the nemmetrics and the bennies. It could be a completely new company. It could be a palantir that buys something. It's the level of knowledge that has to go into an NEP agent or a level of knowledge or mentors are spending a lot of money in a lot of time building those NEP agents. I don't think it's something that could be done again and again and again. And then this is the problem for the big software firms where if you're on a desk and you've got NEP, you've got Ravya NEP and you've got stuff there, but it's done. And then you've got an augmented which is solving overnight, but maybe augmented gets bought by a cowly or an e-com. Wow. Okay, there's no one who's spent five years and ten million dollars to do all of that. So what you're going to do is you're left with whoever's next or developing it in house. These things are going to be like rocking horse poo. And that's why you can see they got bought by a-com because they're off very well. Let's take a take it out of the equation. How's it up? Well, there's augmented. That's really the only other company AIA agent in this area that I know of and then Kendra kind of waved the flag. One of the guys thought that the Nordic desk told me about them. So I was kind of like had a conversation. Yeah, the rarity of the knowledge of the skill might be the thing that keeps the industry as it is for longer than I think. And maybe I've underestimated the lack of the time it takes to build an AI agent in a specific vertical. But you got DDS CAD, but Nemechek, they've got all the piles and the libraries to make it 3D, NEP, but they don't have the AI logic to do it automatically. What does this could do it automatically within house. But then that's specialist knowledge and there are going to be that many of those people to do it. Architecture for me is the difficult one because there's so many things that an architect has he said before as a degree goes. It's a brilliant baseline to cover whole manifests of different types of artistic, physical, structural, aesthetic, historic. A lot of my friends who went off to do architecture ended up doing set design or interior design or nothing to do with design whatsoever. It's a very good basis. So maybe it's something that we should be recommending that people do anyway because even if there aren't brilliant AC jobs in five years, three years time, then there might well be other areas where that skill set that you get can be used or put to use. I don't know. Yeah. Well, let's talk about your conference that's coming up in when it's made this year, right? 10 weeks. Yeah, that's so. Are you winning? You've got a lot to do. I've got a lot to do. So it'll be going for 10 years and this is our 10th year of doing it. I do not have a shortage of people who want to speak. The challenge is to get the right people to speak who are doing things that are out there. And if I get a BIM show nine times out of 10, it's so impitching their product. That's like, doesn't really tell you anything. Next bill for me is like next. It literally is what are we going to do in the next three to five years? I've always decided that old BIM is a dead thing walking. Not a topic. What's coming? And while I assume that everyone in the room is using dead thing walking technology files or less stuff, we have to start thinking about how this industry changes. AI is going to be huge. So I tend to AI and AI is going to be like a substrate for it. The BIM 2.0 is a now kind of, they've got product out. It's not the product that we won yet. They're all coming. So they're going to be there, obviously. I've mentioned the engineering side. I've got everybody who's doing AI engineering design all in one place. We're going to be having a discussion about the impact of that on the industry and where that's going. I've got off-site construction firms that are still willing to try, which is like this the biggest graveyard in the industry. It's just unbelievable that I'd hate to add up all the money that's gone into off-site construction that's been burnt. So obviously we're kind of having two things. The events over two days and normally have kind of like a communal session and then it breaks into two streams of which it's a bit random in bedrooms. This time I'm going to have to keep architecture in one room, construction in the other so that we can go quite deep into these very specific areas. And then death is still death where unfortunately a lot of people think it's about developers and it's not software developers but developers are like, no, no. It's the day when we sit around and we try and work out where this technology is going. And the important thing about death is that people from the design IT guys from AAC firms are there to tell the software developers and the money people where their problems are today and what still needs solving because we're fed up of seeing stuff that gets developed that we'll have 10 different CDE modelers or 10 different kind of like you've got test fit and then suddenly they'll see 50 of them. Well we don't need, well this money's been poured into developing the same thing again and again and again to do coordination. What can we tell you what we need? So you can spread the load and maybe solve some problems. So next build this year we're still continuing with the automation of drawings. There are now more people who are developing auto drawings applications. I think there's going to be some launches at the show so we're going to see some more auto drawing technology server based in the cloud doing all sorts of stuff. We're going to be looking at how do software firms and AAC firms get together because if these startups don't get some engagement and feedback they're going to die and they have to have some level of engagement to get these you funding. And so we're going to have a debate about how we can get these startups to work better. I mean AAC firms had a time poor, they have a lot of time to invest in stuff so how can we cross that gap? What could we do? The other one is looking at future business models. I already talked to Oradisk, talked to Ben, we know they're looking at different business models because name-util licenses is going to die. Because automation is going to start killing seats, not growing them. So we're going to have a look at token value based outcome based. Let's have a