219: ‘Keeping Architects in the Driver's Seat’, with George Guida

219: ‘Keeping Architects in the Driver's Seat’, with George Guida

Evan Troxel: Welcome to the TRXL Podcast. I'm Evan Troxel, and in this episode I sit down with George Guida, an architect, educator, and entrepreneur who is at the forefront of integrating AI into architectural practice. George brings a unique perspective shaped by his education at institutions like Oxford Brookes, the Architectural Association and Harvard Graduate School of Design, as well as his professional experience at Fosters and Partners and others.

As the founder and CEO of xFigura, George and the team are building a collaborative generative AI platform, specifically designed for architects and designers. xFigura acts as an AI sandbox that streamlines 2D 3D and visualization workflows. Addressing the challenge of tool sprawl by aggregating multiple AI models into a single cohesive environment.

This approach keeps architects in the driver's seat supporting ideation, collaboration, and experimentation without the need to juggle many disconnected tools. Beyond xFigura, George is associated with ArchiTAG, a design and research practice that explores the intersection of architecture with advanced technologies like AI and immersive interfaces.

He's also a frequent speaker and advocate for applied AI in the design industry, contributing to conversations about how the profession can evolve in the face of rapid technological change. Our conversation today covers a lot of ground. From the practical challenges of adopting AI in architecture firms to the philosophical questions about the future of design interfaces, we discussed the potential of AI to transform workflows, the importance of maintaining architectural judgment in an AI augmented world, and even speculate about future interfaces that could decode thought directly into spatial designs.

George brings a balanced, multifaceted perspective to these topics, drawing on his experience in academia, practice and entrepreneurship. So whether you're still skeptical about AI, curious about its potential, or already experimenting with these tools, this conversation offers valuable insights into where the profession might be headed and how we can thoughtfully navigate these changes.

So now without further ado, I bring you my conversation with George Guida.

George, welcome to the podcast. Great to have you. Great to meet you.

George Guida: Yeah. Excited to be here. Thanks for having me.

Evan Troxel: Yeah, it was great to catch up with you at the AECtech conference. I know we chatted before that, but then we actually had the chance to get together in New York City a few weeks ago and, uh, got a, a glimpse at what you're up to during your short presentation, which I think was well received. Um, anytime you're kind of mixing a beloved.

Kind of, uh, tool interface like Grasshopper with some new technology. I think all the, the nerds in the room were very excited about that. So, I'm excited to talk about what you're up to, but welcome to the show and, and I would love it if you could start with kind of your path. How did you get to where you are?

Let's go back to, I know you started in Europe, right? So let, let's talk about where you've come from.

George Guida: yeah, it was great to, great to catch up and your session was awesome with. Some of the leaders in industry at the AECtech. So great, great episode in itself.

So I go back, so I'm originally from Italy. I, um, uh, then moved to the UK and I took the traditional architecture path and went to Oxford Brookes, went to the Architecture Association and worked, uh, across two firms, one of which was Foster and Partners. And that was real tipping point between the AA and Fosters, which really opened up, you know, this whole computational design, but also computational side of things. I am the, I'm slightly different from the typical transitioning person, from architect to software. 'cause I'm very proudly still an A for architect as well, and still run, uh, in part in architecture practice.

So I still very much care about these two worlds.

Uh, foster's really exposed this whole computational technical side. So then I went to Harvard, started my own practice. And that one thing after another led to X ura, which I'm sure we'll dive into everything, ai, everything, concept, and, uh, happy to share more on that.

Yeah,

Evan Troxel: Cool. So, so this, you, you said, you know, you, you very explicitly said you, you went to school in the traditional practice and then, so obviously we've seen a huge evolution, uh, maybe not in academia, I don't know. You're actually still have a foot in academia as well, you teach, right? So I, I'm just curious, like what, what do you feel like that brings to a, a modern architect turned technologist?

Like what does that give you, that you feel is useful in the way that you view things, in the way that technology is kind of changing our landscape of architectural practic?

George Guida: yeah, it's a great question. I, I am, I love the academic world and I. It's really a medium to test things out and to really get inspired and, and see through the eyes of students of, you know, what's next, what's coming. There's obviously a bigger conversation, especially we feel it a little more so in the UK than I do at the moment here of why do we need to go through five years of architecture when everything's gonna change by the time I graduate?

And so the, that is exactly the type of questions that I teach now in

an elective at UPenn and next term, back at Harvard. So it's rethinking, you know, education, but also thinking our roles as architects, as designers, as future, future technologists. And this, you know, in the future of work.

Evan Troxel: Interesting. Well, so let, I mean, can we dig a little bit more into that? I, I'm very curious what the current conversation is. Like it's gotta go deeper than, well, why, why do we have to do this now when we all know it's gonna change? 'cause e every academic institution has gotta be dealing with that at some level.

Right? We're, we're watching. I, I think that, that the trope is that academia is always preparing students for the way we used to practice architecture. And you could say that basically with any, any era, but, you know, architecture is, is a slow evolving practice traditionally has been. Um, and not only that, but like the technology that's been introduced, it's kind of the new version of.

Doing, you know, producing drawings, right? I mean, that's what we, we called AutoCAD, the diGuidal drafting board. And then Revit is not that different, even though the, the workflows and the processes are different. Right. And you're, you're working primarily in 3D but the outcome is still two dimensional documents.

Right. So I'm, I'm curious what the conversation is in your circles around that.

George Guida: Yeah, it's a, it's a great question and there's one of the key topics is clearly from the ethical and agency perspective, there's a huge scare, let's say, even from academics, even from students of, you know, how we can be positioned when using ai. There's always the, the early camp, which is, you know, don't use ai.

There's not clear policies, especially in universities of where students, where electives or classes should be positioned at. Uh, but there's a, a real conversation to be had in terms of the emerging, the emerging agency that these tools can give. And so I like to call AI augmented imagination versus, you know, the traditional artificial intelligence.

'cause learning how to these, how to use these tools really opens up that solution space. Now the second topic is clearly from an ethical perspective, we need to have certainly deeper conversations and simply understanding basic technical understandings of how these models work, what underlying technology, and what a black box model actually means. And that comes hand in hand with our standard of care, responsible control. What does that mean in practice? What's our duty, uh, to the, you know, really to the public once we graduate after?

Evan Troxel: Yeah.

George Guida: So process agency is really evolving.

Evan Troxel: It that, that ethics question is interesting and I, I totally get it. I architects, well, first of all, there's liability, there's risk, there's professional licensure, and I could see that becoming more important in a world where the tools have democratized access to creativity. I don't know if that's the right way to say it, but that's kind, I think that's how a lot of people feel that it, that it's what it's doing, right?

It's just like, oh, now I can prompt a solution. And that solution is in air quotes, right? Uh, it's not really, I think architects know it's not a solution, but at the same time it's like we've got a responsibility and so we're going to approach this evolution with that in mind, and, and then I think there's a lot of the world that's just moving on.

In spite of that, right? It's just going full speed. The technology world is really pushing and I think there's a huge push to just kind of own these spaces. Rules be damned, right? And so it's like, I'm very curious what you think about that because it's definitely slowing down the adoption, but for very good reasons, right?

I, we have to be responsible, we have to do this responsible, and we're asking those questions, but I'm curious if anything's actually coming out of having those conversations from, from your perspective, are you seeing anything literally be addressed when it comes to ethics, governance, responsible control, when it comes to these black box models?

George Guida: It's a great question. So these, these are the exact questions I'm now actively collaborating on with the A I A and reba. I'm actually a British architect in their AI ex, in their AI expert advisory group.

So, you know, how can practices from small to large position themselves, especially from a policy, internal policies per perspective, vis-a-vis clients end user agreements, NDA agreements, confidentiality and all of that, how can we well equipped in this landscape, especially also since everything's getting much more competitive, moving much more faster. So there's a clear distinction, let's say the European side versus the US side. In the eu, actually in the uk it's not far off. The EU AI Act certainly put a lot of guardrails in terms of AI developments, in terms of transparency of models, which it's the usual question, you know, guardrails versus stifling innovation.

Where do you position yourself?

The US on the other side is much more

aire. It's more, you can think of it more as a train, which is great for in innovation, but a train without, you know, distinct breaks. And this falls more under like the state to state level and the clear, the European model is the Belgium effect.

You know, Belgium adopts policy and it kind of, it's the Brussels effect. It's, it then spreads, uh, in the US the equivalent is the California effect.

We've had a lot of conversations with NCARB in their, you know, national meat. I gave workshops on regulation, AI, and architecture. And so how can we, you know, bring that into industry?

California is very progressive in this compared to the national conversation. You know, let's talk about DeepFakes. Let's talk about transparency of our models. So then bringing it into architecture. Okay, how can NCARB and a, i a put up a clear stance so that as an industry we can be more responsive or make custom solutions? So I think there needs to be more conversations. There needs, we need to be more proactive as an industry versus where we're currently standing as being more reactive. And so, hopefully in the next few months, at least on the a i a side, we'll put out a report which gives us advice of, you know, how, how companies can, you know, be stand, how can they be positioned in this age of ai? It's a little more proactive than, than reactive, let's say.

Evan Troxel: Do you have any, just kind of the, the, any information that you can share regarding that or just your, it may be even just your thoughts in that space around how companies can be more proactive without maybe having to wait for this report from the a i A. Just, just as maybe talking points, but also things that, to get the gears turning in the head so that we're not in that reactive space, so that we can start to reposition ourselves into the more proactive space in regards to AI in these tools.

George Guida: Yeah. Uh, it's a great question and I've, as part of my, let's say as our architecture practice, we've done a lot of cons, AI consultancy work for

medium to large sized practices. So it goes from like working with management on AI strategy to custom tool building to onboarding training that was pre x una. The clear first step is management approvals, so there's, you need, it needs to go top. It's, it's an equal conversation. It's come bottom up from, let's say the data users. You obviously need management approval. And so there's a clear cultural shift happening. So the first step is very much sitting at the table once a month, regularly, once every two weeks, you name it. Let's have a conversation of the state of affairs, uh, and let's see how, you know, how that affects the day-to-day practice and what that means for responsible ai. Every practice, I think, should, at least in my view, should come up with its own policies, but its own stance of defining, you know, how do we position ourselves vis-a-vis our clients, vis-a-vis our contracts. And, um, yeah, it's, it's definitely a, a needed a needed conversation. Uh,

but Yeah,

Evan Troxel: I mean that when you think about. That cadence that you say is required, which I agree because things are changing so quickly. Like where would is the best place for people to get that information? Do you feel like, I mean, is it just through osmosis of absorbing the fire hose of news that's shooting on everybody's LinkedIn feeds all day long, every day?

Or what is there a, is there a place that you feel like really can dis distills it down so that people can have the conversation on what has changed literally in the last two weeks that we need to be considering when it comes to these conversations?

George Guida: Yeah, it's a great question. Um, oh, okay. So some resources, high level,

Evan Troxel: Mm-hmm.

George Guida: uh, with Reba, we put out maybe a few months ago, just a few months ago, the Reba 2025 report that covers, let's say, you know, across, um, it's AI across spectrums. It goes from education to policies to, uh, yeah, ethics. There's, there's a bunch of interesting resources just there. A, i a were due to launch the new version, but last year they also had a report. Then there's several books in the space. Obviously, publishing in the AI space is a big risk because the moment, the moment you send it to your publishers and the moment

it actually releases, It's it's pretty much dead.

Uh, but I, I do believe, uh, Phil Bernstein's books are really great and he has, he, he took a really proactive stance in

one of the latest, the older now, uh, Reba books. Um, from a webinar, let's say online resource perspective, you're welcome to listen to the ncarb, uh, webinar, uh, that I hosted as well

on regulation and ai. Uh, apart from that, uh, yeah, link it in is a great

resource. Twitter or X is actually surprisingly a space,

Evan Troxel: It's just a lot. Right? It's just a lot.

George Guida: Yeah.

the AI noise is relentless.

And so, but I will also say that you don't actually need. A large team, there's this common misperception that

you need to be the Fosters, the Thom to, or the, you know, the big practices to engage in these conversations. Bring it down to lower level.

You know, let's have the interns or let's have junior staff really be actively engaged and at least, you know, starting that conversation.

Evan Troxel: it, it's interesting to think about how quickly that landscape changes like you and, and so anything that's in print you're like, like you said, is probably outdated already, right? As soon as you send it. So it, I, it seems to me like this is probably one of the more o overwhelming points when it comes to this conversation is like, how do we even keep up with the current version of the conversation?

Uh, it's moving faster than ever and that that's something that firms didn't have to. Used to do. And so it's hard to modify behavior when projects are like, they're the train in the architecture industry, right? That that train's going and we can't put on the brakes because we've got deadlines. We've gotta put all of our focus into these.

And so it's hard to take your focus away and just keep up with this constant scrolling of, of information that's coming at us and changes

George Guida: Well, the best place is maybe this podcast and just,

Evan Troxel: maybe. But I, I don't know. It's, I like to think of this as more broad than, than specific, but there has been a lot of talk about AI for sure. But, um, and, and there are some specific AI and a EC podcasts that, that are probably better resources for this kind of thing. But I'm just, I'm, I'm very curious like what your students think about the, uh, inclusion or the infusion of AI into.

academic side of things, are they excited? Are they nervous? I mean, to me, and I'll just, I've been saying that I've been kind of theorizing or just humbling, um, ideas around, and I'm, my latest stance is I'm extremely concerned for students because what's actually becoming very apparent is that experience, wisdom, time in the trenches is going is, is what is actually gonna be the super important part about getting projects built.

Um, licensure is, I think, gonna be more important than ever, just from a risk and liability standpoint. I, I don't necessarily agree that the, the way we go about licensure is, I, I, I think that's gonna have to change as well. I think that's gonna have to evolve. Um, that's always been. A, a lagging, um, process, I feel like, to the way that practice actually runs today.

But, um, I, I'm curious what your students think and if they're seeing the same thing or if they're like, you maybe more excited about this augmentation. And, and I, I would love to hear what you think about that theory because, um, I'm concerned for the younger people not having anything to do because the tools can do a lot of it, and you also need that extreme oversight part of it, um, to really, if you, if these tools are being used to really go through, like with a fine tooth comb and make sure that everything is still falling under the standard of care, that there aren't errors and emissions, that there aren't hallucinations, et cetera, et cetera, et cetera.

George Guida: Yeah, it's, it's a, it's a really great question. And, uh, the answer is that we are experiencing a, a, a broad cultural shift. And it's, it's not specific just to architecture, it's across all, uh, it's across all industries really.

Uh, there is, uh, it really depends. Per school, I'm clearly positioned either at Harvard, UPenn, I also taught Northeastern and Wentworth.

There's, there's definitely more of an interest in these schools, uh, towards these, um, as an average, but I would say. We've, I've been to several conferences and seen the academic landscape very much, and there is still that healthy share of skepticism, which I certainly do. Welcome. Uh, let's, let's be critical of ai.

Let's talk about our agency. Let's talk about a role in the upcoming years. And emerging trend, uh, is definitely, uh, or a common theme, I would say is certainly the question of energy consumption of these models.