discussion because I think it's important that AAC firms get their say into the software firms and the VC money people. It's just on the podcast. It's come up here many times because you're not in the same phase of design all year long. You're not in the middle phase of design. Maybe you have multiple projects but maybe you don't. And so it's complicated. Is the answer and firms are spending a ton of money maintaining seats that they're not using all the time. The subscription model was supposed to give flexibility to the customers. No, you've got to sign up for three years. There's no flexibility. I think I have a conversation with Andrew Annaghan. I said to him, I believe the KPF hired Goldman Sachs to come in and look at their token usage so they can estimate it for the EBA agreement for three years. That's the level of crap you're putting on your customers that they have to get. An accountant in to help them estimate their token usage. This is insane. And managers. Like you have to have roles in your firm just to manage that process that you never had to have that before. Someone told me the KPF decided that the thing that they were going to do is just delete AutoCAD. Because what would happen is everyone would go into work that start up their AutoCAD and their RABB and just make that grab in the morning. Not there's network licensing stopping them anymore. But every time they started it, then like nine tokens or whatever the token rate is would just disappear. And they were spending something like one and a half million dollars on unused sessions because people were just starting the application. And that part of that is habit. It's a very expensive habit. And so by deleting AutoCAD, then everyone stays in RABB. And that sounds like, wow, you doubt me. Or in some ways I'm kind of slightly pleased because it means that hopefully they won't break the model or the drawings by taking them to AutoCAD to do something which they'd have to do in RABB. So it would be in some way coordinated. But anyway, so that was one thing. And then the other thing that's come along is. So I'm understanding that all the desks kind of introduced some kind of like a higher token pricing for seats of RABB that are doing automation. So you as a company are looking to save money. You're looking to deploy automation. Now, before you might have had like 500 seats where everyone did their drawings. But maybe you're just like going to concertina that down to two with some clever automation tools to do it. Well, that's kind of they're going to cost more. So they want a slice of the productivity benefit that you found for yourself by buying some of their pie software. Because you've lessened the need of the number of seats that they've got. And so that's another kind of problem that I think the industry is there's going to be a rearguard action as we as we move from name to use the licenses to whatever tokens and something else. So these companies are going to come back to you and go with our automation. Even though I'm not be there. It's going to be a third party. We've saved you five employees wages. We want our we want our slice. And I find that outrageous because the architects are the ones who've brainstormed and come up with a solution. And they've delivered the automation and with AI they're going to be delivering more enhanced automation. And yet the software vendors are going to see there. The number of licenses shrink and they're going to come back from for a bigger chunk of whatever it is that you are using. I think that's my my big fear is the software companies and capitalism. I can't keep saying this big. They all have to grow. And if they don't grow, then there's a problem. And so therefore they need to do the great thing. And they've only got one source of revenue, which is customers. And unfortunately you're going to be paying more. So I kind of think this the price of software, even though we're in a world of a eyes, actually going to go up. Because you're going to need less of it. And if you're not me to to hell and paying high prices for the metering, they're going to come after you for a slice of how much that project was worth. Because they need to maintain. But you know how long is the project last when does the software company get paid if you're willing to accept that augment has got an interesting business model where they'll get the client. They'll say how long would it take you to design this data center electrically? Oh, six months. Okay. We'll do it at 10% of the of the cost that you estimate that you will do to do it all doing overnight. And you can do as many times as you want. You can run the run the outputs as much as you want until a set date and go ahead. Okay, fine. And that's that's one model. So everyone's going to be testing this out. The sass world that cloud sass world that these guys thought was going to happen five years ago has is vanishing. And automation is is now that biggest threat. The biggest threat is that customers being more efficient. It always makes me smile. And that's what's going to happen. Small companies are going to start looking for automation. And if you can get the workforce that will work towards helping you finding those automations and saving that money. You can either spend the money on going after more projects and being a bigger company. Or you start actually doing jobs away and become a smaller company. That's quite profitable. Right. I'd like to think that people would keep their jobs as opposed to firms deciding to be more profitable. But again, if you look at what's going on in the tech space, that's not happening. Right. There's being they're just being slashed and slashed and slash because of automation. Yeah. So yeah, I fear. I think you're going to so look at the engineering world. So electrical cars come out. There are normally huge design. that work on cars. And most of that is because of transmission and the engine. You've got body-shop people and you've got your interior design and stuff. But a lot of that is down to the mechanics of car. How comes an electric vehicle? You've got batteries and you've got an electric motor. Most of these things are built on a standard platform. The batteries you have met, say in the electrical engineers, the stuff you buy in. So the only thing that really needs to be engineered is the