Why should we be using these when they're, you know, they have such environmental impacts?

And it's a super valid question.

And, uh, the big, you know, the, the big companies are not always transparent for the consumption of, let's say a pre Google pre search, you know, pre AI search, Google search, and a post, uh, ai, Google search. Like what's a, is it a times five factor? What's the implication? What's the implication of data centers on supply chains?

Lithium mining, you, you name it, like the supply chain is so complex. Uh, so that's the environmental side. There's clearly the agency, the control side, which is definitely apparent, but I would say, at least in my classes, we go from ethics to building tools versus using existing workflows. It's very much about, you know, learning how to vibe code, learning how to, from no code background to building functional apps, 3D viewers or actual tangible interface. So having control over that end result over the workflow versus just being, let's say, passive reactive or strictly from the creative side. So that really, this exercise, at least from my perspective, really opens up the possibilities and how students can keep control and, you know, keep agency within that process. Uh, so there's, there's definitely, at least our final review is in a week. Uh,

so it'll be exciting to see, uh, some of the, the final results. But, um. There's, uh, yeah, there's always this mixed bag of excitement and why not? I can dive deeper into risk and liability if, if

Evan Troxel: Yeah. Well, I'm, I'm also curious, who do you think this impacts more, or, you know, more quickly even, maybe, maybe it's more of a timescale than a, than a question of roles. But, um, I'm, I'm curious what you think about that, at least, you know, in the, in the short term future. Do you feel like there's gonna be people in industry who are more impacted than others?

Do you think it's generational in that way? Do you think it's role based in that way? Do you think it's time from graduation? Like, like do you feel like this really impacts new graduates, or do you feel like there's still gonna be the tool operator hires for, for these larger firms because, um, they have absorbed or they have the ability to absorb new technology and new concepts so much more quickly?

George Guida: Yeah, it's a great one. Even if the a a c industry is, you know, still today quite slow in terms of adoption generally of new technologies. I would still say all of the new graduates coming out now will certainly live this on a day-to-day basis. AI tools will be an integral part. It might not be sold or, you know, described as ai, but underlying technology will certainly be, be there.

The, the, you know, the garner hype cycle, we're, we're, we're kind of the use case stage of that is, is starting to pan out.

Evan Troxel: Hmm.

George Guida: Um, what the va what. At least a question, at least from a practice perspective, is very much do we invest in our people or do we invest in new technologies? Is like huge dichotomy. Like where do we position ourselves?

'cause clearly what you described is to be the best architect, you need to invest those years of, you know, the grind of man manually making the cad, the rhino, the details

and learning to be in the trenches. But the

Evan Troxel: I don't know if it's tool, I don't know if it's tool based. I feel like it's more just like, oh, we've been through it. And so of course there's like the waterproofing details and Right, the hold downs and et cetera, et cetera. And the shop drawings like are, do they actually match what we designed for, do they match the design intent?

But there's also just the, you can't do that because in this climate zone, like I, you know, and this and that and the thermal expansion, and it's like, well that's locked in here. And so that's where I'm getting at like, like that oversight idea. It is more of like the battle tested, we did it wrong 40 times and then we figured it out and this is what works.

And, and there's like two guys who know that in the firm.

George Guida: Well, my, let's say my provocateur hat on

to putting it on is what if we can invent that enterprise knowledge within an AI model

Evan Troxel: For sure. Can you do that though? Can, can, we can. Well, I know we can. The question is, are firms doing that? Are they investing in that part of it? Some are for sure, but a lot of 'em have no idea how to do that or like, or what is not the, the slowest, most painful way to do that. Because I think everybody thinks, well now I need to write everything down that I know people don't even know what they know until it comes up with a situation.

Right. So,

George Guida: We have gone through this cycle with several practices, um,

medium to large actually, in building custom tools, and I'll give you one example of what this could look like. From the, for the technical people, this means it's an agentic rag model

retrieval augmented generation model. So essentially at a high level, you're removing the hallucinatory aspect within a traditional LLM, and you're empowering the LLM to essentially access your enterprise knowledge, drawing sets, uh, RFPs, project proposals, any documentation if you want, even all your whole drive or SharePoint, you name it. That means we're adding a semantic search level on any, you know, any types of documentation. So it's, it's becoming more accessible to the smaller firm. There's certainly, it's say, a cost associated in preparing those drawings, going through archives. There's a manual process of course, in that,

uh, but for the data ready practices, you've overcome quite a bit of a step.

If you have the intranets, the open assets, and these system, you're already a huge leg up to be able to plug in an ai. Rag model and gentech rag system and then move fast and, and then you're a full AI enabled practice,

Evan Troxel: Can you, can you u can you explain that? Without all of the AI terminology, like without I, because I, I know what you just said. I, I understand. But I know there's people who are like, what's rag? What, what is semantic search? What is this? And that and the other, like, explain it without like, like you're an architect, so you know how to talk to architects, talk to architects who, who are not technologists.

George Guida: Yeah, of course. Um, I'll give you a, a fun example. This is, this was, let's say more jovial in, in passing, but, uh, from a, a big firm, one of our clients, um, they have one of the lead architects is, um, retiring. So he's got maybe 40 years under his belt of experience. And in conversation we were like, what if we could, uh. Train a model based on all Billy's or Bob's knowledge. And so could we record him for 24 hours? Maybe this is not the best example. Could we record all of his knowledge or take all of his

past projects, drawings, details, uh, and embed them into some sort of local collective intelligence? Okay, let me, let me bring that back a second. Traditional search is very much, um, is very limited in the possibilities. Um, when you search a document, it's very difficult at times to find that correct detail, that correct drawing or RFP. Now, AI unlocks, let's say, an understanding of meaning within each page, within each paragraph. So it's essentially summarizing or extracting, um, a new level of understanding of all these documents. What that means is that we can actually find documents and then generate new texts. New RFPs, new drawings from that past knowledge, and so you are moving at an incredible speed compared to previous, you know, finding the right document, not getting the right fit to what you're searching for and then informing your next step.

Evan Troxel: And, and the rag part of that conversation really comes down to kind of your firm's way of doing things, right? Your vocabulary, your terminology, your concepts, the way you do drawings, it's more of a specific approach to the way that you operate, right? So that's what RAG is really correct me if I'm wrong.

George Guida: That, that's right. It's, it's based on, it's trained only on your own knowledge plus the chat GT five model, for example, the latest model. So you're connecting all the power of a large language model with all the knowledge from your practice, and so merging the two together. Within the, this is all data secure.

If you're, you know, if we've built custom solutions, there are some on the market today. Data security is key, but it means any junior level person talking about like investing in the person or investing in the technology, it's supercharges the junior person entering into practice and means they can go or at least have contextual knowledge of, you know, what does it mean to go into DD stage

or like, I'm at DD stage, I have no idea what to ask. And the way I grew up, obviously similar to many in the audience, is, um, ask the guy next to you, go to the partner. Like, do that

manual or like, go to the actual library and find the doc. Now it's just so immediate. So you can, there is certainly still that value of empowering all even that junior and really pushing that process.

Evan Troxel: Yeah, I mean, the way that used to happen was exactly as you described, or it was just through osmosis of being in an office, right? You would just overhear the project manager yelling at the contractor about something on the phone as they pace up and down the hallway and it's like you can't not hear what's going on.

And that's part of the mentorship process. It's like you're men being mentored through osmosis of being in the office or you're, you're just hearing the conversations that are in passing. And I think that's where, that's where like there's all this gold to be found and that to me is the really difficult stuff to actually pull out and, and codify so that it's somewhere in some.

Rag, right? That allows you to pull from it. But because a lot of times, you know, stuff goes into the email, but it's not everything somebody's thinking, so you can't just get all of their email. It's like, well, that was the stuff that was safe to say in an email. There's all this other stuff that was way safer to say on a non-recorded phone call, but it was still needed to be said kind of a thing.

And that's the stuff that's not getting captured in any way. That's the stuff locked into Bob's brain who's about to walk out the door. And so it's not a foolproof solution, but it is better than doing nothing and just saying bye as Bob walks off into retirement. Right.

George Guida: Yeah. And that really, it, it proves the thesis that early adopters will certainly reap the benefits. Now, you know, the democratization, the access point will certainly flatten in time. So yeah, dive, dive in.

Evan Troxel: So, so do you have any kind of. Examples of how people could do that right now. Like what, what is out there right now that isn't maybe a custom solution to have a chat GPT five reasoning model, look at a rag, like is there, how could somebody spin that up and just give it a try and, and actually say, Hey, there's value here.

And then show that to the partners and say, we, we should be doing this.

George Guida: There's two, two examples I'll give. Um, the first one, the hello world. Let's say the entry level tool to use is one by Google. It's called Notebook lm.

Conceptually, this means drag and drop your 10 to 20 documents, your notes from past meetings or whatnot, and then you can retrieve information from it. So you can literally make a podcast of that embedded text or those reports while you're walking to work or something.

You're preparing for a meeting. You can literally prompt them and say, make me a podcast about this and make it in this tone. Or like, for five minutes

you can make diagrams, videos, presentations, like that's like the hella world. The second step, maybe not to jump into, um, other tools is our own. So xFi good.

Now we're, we're looking into embedding these systems internally, uh, and that will be much more accessible in the coming few months, so

happy to chat.

Evan Troxel: Yeah, I mean that if there, if there's a way to do it now where you can actually look like, so Notebook LM is a great example. I, I, when, when you were saying that, I was thinking, oh, that's a, that sounds like a great way to study for the architecture registration exam, right? It's like, oh, let's talk about risk and liability and, and a i a contracts and just make me a podcast about that, that I can listen to on the way to work, um,

George Guida: And make it fun

Evan Troxel: And make it fun, right? I mean, because the alternative is to look at, you know, the flashcards or the books or however we, however people are studying now too. Um, I, I know there are a lot of, um, on demand kind of recordings and videos and things like that, but if you can make it fun and in a certain tone and explain the concepts, uh, in a way that is within the timeframe that you have when you're commuting, that sounds amazing.

And so I think that's a great example that you just gave, because it applies way more to studying for exams. It could be, I, I had this, these challenges coming up on my project. We have to, we're gonna figure out a way to do it. Uh, how have we done that in the past? How have we gone about solving these, these kinds of issues before?

And, and at least have that under your belt so that when you show up, you are supercharged, as you say, to kind of take that challenge on and, and bring up those ideas for that discussion that's gonna happen around that issue.

George Guida: We've had clients who, who have

used this, these implementations we've built, and they've used it in real time on a meeting with a client.

Okay? right there. on the spot. Okay, what did we do in that past project? What's the summary of, it's like an immediate answer.

Evan Troxel: Yeah. I mean, and I think there's probably some, uh, allergy to showing the secret sauce sometimes. And so you're clearly saying no. Like this is, we have this ip, we have this rag model, we have this information at our fingertips. Why, why would we not do it right in front of our client and show them that we have these, this kind of capability to solve the issue way more quickly than, oh, we're gonna go back to the office and pull, pull this out, and then bring it forward later.

I mean, that, that seems like a no-brainer, but at the same time, I'm sure there's an allergy to,

George Guida: Yeah,

Evan Troxel: to showing the, the secret sauce

George Guida: an analogy, it's an allergy, which is synonymous to what the Val evolving values of architecture are

because there's an inherent or implicit push towards efficiency first,

and that's more industry led than anything

versus what we obviously want, which is, you know. Uh, to produce better architecture and to produce more sustainable designs.

Like there's, there's so many different values or lenses upon Atric, you know, approach The same, same question.

Evan Troxel: You talked earlier about this kind of idea of augmented intelligence, right? So tell me what you mean by that. And, and, and I guess I would love to hear some examples of it's like, whoa, I would've never thought of that. I, that to me is kind of, is what I pull out of a general statement of augmented. It's like these tools give you access to connections, or they make connections between things that maybe the person who's running those tools has never had themselves.

And so, so tell, tell me where you're coming from with that.

George Guida: Okay. I'll give you two examples. The first one is from high school students, and then I'll give the X vlu example.

Evan Troxel: Okay.

George Guida: The first one. So, uh, last year I collaborated with a company called monday.com, their foundation specifically, uh, in collaboration with, um, some, with the larger team. We built out a full curriculum for 600 high school students on design architecture and 3D printing.

These are students who were taught first how to sketch, you know, entry level, let's keep agency, let's do everything manual,

and then empower that workflow to go from zero to product through images, 3D models, 3D printing to actually tangibly see something they imagine. Let's empower that imagination.

Let's give young students the tools that they can thrive to really push at least that vision of what they could be or what they could do. And that was super empowering. We had hackathons. There are students who, you know, didn't have any background or really knowledge beyond the Snapchat AI feature, the Instagram AI feature of like how to do these things. So super, like super validating, seeing, you know, giving a tool to that imagination. Now the second one is obviously from the ex URA perspective. We see this day in, day out, uh, especially like recently, we hosted a webinar with big and Zaha one. So our bigger clients using the tool and really to see the quality that they're reaching from an early massing model or a sketch to do iterations and push the design direction without even going too far into the physical model or the massing model. It really poses like a separate question of like in concept stages, how, how much should we actually rely on the, that manual adding all the details to make all the options that spent three weeks making like five options or if you know how to use graph ar, push that needle when we could have a one day sprint or a shorter period smaller team potentially. Really map out all that solution space and then go manual after.

Evan Troxel: What do you think that does to the, to the profession? Because you, you explained like a really exciting thing. I totally get it. And faster with less people. Right. So what do you think that does to, to the way that architects work? Or, or maybe even just, I don't know if we know what it does, but, but like how, how do you see that affecting the number of people in the profession?

Do you see that? Like, we just don't need as many people in the profession if these tools become super prolific inside architecture firms?

George Guida: Yeah, there's multiple ways of answering this. There's not obviously a clear answer for it, but um, certainly what we're noticing is if you are a great designer before pre ai, you are an awesome designer post ai.

Evan Troxel: Great is less than awesome. Yeah, for sure. Okay,

George Guida: Yeah.

Evan Troxel: get it.

George Guida: Uh,

Evan Troxel: that's the augmentation.

George Guida: the huge augmentation, it's

like, like you can really, really push the needle and that's those mark, the outliers versus the noise, which we're commonly seeing

you, the Instagrams, the links, the general posts. If, yeah, if you embrace these tools, learn how to use them, you

can certainly have a different competitive edge than the common perception of a flattening curve. Uh, and, and there's, I oppose the, the, the idea of homo ai leading to a homogenization in like aesthetics, form and typology. I certainly can push the needle.

That solution space will really, you know, you can create crazy hybridity, crazy shapes, crazy forms if you know how to use them. Um, there is, uh, a huge value, certainly in at least po in terms of positioning oneself as an office. It's important to be aware of its, you know, risks. That these could bring in the next decade potentially. Uh, what I, you know, some potential roots for this is certainly keeping an eye out of these profic peripheral industries. And so we've commonly discussed pre ai. Obviously architect as developers operate as prefabricated. You know, there's so many models that AI can really supercharge us versus, you know, why do we have to wait for real estate to then hire the three architects to do the amount of work that a whole firm would do when we could flip the model?