body, the interior and the suspension. Maybe okay, you can think about the cooling. But the design teams for cars have been hugely slashed based on the technology of the car. And I kind of feel that in some ways the failure of modular has saved a lot of jobs in the industry because buildings haven't become things you order online, particularly easily. But at some point that we all feel that that's an inevitability that there will be sort of a modular solution to a lot of building types. But I don't know how far away that is. I don't know. There's a big graveyard there of companies that have started. Oh, yeah, I wouldn't see one company. They had a load of all of their modules were on an airport. And the reason was because the delays on site meant the factory was turning out on the modules while the site had problems. So they had all the modules in there to store them somewhere. So they stuck them out in the open in an infield. The process is so shockingly inept between the different departments from design to construct. It's a massive coordination problem. Anyway, so we'll be looking at that as well. The great thing, we got an increasing number of AIA large film rant table firms come to the next build, which is great. You have amazing conversations in the hallways. So it's not just the software companies that are showing up to get there. You got to have all side to the conversation there. In my event, the AAC firms are king. So the order desk have come along, which is great, but they get 20 minutes. So does Karnik, who I don't know how many things they've sold. But everyone I've got last year's CityWeft, a company that didn't have any funding that I knew of, or they got 20 minutes. It's a case of we've got a level playing field. Everybody has got that. I wanted to bring the A Games, say exactly what you're doing, not what you've done, but what are you doing? Last year was a bit of a special because we had all the BIMT point-offs in a row. And it was kind of like the first time that all of them had been available. But this year they're all going to be looking at their agentic features and what they're going to be offering and how they're going to bring that in. I've got changed in a year. It's only been a year, but the topic has changed completely. I've got Raven coming. The guy's doing the sketch to Moll in Rhino. He's flying from the States. I hope Dave Miller, a mega-minute stoke, he's got to still give me a few ideas of what we're going to use him for, but he's talking about Levenate. So that'll be great. Now I did ask Hyak, I need to get Circle round to see, they said yes, last year, but I need to circle round because even though what they do is not applicable to the Europe because Timberframe, but mansions, we don't have them. But what they've done with their software is absolutely beautiful. I think it's a demonstration to this industry of you pick one building type, you build one cloud-based modeling solution for it, and you can do the door-eighing things. You can do everything. You can even include people who can't model. You can take a sketch, and with AI, bring that in and turn that into a model. It's a delight. I think that it's just such a shame that they've come to market while there's an idiot in charge. The building market in the States is in terrible, terrible state, but when things do pick up, I think that kind of technology is just going to be an absolute killer for house builders. I reckon there'll be an equivalent of that for office buildings. Are we equivalent of that for, well, if we still have jobs in offices, the hospital version. It's the perfection of computer science. It's what high up is. Everybody needs to see that to realize what's possible in a multi-discipline single solution. It sets the bar. Then the breakaway from that is maybe at a conversation with somebody yesterday, maybe the generic red-air kind of building model is not going to exist in the future. Maybe everything is an expert system. Maybe because how many new types of buildings are there that are generated. Why do you need a generic thing that doesn't really know much about the building type? When you could have a building solution that actually understands, okay, these things are important. The fire regulations in office are important or they can run that all the time. But then that's out there. But the moment we're dealing with the world as it is, I'm just seeing the breakdown of that monolith happening very quickly now. Those modes, I think RBT, file format issues, they'll be dissolved. It always reminds me of this. That was that film with Pierce Broncinal, where he's monitoring some volcano that goes off and to escape they get into a boat. The boat's made of aluminium and the lake that they're going across turns to acid. The boat is slowly melting and the propeller finally gives way and it just goes down to a little stub. I kind of feel that's the future for a lot of software in the AC space. There is a metal boat in a sea of acid. We'll see which bits of the boat survive the journey to the other side. But yeah, it's my brain jumps to visual, I'm trying to send it to myself. My brain throws in these kind of like, "Oh, yeah, that film." Obviously, it's a bit like rowing across a lake of acid. Obviously. Well, this idea of having an in-person event and just the value of that and getting people in the same room to have these conversations for two days is still super valuable when obviously the fire hose of information and the internet is just constantly flowing. It's the best thing I've ever talked about. It's the best thing I do all year and it takes three or four months to plan it. So then a night before we get the speakers and some of the exhibitors and some of the design IT royalty. We have like a few beers, so that's like a good start point. Everyone's super social so they know each other. Next day we get going. There is absolutely manic because every 20 minutes or each room is doing something different. People send teams because they scared of missing things. I think next building, next step we do, next build, both streams do live. So keep an eye out on the website for that so you can join in. And then at the evening it's kind of like more beer, more meats, more little conversations. It's always a great place to hear scuttle but and then the next day is dev which is a