There's still democratization on the other side of the spectrum, and let's really dive deep now in architect as developer or, and that's, these are like the common use cases. There's, you know, it's a little, it's died down of course now, but metaverse vr, spatial computing, uh, there's, you know, real estate, um, there's a whole world of peripheral.

I'm not saying. This should be the trajectory, but there's obviously that, you know, to keep these things in mind is definitely something to be on the table. There could be potential gaps.

So imagine I've described fewer people faster,

which is not necessarily the values that I'm pushing for, as opposed to higher quality designs, higher quality, higher value to your clients, which you could upcharge by itself. Uh, but you can, um, but there's, uh, uh, sorry I lost my train of thought. But there, there's certainly, um, there's certainly, uh, values that, uh, in those, sorry, in those gaps, you can, you can start taking on potentially more projects or you can potentially extend those design stages, which is not necessarily always where a client might save.

You can produce faster,

you know, if you could do produce the same amount in a shorter period of time. They might not, not necessarily be up for a longer concept design stages, which is what we might encourage. So it's, it's, there's multiple routes. There's not one clear answer. Um, but all industries will certainly be changing and we're definitely seeing it so radically, especially in pro programming.

Now, obviously, so many programmers using coding or building AI models or that's their first beachhead, that's their whole first market. And in the legal industry,

legal is such an interesting use case. 'cause it's so text driven.

Text is like the 1 0 1

text image video 3D, that's kind of the

like el like we're in 3D so we're a little safer, quote unquote.

There's clear complex for now, there's clear complexities in the AC industry. You know, supply chain 3D modeling, like there's enough complexity to protect us for let's say the short, medium term compared to other industries. Legal text. We just need a plugin into Microsoft Word, and we have automated 50 plus percent

of some arch legal, you know, per tasks.

And that's happening now.

And that's an industry which is time has value, like one, like so centrally embedded into its pricing model. So that could lead to, it's, it's being aware of what's happening out of architectural

Evan Troxel: Yeah.

George Guida: The discussion too.

Evan Troxel: I, I feel like where it typically falls or has fallen, and maybe it just continues to do this, I don't know, is Yeah, just doing more, I, it's like. Like BIM didn't make it so that we could go home earlier, right? It just made it so that we could do more. Uh, and so at least that's how we've applied it to our practice.

It's like, okay, yeah, LOD 400, let's go. Um, let's, let's figure out all the things and cram it into the time space that we u have always had for SD, DD and CDs. Right? Um, which is, it's good on one level, right? Like design could never be done as far as a designer's concerned. And we can continue to design at finer and finer detail and work out that design intent and get it into the documents.

And I kind of feel like that's how these tools are gonna be applied as well. It's, there will be pressure, of course, internally and externally to say, okay, now you actually get half the time because the tools are allowing you to produce this information a lot faster. To me, it seems like if you're actually gonna be producing better, design, more back to your thesis of, of really wanting to push the envelope of what we are capable of in our firm with our designers, then you're gonna say, we get, we still get the same amount of time, but we get to iterate it on 50 x instead.

For example, I mean, I see architecture still really falling more into that, that side of things, but still having to have that internal conversation of the pressure of doing things for the sake of efficiency and productivity. Um, and, and I and I, I feel like design needs to sit in front of that. It needs to have more importance because the value of architecture, it, it sits there.

It's like more design, more time in the project. Figuring things out is a better outcome. It always leads to a better outcome. Of course, we have to cut it off at some point, but less time has never been a good answer for producing great

George Guida: Yeah, that's that a hundred percent in line, that just becomes a, a, a fun negotiation for clients. Uh, because,

uh,

Evan Troxel: Fun.

George Guida: if you, yeah, this is something, uh, Eric sees. See l uh, from

Harvard is also described essentially that curve of value perception from a client. If you reach that 90%, imagine you reach the 90%, but you need two more weeks to reach the a hundred percent of the best value per, you know, co

per time and whatnot. Is a client going to want to invest in that extra 10, 15% maybe. We're obviously in our, our own perception we're perfectionists in architecture. It's never a finished game,

so we would like that, but it's, it will be exactly that fun conversation to express that value add.

Evan Troxel: I mean, yeah, if you look at it like a game, then it could be fun, I think, and it should be, maybe it looked like as a game, like, oh, we we're gonna figure this out. We're gonna play this game. That's a, I think that's a good approach for a lot of things, but a lot of people don't, don't think of it that way.

They're like, oh my God, that sounds, that sounds like a horrible thing to have to figure out with a client, but it's coming, right? It's literally coming. This is one of those things that these negotiations, these conversations are gonna have to happen because it's happening all around us and the pressures on to address it in this industry, and I don't necessarily mean that addressing it means adopting it wholeheartedly.

I know you're probably more on the side of the scale of like, yes, this is a good thing. I think it also makes the argument for boutique firms who say, we absolutely don't use any of this stuff. It is a hundred percent human creativity and, and I'm curious what you think about that idea in this space. Do you think there's value there?

George Guida: we, we've seen this historically

through time and uh, we're definitely, I think this the technological divide,

the made by a human or made. Non AI generated architecture, there'll be inherent value in that. I'm not opposed whatsoever with that approach. So, um, yeah, we're gonna, it's gonna be interesting.

There'll be the, the heavy adopters, but they'll, there'll be a, I feel a bigger and bigger camp over time of the non, you know, the non-AI users. And as value a

client, you know, a client who doesn't want a company to use AI

Evan Troxel: Yeah.

George Guida: to that.

Evan Troxel: Right. I I, I'm, I have a philosophical question for you, George. So, do you feel like AI is a tool or a service or, or some combination? Um, I've, I've recently been thinking about it, right? Like Grasshopper, rhino, Revit, whatever, ArchiCAD, those are tools because like you're making the decision to do this thing, but then there's like this kind of outsourcing of maybe it's, maybe it's a little bit of creativity.

Maybe it's a little bit of, um. I mean, obviously with this idea of large language models and, and seeing context and things that may be, you know, pattern recognition and things, uh, that, that, that it can do for you is like, whoa, this is, this is pretty amazing. But do you feel like that is, is doing a service to you or do you feel like it's a tool that you're using because you talk about it as augmenting and so then I, I start to think of it as a tool, but, but I'm curious what you think about that.

George Guida: I have shift my lectures, uh, from describing a AI as a tool to AI as a collaborator.

And collaborator can be seen analogous to service. And you could potentially in

Evan Troxel: when there's a subscription involved,

George Guida: Yeah.

Evan Troxel: but tools cost money too. I mean, it's just a, yeah. Yeah.

George Guida: Yeah, and this is, this fits well within this year's, let's say, development of agents. I would

say this year is very much the year of video and agents. In the next 12 months, we're gonna see a lot more develop in the 3D space specifically, uh, which is what we're obviously pushing on heavily in ex ura. So combining all of these agentic workflows 3D, how can you make it fully moti multimodal? Um, so agents, um, can be seen as a single agent. And just to kind of give context of what an agent is, an agent can be seen as a single chat GBT model or single large language model in conversation with other models. They have contextual history of past conversations. These context windows, which is what they're called of that memory size is getting bigger and bigger, bigger by the month. They're able to formulate or hire, do hierarchical reasoning to a certain extent, to plan to use tools, cps, or to use, um, this reason or get an answer, prove, validate that question, or go

back and continue that cycle. So from the single agent, it becomes a multi-agent process. And so can you as a practice build out customizable multi-agent frameworks.

I'm going way too deep into the jargon, of course, but um, and, and so then you can start moving to create custom solutions for not everything. We can't replace everything, but for like the documentation for optimizing some 3D workflows, creating drawing sheets, doing compliance optimization, checking all your drawing dataset.

There's more and more tools coming out there. You cannot obviously replace. Replace client facing, you know, soft skills, people to people, public facing. There's a whole world just in there. So I think the service idea, this collaboration of seeing, um, AI as a potential collaborator is definitely going to emerge.

And I, we are exploring it as well. So clients are coming to us and saying, could you help us across, you know, in this competition, across these stages, I have this unique scenario in which we still run the architecture practice, but we have this SaaS model.

So what it becomes not uniquely for us, and this is ycs for example, Y Combinators, let's say ask last year, it's SaaS plus service. You build your custom tools and future in the next few years. You can just vibe code your next app, your next workflow, make a custom proprietary solution, and then offer that service and you become an AI enabled practice.

So it's kind of this hybridity between empowering you as an Arctic, the traditional Arctic, but

also somehow offering AI as a service to

Evan Troxel: Yeah, it's, uh, that, it's really interesting to think about it from that, from that point of view, because you've got

the ability to make tools on the fly, right? And you're using this kind of outsourced. Body of of information. Body of knowledge, and you can just plug in whenever you want and create stuff. Like, like one of the things that we saw at the AECtech conference was Turner Construction. Do you remember that?

That presentation? It was where they basically said, okay, chat GPT for everybody, 15,000 employees. And now they've got people in HR and people in finance, vibe, coding apps to do things that they could have never, obviously they would, they don't have developers in those departments. They don't have time to do that.

But now they're just doing it on the fly and it's completely unlocked certain things that saves them a ton of time. Right. And And it also gives those people agency to solve their own problems.

George Guida: Yeah, it's, um, definitely an interesting model. My, I, it's incredible to see them doing it at that scale. And there's a, there's an obvious cost and implied with getting enterprise licenses with chat GBT at that scale

Evan Troxel: Mm-hmm.

George Guida: and that case, I, I would almost want or encourage an a EC specific solution.

I mean, you want it to be catered.

You want it to have that enterprise knowledge of rolling it out with an incredible speed across all, you know, all parts of an office. Can potentially be a, a risk in terms of, you know, not falling, not letting anything fall in the cracks or losing accountability or

Evan Troxel: Yeah.

George Guida: so it's, it has to be done, I think at a, a certain pace within industry. Yeah. But Awesome.

Evan Troxel: Short, short digression here, but I think something that I'm seeing lately, we see it with resumes, we see it with RFPs. We're probably seeing it with legal documents, and I could just see this snowballing into architectural documents. Right? Which is. Okay. The AI writes the RFP, the AI writes the response to the RFP, the AI checks, the AI's response to the RFP.

You can see where this is going. I mean, is this, is this, I, I, to me, this is something to be concerned about. Like it's, it's went, especially when it comes to like a design competition. Are we just gonna have the AI tools help really augment us and, and create the entry for the design competition? Only for another AI to review the submittal to the design competition?

Competition?

George Guida: Yeah, it's, it's, it's

Evan Troxel: We're outsourcing everything.

George Guida: I, I, I don't think it will be universal. I like to think it won't. Uh, but it's, yeah, we're gonna definitely see more of these use cases, especially from a validation perspective.

What I, what I find scary is, is seeing this in light of licensure, in light of our role, our responsible controller

or our standard of care,

because the ball falls on us as

architects. And that's today, and I'm just gonna add this as a provocation, but in the eu there's, I think it's a liability act, which. A product liability act, which essentially, which conceptually can shift an AI model's liability to be similar to that of like manufacturing. If there's a defect or hallucinated response, could that AI model assume certain level of liability

that changes our current system?

Which means could the outer desk, could the Revit license of, or all licenses. Now they give you full liability of anything you do on

these tools, which is fair.

Evan Troxel: right.

George Guida: Could that model in future, I'm not saying in the next decade, maybe slightly longer. What if that model shifts then we're really like, you know, if we're, we need to have like a, a different discussion about what licensure means and what, you know, the cyclical, the scary cyclical loop of making, verifying with AI and us having positioning ourself within it. And I will, I'll caveat just to say, and this is what I teach, how can we embed from a tooling perspective the necessary feedback loops?

Within these systems to keep our agency to keep the rights, check the right checks and balances within the system. xFi Good is all about this. This is our thesis process As project can we dissect your whole workflow,

give you full visibility in a node-based accessible system to see what our inputs are, what our outputs are, and across all the different modalities, text, image, video 3D. So it's that. It's designing, it's at the core of it's, you know, pre-use, it's design, the right systems to keep the right values as a profession. That's what I'm really trying to push on.

Evan Troxel: Okay, we'll get, we'll get to that in a second. I, I want to talk about ex uras, but the, this, this kind of snowball effect that I'm talking about. I think it's just, I to touch on it one more time in that the concern really is like convenience wins, right? That's human behavior is convenience wins. This is why DoorDash, this is.

Right. This is gig economy kind of stuff. Um, it's like, okay, what's the easiest thing? Because we have busy lives and there's even more constantly knocking at the door of things that need to be done. And so we're outsourcing the email writing, we're outsourcing the summarization. We're out, you know, it's like one person shows up to the meeting and gives a lecture and there's eight ai um, note takers that have also joined the meeting, right?

It's like everybody sends these as proxies of themselves to the meeting because they don't have time to be there. And so everything is getting to that is, is get, like, we just seeing more versions of that, right? So it's like, oh, now it's gonna, now it's the image making, now it's the 3D models. Now it's the video.

And it's just like this constant kind of orchestrating. And so, I mean, how does that fit back into, and I think this is where we segue into exFigura, right? Which is this, okay, you you call it augmentation, but it, it's like. How does this weave its way back into responsible control? Um, but also just kind of, you know, like the, the, the lame way to say it would be master builder, right?

It's like this, you, you have this orchestration, you're just constant, but all these things are doing the work for you on your behalf. And of course you need agency and you need the way to check into all that. But, but is this, is this getting, is this making things worse or is it making things better for architecture?

And I guess what it really comes down to is what do you think that, and you have a foot in practice, you have a foot in academia, you have a foot in technology, the value of architecture, like what is the value of an architect and architecture to society? Do you feel like it's gonna be more widely distributed?

Because architecture is really for the 1% of the 1%. I mean, if you really step back and look at it the most. People hire architects to do the, the, the 1% of buildings in the world, or maybe it's two or 3%, but it's not a lot of percent of total buildings in the world. So is the value get more distributed because it's more available to everybody because more projects can be done?

Or is it actually gonna water it down even further? Um, and just make, make it so that more people can do more buildings without architects? Like where's, where's the value proposition lying there for capital A architecture and, and could that become more widely distributed?

George Guida: An interesting analogy is, um, what I recently heard in, um, a talk with radiologists, there was that same perception that. You have a machine learning model,

you train a data set on non-cancerous or non-malignant or mag malignant tumors, or like these type of examples, medical imagery. So can you know the levels of accuracy, especially in machine learning, which is more deterministic than, you know, the hallucinatory side of generative ai. You know, considering that efficiency gain, the, the response, the accurate response rate, will we have less radiologists and the, at least the last few years have actually seen a rise. And there is that question, this is what you're describing here. It's a broader democratization. There's actually higher, uh, there's higher demand for these types of services.

Evan Troxel: Because there's more need there. There's more cases

George Guida: there's

Evan Troxel: there was a huge bottleneck before. And now that that bottleneck

George Guida: and the big bo part of the bottleneck was, you know, cost and. Maybe time or like,

and so now can we move? I do think there's definitely an, a really interesting conversation about that post 3%.