different field. So that next build kind of vibe is very much about seeing practitioners. So it's like the fosters and the HOKs and the Perkinson Whirl talking about their workflows and how they're adapting to adopting new technology. But the next day is very much more software people talking about where they're at. Brand new things we haven't, people haven't seen before. Some guy up in Estonia that doesn't have any money but is doing something really interesting in something that I've kind of found out that I might have written up in the year and this guy is really fascinating in Gaussian splines or whatever. And then at the end, it's a bit of a, after two days of talking nonstop, a lot, I mean, all the night's a late. It's a bit like a collapse and it's a great, how the conversations with people that, the first time, she heard Julian Moote from Bentley came, I thought he was just going to be there for the day and gone. This guy's way too important to hang around. He was the last person to leave the bar in the second night and he was super hyped and he'd, he'd like to say, you know, the one thing I've learned from the last two days is that the market is completely open. Talking about Moote, the future hasn't been decided yet, but in London for some unknown reason, there's a lot of signature architects, I guess, who are kind of out there in thinking about workflow changes. And now we're getting to come New York round coming in who are also better f*cking into it, probably have, have equally big ideas of what they're doing. And there's a really interesting mix between US architects and UK based architects. And you sit down and listen to that conversations or we try and take part in it. And you see that they're all doing, they're all doing a lot of things that anybody who thinks that that the AAC space spends as much money and time as agriculture and fishery really needs to kind of like leave that at the door because these people are doing amazing things that all training their own elements, they're all totally experimenting with piping data around all different, all different products. They're all looking at automation. Yeah, it's just, yeah, the stuff that these guys are doing is amazing. It really is. Yeah, and some of it wasn't possible. The exciting thing is that these films are large and they've got people who can program. I reckon in 18 months time there'll be people coming to the next build who can't program, but equally can get their own dispute programs and they'll be coming along and they'll be on the same level because suddenly they have that ability to to play architects have always said to me, well you're not fast as we can't afford, we're built for our we can't afford to have computer programmers and like they can. The world changes, the world changes and every democratization of tools to this is a democratization of research and development. It's a democratization of application development. It's a democratization of data because suddenly it's not locked inside of the modes and you can do what you. The one thing that everyone will realize very quickly is if you don't have your data on your own server, on your own storage, then you're limited as to what you can do. I mean, this is the one thing I'm pretty sure I said earlier, so I'm talking about APS. So I was talking about one of those KPS thinking, all right, so I'm talking about that. The NCPS servers are going to be shoved on the end of APS. So if you want to do a AI check on your data, you've got to go through APS, which is all tokenized. So, you know, however, if you want to touch your data, all of this is kind of open in that respect that you can just go in, pay the money, get the data. The problem is, is that APS has limits on how much data you can take out. It's actually bandwidth protected. So there's only so much data you can take out over, and there's only so many times that you can ping it over a certain period of time. So it's actually constrained. I think that's okay for the next year, where people might want to do, I want to have some dashboards, I want to check my project information every week. Roll forward to when you've got agentic systems that are hitting your data on an NCC constantly. You're going to have a huge token bill, and you're going to run into bandwidth problems because this thing is constrained. And I don't know why it's constrained. I don't know if it's, it's, it's a server limited or maybe they, they can do something about that, but the cost alone would be enough for you to go. This isn't a great place for me to do any agentic, real agentic stuff on my information. And so, yeah, I kind of, APS is 10 years old. It was never intended to be an agentic system. It was intended maybe, it was, it was very, it was amazing. It was an amazing thing for all of this to do 10 years ago, plus to realize that the cloud was where they wanted to be. And every time they got all these applications, they needed a port to the cloud. Well, a nightmare. Can we break down our applications into core areas and then assemble them in the cloud. So when we came to port them, we could rebuild them in the cloud with common, common DWG, common document management, common view. And then third parties could also develop on top of it. Customers could also develop on it. It was a brilliant idea, really, really fast-sighted. But all of this didn't really ever, because inventors still on the desktop, revots still on the desktop, all the cancer still on the desktop. 