Evan Troxel: Mm-hmm.

George Guida: And so these tools will definitely open up a whole new category. Um, there is definitely, definitely that scare, which, which is what I call the mid journey effect of, you know, the, the lay person, the non-architect who comes to the client, which is, you know, scary, comes to the architect, which is a scary thought. Okay, build this, this

is my mid journey image or

Evan Troxel: already done all the work. I just need you to draw it

George Guida: execute, let's

go. That's, I guess

Evan Troxel: That, that is not a new, that is not a new conversation that's been happening forever. Right. Except it used to be an image outta dwell, or it used to be a sketch on grid paper or whatever. It's like, I've already, I just need you to do the drawings.

And so again, there's like a mismatch of like the value of architecture, which these tools, I think are, are putting a magnifying glass on because they're, they're, they're, they're super heating. That that idea, which is like, I really don't need you except to get the permits approved. Right. That's your value is because you have a stamp.

George Guida: So, um, coming back to your idea of Orchestrator,

the terms that are being used now are, yeah. Orchestrator curator,

let's say that, that higher, you know, controlling manager of these systems. I think that's where our value add will certainly lie in,

you know, connecting the right pieces. And it's, it's like we're, we're going really deep into like this, the potentials of ai, but I, I do want to take one moment to step back and say in a EC, the actual development of AI tools, the real, real use cases are so typology and use case dependent, even today. So we're talking about like many small segments, a lot of noise happening. This will flatten over time. But really if you think of like the, the crux of like CD or DD stages of like going from an LOD 100 to 300 or like these type of workflows, it's so, it's very niche and its adoption is still on the much, on the lower end.

The potential is huge, but we're like just the frame. Like we are still, in the earlier days, I'm different because we've entered into the market from an ideation and visualization perspective, it's much the ROI or the value add is much more apparent. Uh, but for the later stages it's still quite nascent.

Evan Troxel: Yeah, and it's, it's difficult for people to adopt a tool that they are, have no trust that will be around in six months with the landscape changing so quickly. So why even invest in this tool? Because we just don't have the foresight to say, are they gonna be here to support us when we actually need them?

Because these projects take five years.

George Guida: Yeah, absolutely. Um, so I, I do think this orchestration level of, you know, being in charge of the, the human to human, the client facing, the public facing for any public consultation, you can't send an ai, you know, nothing's really gonna happen in that case. Uh, so

there, there always needs to be a factor of, you know, that that human in the loop and I can push the needle as a provocation on the other side if you want to.

Evan Troxel: Yeah. Let's hear it.

George Guida: So for the Venice Biennale this year, we, um, we submitted with a team, um, the, a project essentially, which essentially conceptually we tried to close the feedback loop of projects. Post completion. So, and you think of the architectural cycle, you do a project and then you, you finish it and that's it.

I've never

Evan Troxel: to the next project,

George Guida: you go to the next project, I have never seen myself post occupancy evaluation details or like real feedback from users in that space 1, 2, 5 years down the line.

So could we conceptually close the loop and bring agents in the workflow? So

I'm going into the weeds, but what it means is, from a 3D perspective, we're using Unity. Can we have a building before, imagine an A parking lot. We scanned a parking lot in Boston. We converted it through AI workflows, video workflows, 3D Sian splat workflows into new designs, more sustainable facades into residential buildings. And we add, we ask physical agents, so large language models connected to values, names, professions, to each respond. Certain metrics, which we asked them.

So those could be like, is this walkable,

is this safe to you? Uh, what is your perception of this compared to the previous design? Some metrics which we previously can establish. So you've got 200 agents walking around your new design assessing based on your criteria, so you can really inform that full cycle

and then keep control and keep iterating as a loop. I was just attend it as a provocation, as a way of potentially re-seeing what simulation in its traditional sense could mean versus

Evan Troxel: So you're looking way farther out, uh, through like, okay, you're, you're saying, okay, this is the design. It has been done, it's now been occupied, and then we're gonna go into the future and we're gonna have these agents review all of those criteria that you said. And then you're gonna take that information and you're gonna infuse it back into the design phase and, and continue.

So you're creating a way forward looking feedback loop into the design process.

George Guida: Yeah, it could be fun.

Evan Troxel: It could be

George Guida: it's, it's,

I mean, I think

Evan Troxel: legitimate use

George Guida: I think it's, it's, it would really help validate, and, and this is back to the value proposition of, you know, that quality, that design matters approach.

Could we start really informing, validating beyond the obvious need of optimization,

performance analysis, environmental, all of that?

Of course,

it's like the qualitative side plus quant

Evan Troxel: Is there a value add or a, I should say, a monetary, um, or a contract reason to a, a version of like, does that change the way that architects operate now and how they charge their clients? Or is that just like the evolution of current version of practice, do you think? Assuming that provocation becomes reality, let's just say.

George Guida: I like to, on the short term, this is not for this use case, but across the spectrum, on the short term, I think there will, you can certainly up charter there, there will certainly be a, you know, a universal value add,

unfortunately that the the flattening of that curve is, you know, there is potential for that to happen.

So

it might be

Evan Troxel: It just becomes normal. Yeah. Interesting. Well, okay, so let, let's segue into what you've been working on with xFigura and gi. Give us an overview of, of your platform and how you're, how you're going about solving an architects early design concept Phase issues, their problems, their, their opportunities.

George Guida: Yeah, it's, so xFi URA is a platform we launched about eight months ago. We've, we're now a team of nine about, or starting fundraising now, and it's seen a huge growth in the last few months. So been really excited to, to push into this new, new space. We are essentially, you can see us as a collaborative platform or a mirror-like interface where we combine all of the best generative AI tools into one. So essentially there's, um, now there are 52 models live with, together with an AI agent, which help accelerate or help improve that design quality from a concept design stage. The idea is, um, is simple, super accessible interface. You can create presentation boards, you can collaborate, you have full data security, which is essential from an enterprise level, and you can bring all teams into one space to design those concepts and actually not be sucked into the reality of the now. The noise is real. We are the noise catcher and we're the few, you know, some of the, one of the few companies that can approach a client approach a large, medium, or small practice. And. If you've been testing some AI workflows, dunno where to start, or you've been sucked into this noise, we will remove tools from your plate and we've packaged them all into a single accessible interface for a EC specific workflows.

Evan Troxel: So, so by taking, okay, I go to Chachi Petit for this. I go to, you know, notion for that. I go to, you know, whatever. You've got all these different tools out there, and they're all kind of, you have to go to each one of them to use them. What you're saying is you're pulling all those into one canvas. So that all of the work is happening in, in one place, so that, that simplification is that I go to one place to access all of these things and apply it to conceptual design.

George Guida: Yeah. And um, uh, example of that is using our rhino plugin or spec or speckle integration and you can import an image or a 3D model

from this and you can do your whole workflow. So it's. Early massing model. Let's do an image to image workflow. Let's use the latest nano Banana Pro

Great marketing term.

It's the, it's the fun

hype of the moment. Very catchy, but the image quality has changed visualization forever. The level of coherence that we have now is unprecedented to what we had four or five months ago, and so we're able to take that massing model, ideate across options, and then potentially make facade modules in 3D or export 3D models or iterate or go back to an aesthetic style, or there's so many different use cases and giving users that full canvas and control over the whole process is such a value add.

Evan Troxel: I, I would love you to give a, your best shot at explaining what the interface looks like and feels like to use. I, of course, want people just to go over to the website and just see it, because the image definitely will speak for itself when it comes to it. But when you describe kind of what the xFigura interface looks and feels like, how do you explain that so that people can get an understanding of it?