10 years later, yeah, the, the only real product that they developed, I know of his tandem, that was the first real APS product, but all the real money makers still on the desktop. So APS is a bit of a failure in terms of what it should have brought to a desktop. They were very early in identifying the cloud. Perhaps you would have never realized it's full potential, basically. Never really out for it. Now you'd think, oh, now it's coming into its potential where the Genetic Workplace. But from my analysis, with the limitations and the price, it's a stopgab, it's at best. It's not a, it's not an agentic system where you have huge amounts of data being ingested and processed by AIs as you work and as stuff is developed and dropped into the repositories. It strikes me that all the desks will need to come up with something else. And yeah, software developers should be aware that this is a weakness, that we, the current week don't have anybody visually on them, on the market saying that they've got a genetic solution. It's worth what, this edition of the mag, Kone, Carcola, TestFit, Motif, what's it there? Well, all the BIM 2.0 guys have given their views in the edition of how they see a genetic. And again, you can copy and paste that into Claude and then ask Claude to analyse the differences. If you, if you, that's that's quite a good way of figuring out what boiling down what you're thinking. Yeah, I think how tough, how tough exactly is it was all the other one. But Mars is very interesting and very, if you read it with a ordered desk hat on, it's incredibly pointed as to the. From Mokif. Yeah, yeah. Now a Mokif is, yeah, I'm very interested in what Mok Motif will come up with. I know that they're licensing the parasolid engine. I know that they're very heavily into a genetic AI. I think that's been a bit of a switch from their view maybe from two years ago. But like you said, they're kind of the latest one to the party, but also maybe in the best position to implement. And Amar is an old dog. He is not developing his stuff in the open. Everyone else is, hey, here's some new features. Here's some new features. So it's very easy to go and see what other people are doing. I think Mokif is going to come left field. No one's going to have much of a clear of its capabilities. He's very kind of old school in that kind of don't develop out in the open, don't give your competitors the edge. Yeah, don't show your card. Yeah. And I think that probably puts the worries up order desk a bit because they can go and see what the other competition are doing, the BimTiper and As, but they can't see what motive is doing. And I don't think you'd have much of an idea from the stuff they have sold or showed or tried to sell to this point. And they're a very capable team of people who have a lot of respect in the industry. So there's always that threat. But I think Structrutes impress me incredibly with the conceptual stuff they've added in the last six months, five or five months. And Conic is just, Conic is just amazing. I think I've said this multiple times, but literally the whole time of my life and I've seen software develop, they've developed it for data sets that are certain size. And then over time they end up with customers building more and more. They have to try and find more performance. And so they end up trying to get the. - They get worse and worse. And it's like, oh my God, we wouldn't have built the data set as it was. They started off with just this ability to ingest everything and fly. - Huge FPS numbers. - Right. You just fly into the building, you fly up into the one object, and then you start modeling. And anyone knows, go ahead, there's this. - Right. - Right. If it has the right tool, so it builds on it, it will demolish anything in its path. If it has the right tools, then now it has the right scale, scope, and speed. It's very, very angled towards construction, because it's just seen. It's also quite an interesting greenfield space to be playing in, as opposed to trying to compete against a red end, or they've got their own kind of little sound bit to play. But if that thing delivers in the AC space, so Reb it is constantly, if you've edited it, it's constantly trying to check the parametrics of every single part to see if the any of the edit you made has got any kind of follow-on. So it's constantly in this high-process state. And so that slows it down, because it's got so much stuff to go through. - All these lintries. - It's working out for you, Martin. It's looking out for you. And if it had just been to use more processes, more cause, that might have been okay. What. what. Conic doesn't do that. Conic, you just infer parametrics. You infer relationship between the geometry, then you do your edits, and then you remove it. And it's always done geometry. You just put smarts on it. It's such a refreshing way of dealing with it. - Yeah. - It's. it's. it's chalk and cheese. This is why Reb it can't be this future product. - I want to know what that means. - It's a Britishism. I have no idea where it came from. But. Conic has everything that it's showing, and it's doing like 4D now, and it's doing. Yeah, I'm just. every time. every month, they should take all that stuff up on it. - Right. - You guys are just nagging it. Yeah, it's more like a visual effects workflow than an AEC workflow in that regard. Right? It's just like raw speed and power, and like get the job done. - They've taken no VC money. - We don't have time to wait. We don't have time to wait for anything. And that's totally what they're delivering on. And the thing that's interesting is that they're. I kind of must have gone over and see them maybe a year or two after they'd started, and we went for lunch, and it was like 25 people for pizza, and all of them had been working together for over 10 years, but they were at Brickscat. It was like a family. So these people immediately were a team, and to have that velocity of people all working together, pushing forward, rather than all higher him, while higher him. - You're going to buy Brickscat. - This is a massive advantage. - Massive advantage. - And then. No VC money. It's all self-funded. Thank you very much, Exigen, for buying Brickscat. And so no one's dictating to the many rules. So it. They've got a lot of things that are very fortuitous. Some of it happens to dance where they've got. Exigen realized that Brickscat was going to be really useful for them, and so they exited with that. But also. I just. The first time we did dev, I asked Eric to give a talk about. I always try and have someone who's older, older, generation to be there to try and tell software startups, a little bit about their story about how they managed to exit or whatever. And Eric had me in tears because he literally built a company then he got not greedy, but he thought the great thing was to go off and go public, get a lot of money in. Then that kind of caused problems. The company goes bust. He's like. In that process, he's going back to his staff saying, "Look, to survive, I need to cut your wages and they all went yes." And then kind of eventually I think it did go bust. And then he said he was sat there in front of his bank manager saying, "I owe you three or four hundred thousand euros." The anyway, you're going to get that back as if you let me more money in my start my company again. And. The guy went, yeah? It worked. And it worked. The thing was the story about the staff and Eric and Eric and the staff being Tani, Sympathico, and just being a team. And it was definitely worth watching. All of these people have been through the mill. It's inspiring to hear their stories and to hear what went wrong and what went right. And some of it is luck. They will always say that. But Eric, even though he sold his company, and he could very easily be living on a beautiful island whether you're decided to start again and build another Bim tool because he tried twice before and now he's doing it again. And he's doing it with the same people that he worked with before. But all those lessons that he learned along the way of being applied to to Conex. So, you know, I'm deeply impressed by Eric and the team, but I'm deeply impressed by everyone. How tough. How tough was in this space when nobody wanted to develop a new rabbit. And they just. I spoke to someone at Hexigan. Was it Hexigan? Yeah, I spoke to someone at the CTL of Hexigan. I said, you know, there's only these people who want the next generation of tool but no one was willing to do it. They're all the open-knit guys who are willing to talk to anybody who's listening and after about six months they came back to the house. He'll cost 250 million. No, not interested. And I'm like, there's a guy in India. He's got. he hasn't got any money. And he's willing to sort of develop and have a go at it. And you're a company with unlimited resources. It's a thing that is very short-sighted. And then, uh, until we do other firms that are similar. We're volume players. We can't go and talk to these large signature architecture firms because there's no future in us selling volume to these people. And then AI is coming along. I'm okay. How many volume seats are you going to be selling in ten years' time? Because I think we're in a different world now. And I'm talking about making assumptions, right? Right. Yes. My favorite one is the Phantom Jet, where they designed the Phantom Jet thinking that all future warfare in the air was going to be beyond visual range. So designers playing with loads of loads of racks, loads of missiles. The Americans sent them into Vietnam. And then they were suddenly dogfighting with migs. And they didn't have a cannon. And they only had. They only had these long, but beyond visual-rained, amaran rockets, which made that they were completely useless. So these things were sneaking up behind them with their radar off. And then they were in a dogfight and they had no gun. And so they had to take the Phantom Jet and reverse engineer a gun inside it, which meant that they bloated the nose of the thing so much that you could hear a Phantom Jet coming from ten miles away. It's all about making design. This is like. Yeah, design's really important anyway. Design is important in fighter jets. It's important in CAD software. And I don't think there will be to stock home. There's an amazing boat that capsized the. Oh God. There's this one boat. It's like 16th-17th century boat. And it might be even older than that, maybe 15. And it's the valder sitting in my head. I'll see if I'm looking up. But this boat is. It's. They were building it. And the king heard that in Britain we had an extra layer of guns on our boats. So he midway through design's boat. He says to his guys, "Another layer, more guns." More guns. More guns. And so they do it. And every ship does a test. And the test is they have the full crew go from the left port starboard where they run to see how it rolls. It failed that test because it was so top-heavy with all the guns. He insisted that they go out anyway. So the boat goes out. And the first gust of wind, it leans to one side. All the lower port, the gun hulls, were open. Water flooded in, it sank. It never did anything. It never blew anything up. It never shot a gun. It sank straight out of the. It's a lemon. Yeah, right. Because they tried to build something on a platform. that was not fit for purpose and I wish I could remember the name of the damn thing. It is the most beautiful thing you've ever seen in your life. You go into this darkened, huge whereves and it's there with all the ropes and all the guns and all the paint and all of it. It's just right next to the Abba museum. So you can compare the two. But for me, these stories resound when I kind of look at technology and I look at products that people are trying to extend the life of by building things on that they were never intended to be. And at some point, everyone has to realize software has a life and then at some point you have to have a new generation and that new generation. I would let's extend that to firms. I think it needs to be extended to firms too. The same analogy can totally apply. And I'm saying that because it's a difficult problem and I think a lot of people aren't putting the attention on it that it actually deserves. And so of course you're going to have attrition and those people are going to just go start something from scratch which is literally a small team is going to compete with your big giant thing because the technology is leverage. And yeah, this is a reality. The is winning that first deal that keeps you in business and then doing a really good job on that to the point where you can go and get more business. And then it'll come down to fees. This is the thing is that everyone everywhere, all at once, is going to be rethinking that process, rethinking. I'm being fearful. Every design IT director should be speaking to their board saying this is coming. I'm telling you early because it's you we're a big oil tanker and we are going to have to turn. And there are going to be pitfalls along the way and we're not going to get it all right. But there's enough technology and there's enough momentum in wine seeing I know that means we're going to have a different world and that world is not going to come slowly. It's going to come quickly. They will impact different parts of it at different speeds. If I was an engineering firm like like I would be, if I was a cundles, if I was someone in MEP in structural, I would be seriously evaluating what's going on because they've got 18 months before this hits. The biggest thing that could happen is if large firms bought that technology so that it wasn't available or it couldn't be weaponized by an auto-desk or to be given to everybody. If that can get available to everybody. It could go one or two ways because there were so few companies out there doing it. But people like Andrew have got a proof. I haven't seen the software. All I've done is talk to them they've taught the right talk and they've said the right things that they're going to do. But it doesn't necessarily mean that it's going to be a great product. I'm just guessing that they speak the right things. But that's something I'm watching very, very hard what's going on there because there were so few players and it's got those products are starting to get very, very obviously beneficial. That kind of goes down to people who do startups. If you're doing a startup, you developing a startup to create a business or you're creating a startup to sell to someone. But vast majority of the technologies I see are being built to sell so they can, here's a five to seven year marathon and then some will buy us. The VC companies are there. That's what they want to do. They want a 10 times multiple. Thank you very much. That's the thing with Conic where they don't have VC money. So the incentive to them to sell is not outside. It's internal if they want to do that. So this space has twists and turns that are coming. I can guarantee you that there will be consolidation this year. All the firms are looking at buying. All of the major BIM firms are looking at the market thinking if I add down and maybe I'll buy them, everyone, every single one of them. Yeah, even ordered desk is sniffing around but that could just be normal strategy analysis perhaps because they've got former but is former delivering. I don't know. I don't see many people using it how they're. It's free. Yeah, I don't know. So I talk to our Colola Paul's doing really well. I know snap treats getting six times the traction that they were last year with the AI stuff. I'm speaking to people who are saying we just want to get away from American software. There's that aspect. The whole trade war thing. Anti-EU, there's that playing off as well in terms of I wouldn't be surprised if the EU is going to be more investigative of American firms playing in Europe based on what's happening with Trump. There is definitely a feeling within governments in Europe to replace traditional American software like Word and that kind of stuff with local brew equivalents. That stuff's really playing out. It's not he's making enemies out of allies and that's feeding through to businesses who don't want to be supporting that even though most of the software companies have nothing to do with what's going on. Yeah, victims along the way. Yeah, everyone has a victim in this sad world. Part of the way I was very disappointed with who can AI signing with the Pentagon. I was very happy with Claude. So maybe I can see what happened that the people voted. Yeah, I got both and I still have both but I've spent so much time training the damn thing that I can think I've got. Well, which bits, the perennel which bits are in Claude? Claude's going to knew for me. But by firing away Claude is so different. It's so much better. I feel like a different animal. Yeah, operates very differently. Yeah. The only thing that was so I had to chat to you P.T. And I've been using it like day in day out and then I moved to Claude to do because I'm using it for the, one of the agent tick tools I'm using is based on top of Claude. I kind of realised this is a very clever system. I maybe I should get Claude. So I moved to Claude and then by two o'clock in the afternoon it said, okay, you've had your credit. It's saucy tomorrow. Right. Right. What? What? I had to ask chat to you P.T. about Claude and I went, oh yeah, the way it is. And so then I upgraded my Claude because I didn't know where enough. I could not do without this thing. So yeah, it's a funny and now I'm very glad. And the, I do not like the way they just like dump, oh, we've done legal, we've done accounting. I don't know the way they dump that out like like bonds. But I was very intrigued to learn a lot from the CEO about his view. And he's the one going, I'm warning you now, I'm warning you now everybody, you know, you know, you know, listening. And he can't say he didn't say, yeah, he can't say he didn't say this. This year feels very different to last year. Like I could have gone 10 years in the land of BIM and I just felt nothing's changed. And now in the world of technology, it's moving at such a, I haven't even tried any of the kind of free LLMs or the Chinese ones. It's a very different world. I'm kind of aware that I'm not really an architect. I'm a mechanical engineer by training. But I love architecture and I look construction. I love the creative collaborative process of it. So when I go and seek people, I, as with Martha and Sheriff from Fosters for an afternoon, and they, they communicate to me through the lens of people who do this stuff. And I communicate to them as somebody who watches the development of this stuff. And we somehow meet in terms of I learned from them about their protests and their vision. And on the other side, they learn from me about the ins and outs of software companies and all the kind of shenanigans it goes on there. What's being developed? who's developing what. And so I think you do a great job of just pulling in all, it's such a great diaspora. There's so many different inputs into this industry. No one's got it all. No one can confess to understanding it because it's multidisciplinary. It's a motive, it's human. It's theoretical. It's, that's the thing that's so good about architecture and studying it is that, hopefully, on a good architecture course, they're going to give you all sorts of input. I mean, I remember all my friends in Oxford University were architects and I had bin skills. So I had, I had, I had, I had, or a cat skills, I could do modeling in 3D. And their first project was like, here is the space, build something and in a crit, in a crit, defend why you build, why you build. Right. And I'm like, oh my God, that's so open. That's such a, wow. And here I was, try write a bloody programming Pascal or do some math. Well, you're going to get a different answer from every student in the class, a wildly different answer from every student in the class. Whereas, you know, you're turning in a project that is so prescriptive, you get the same answer 50 different ways, right? Like, we have two different versions of the same answer. This is a, it's an entirely different thing that you're talking about. And I went and signed on the crit and I, it was horrific to watch. There were rears. Yeah, yeah, it was, well, no, there was, I knew the, as projects got bigger when people pulled their hide than there was mortiers to come. But to see, you know, their tutors ripped them apart and then other people, yeah, their fellow students ripped them apart. It was really quite a, it was like, oh my god, this is so much more than I expected. With that liberal, with that liberal kind of freedom of creativity comes an act of being judged by your peers in front of everyone. You know, it's almost like acting, acting that you have to be an actor, you have to be able to project and you have to be certain of why you've done something. So yeah, I'm very vulnerable in that way. Yeah, I think it's, and people call it the fun job, right? The fun job is designing. It's actually very vulnerable and you have to be willing to take that criticism. But it's just, yeah, I think it's good for people's characters because how often do you have someone coming on to sing that shit? Or, yeah, it doesn't really think about that, do you? It doesn't happen. Especially in my, you know, you were either right or wrong. Your program worked or it didn't work or your math was right or wrong. Well, they're, it's so subjective and you have to be, you have to be a single architecture firm on your own, basically, as you study. Then you work in groups and they have to work with somebody's a bit of a lemon who's holding you back and then when I was, I was still a uni or whip. So we were sharing, I was sharing a hatch of the architects. They were up all night all the time. I mean, they might have worked, they have to do was just insane, but they enjoyed it. And that was, yeah, it's my, one of my friends was most better at drawing people than he was at drawing buildings. And so when he was submitting work and I was like looking at his drawings, there'd be like a huge face covering like a quarter of the other. And I'm like, what's going on? Is that he's like, I didn't have time to like finish it off and then I better drawing people. So there were always more people in. Right. I was just like, oh, how sneaky. Yeah, obviously didn't really, that didn't fly very far. There's stories of people taking their exam when the exams were done on drawing boards for architecture registration. There are many stories of people covering up unresolved issues on buildings with trees. Yeah. (laughing) This vegetation to the rescue. This time I was helping one of them submit his model in drawing in CAD. And he was marked down for doing it in CAD. Yep. They wanted the artistic element, I mean, they didn't want the computer drawing of it or the model that was generated. Which I kind of felt was very backward. I wasn't writing about CAD at that point, but it was, yeah, he was a fricking from my knowledge. He was, when release 13 came out, we were still sharing a high-reson Oxford and I was writing about CAD. And so I had a early beta of 13. And he modeled the Radcliffe camera, which is in Oxford, this amazing Stokeville Library Building Pop-Oxford University. And we worked out that if you modeled one-twelfth of it and then arrayed it around, you could end up with a good model of it. So it was perfectly symmetrical. It was that point that we just kept crashing on release 13 and that was constantly ringing up tech support in the States trying to get fixes for what's going on here or something. So yeah, that was, that was, that there's always very useful. - There's the difference between an engineer and an architect. An architect doesn't call support Martin. We tried to figure it out all night long every night. (laughing) - I made a lifelong friend, actually, I was both of us so much in a Hologres tech support and her dad was one of the singers of Jefferson Starship or Jefferson hair pain and say, "One of those bands that would come to your very name "or whatever." And so yeah, I still told her now on Facebook. - She's nothing to do with it, but yeah. (laughing) Yeah, it was just one of those things where we were trying to use something in anger and couldn't believe that it just kept failing. - So yeah, so yeah. - This has been a great conversation. I want to think the audience for hanging out till the end here for all the stories. I love all the stories and it's great to catch up with you. And I wish you all the best with next build and next dev coming up in May. We'll put links to that so people could go in the show notes for this episode. - You can watch the next build online. Both streams would be video and live streams. And if you obviously are gonna be five hours behind foot wood, I know, it's whatever. I think you can just go ahead and scroll it backwards or forwards. But within one or two weeks, there'll be available on NXTAEC.com, which is where we store everything. - Both show. - Yeah, and you can watch last years and the year before that there as well. Martin, that's been a great, super fun. Thanks so much. - See you there, I'm always good. Thank you. - If you made it this far, thank you genuinely. These conversations take real time and energy to put together and knowing you stayed until the end means a lot. Two quick things before you go. First, if this show has been worth something to you, if it's shifted how you think or introduced you to someone worth knowing, consider becoming a supporting member. Less than 50 people out of 25,000 listeners are currently paying for this to exist. And for $12 a month, you get ad free episodes, full access to my deep dives in the leadership edge newsletter and other special content I save for members only. Head to trxl.co and click the join button. Second, subscribe to the show if you haven't already. It's in your podcast app, it's free, and it takes five seconds, and it's genuinely one of the most useful things you can do to help me grow this thing. All right, that's it. I'll see you in the next one.