George Guida: Yeah, so you start off by creating a board. Uh, this, this can be shareable to clients or to colleagues internally or to other students. We've got a big academic community using this as well. And once you enter into that project, essentially it's a canvas, it's an infinite diGuidal space, like an infinite word document, but catered for any different modality, text, image, video 3D essentially to, it all boils down to a single node. A single node can be seen as a whole platform in itself. Actually more than a platform, because that single node where you can put a prompt in generate text or image or the other modalities, it gives you access to all the other models. The example is for image, you can use open ai, you can use the Google models, the open source flux or SD L models. And so within each node you can then generate images, import images, and then create a full workflow. So it's a, you can see it as a very accessible form of Grasshopper.

You're connecting.

Evan Troxel: I, I was, I felt like you waited too long to get to that point because the, the, the, it's like grasshopper or Dynamo. It is, it is a canvas in which you drop these nodes onto and then you can string them together into different configurations to create basically a recipe right.

Of outputs. And those can fork and they can do all kinds of different things. But, but okay. So, so when you say an infinite canvas like Miro, like an AutoCAD drawing, whatever, it's just this big blank space and then you drop or imp, you know, you create these nodes and then you start to string them together to create workflows basically.

George Guida: Yeah,

and you can create templates. So automations, you can publish these, share these, and or for onboarding you'll have a full set of them and you can, with one or two clicks, you can have a full upscaling in painting 3D pipeline, that's just,

you know, one click away.

Evan Troxel: nice. So you said you have access to 52 different models, and those are large language models. They're also image models, they're video models. Uh, and, and so like you said, this is multimodal, so you could, for example, I mean, could you give a couple examples? I don't actually wanna make any examples of, of, of maybe some simple workflows that you have your students start out with.

For example, to just start to explore how to use a platform like this.

George Guida: one of the hello world, the the first exercise that I do with students is giving them an understanding of how these models think. And so at a core of using all of these models is, is understanding what prompting means, knowing how to communicate a design intent or a general intent of whatever result you want to these models. Now there you have to understand that each model. Is fully subjective. There's no such thing as an objective data set. And you're essentially, you have to prompt each model in a slightly different way. And so learning how to, the exercise I give is describe a a, an image describe, so bring in an image of a, a building that you're interested in.

It could be like a Louis Kahn or something. Drag it in, describe that manually. Okay. Then let's use chat GBT to use the, an AI large language model to describe that image.

Evan Troxel: Yeah.

George Guida: And then let's, like, let's find different ways in which a machine versus human will will describe in a single image. And then let's generate an image from that and compare that before and the after.

Evan Troxel: Mm.

George Guida: we go into the nuance. Okay. Language matters

that we want to give information about. Materiality form lighting. All of these things which are maybe implicit when you're designing in a different way, have to really be pulled out. And that was partly a thesis of mine for many years back at Harvard. It was repositioning the role of language within the design process.

Evan Troxel: Mm-hmm. Do you find that, uh, having an architectural training. Helps in that process or hinders in that process, because I, I think about it from like a, oh, the, like the readability of architecture. You know, it's, it's how, how do normal people, how do muggles non architects, how do they read architecture is gonna be very different.

And, and, and their description of it would be very different than an architect's reading and description of architecture. And so, I'm, I'm curious if one has a more useful, um, part in, in this process or not.

George Guida: yeah. Uh, disciplinary knowledge is still essential, and that's where I kind of. Tell my students as well. And knowing how to describe architecture, knowing how to describe form, that's text. Knowing how to convey an idea also from a physical model. The physical model or craft

craftsmanship for me is not dead in light of ai. For me, it actually gains a whole new value. So could you do a physical model, iterate, test out different forms or sketch? We're seeing so much more use of sketching, you know, go from text, sketch or physical model. I have not even touched a computer in this like, you know, first category. And then from that derived form, derived ideas connected those to different pipelines. So obviously, you know, architectural school, those first years, essential in getting that base, you know, that core knowledge.

Evan Troxel: Uh, you know, this idea of, uh, okay, so, so I used to show this video in some of my presentations. Uh, there's a great YouTube video if I can find it. I'll put a link to it in the show notes of, uh, these two parents, uh, handing their two kids a legal pad with a phone number on it. And then they had a, uh, cardboard box on a table between them and they said, okay, your job is, you have three minutes, or whatever it was.

Maybe it was five minutes to figure out how to dial this phone number. And they told, took off the cardboard box and there was a rotary phone underneath it and hilarity ensued. It, it, it is really funny to watch because these kids have never even seen a rotary phone, let alone like, know how to work it or, you know, they, they barely know what a phone number is because the contact is in their phone.

You just tap one button and it does all the magic. You don't even even need to know how to, so, uh, you know, it's a, it's a few minutes long and, and, and it's, I won't spoil the ending. Uh, so I'll put a link to it in the show notes, but like, is a keyboard the next version of that? Right. Where it's like we don't, because, because as you're explaining what you're talking about with sketching in particular, it makes me think back to a few years ago when I think it was HP.

Had a computer with a, like an overhead camera. And it was really meant for creative people. And it was like watching what you were doing with your hands. And it wasn't about the keyboard, it was about sketching or pulling in, um, you know, arts and crafts and, you know, making collages. And then it was capturing that in real time.

And then it, it could do things with that. It, it would, and it, this was before ai, right? So it wasn't like really understanding what it was looking at yet. But I could see now with AI really being like, what, okay, it actually is watching me sketch this thing and it is building as I'm sketching. Um, and, and what, who needs a keyboard anymore?

Like, that's the next version of the rotary phone to me because of voice. This just natural language aspect, but also the ability to, with this multimodality and, and you could see that happening in real time being kind of a really incredible tool for architects to be able to use

George Guida: A a a clear example of where interface could, you know, at least the tension interface as a question is going towards was open AI's recent acquisition of John Johnny I's Company. And that merger really is, I think, is going to start pushing

Evan Troxel: and AI open. OpenAI just basically said like app hardware makers like Apple are our competition, not other AI companies. Right. It's interesting.

George Guida: there. So we've been, we've been saying this idea in the next five years and the next five years, VR or spatial computing will, you know, see it's new, see a new,

wave, It'll

Evan Troxel: a new version.

George Guida: But

I, but I feel as if we are in, in a potential, like in the, the hot five years now,

and I recently, uh, saw my students do engage in vr.

We

go into spatial, you know, physical computing and whatnot. So I think there's, and this is what I teach, the, the triptych I teach is architecture interface entrepreneurship. And so I teach 'em how to build exactly what the future of interface might actually look like. And so it's fascinating as a topic.

You say the keyboard, but there's obviously the next step is voice.

And so voice. Okay. What's the limitation as a representational type? In language, in voice. And then the next step, okay, let's extend to spatial computing. Okay. What does that mean? Like I'm thinking vr, gravity, sketch of physical interface.

Are we standing to design? What is that like? What is that interface? Are

we doing it collectively? Should we use the World Lab models or Genie three? These, like what is the future there? And conceptually, just to push the conversation, I have a lot of fun doing this generally is, um, really let's think of what the future could be.

Uh, one future could be and that to, to express this. I'll use met as recent research, which is okay now we can decode thought as text. So text to brain

thought or I'm thinking something I can express it as text.

Okay. Step two. And there, there's already published research on this thought to image, can we decode, thought and express it as a video or as a single static image?

So then Conceptualists, think of that in a EC thought to image to 3D model, to construction documentation. Could you imagine, could you dream a building hypothetically, conceptually provocate in a provocative way? What does that mean within, in light of what interface could mean? It's, there's there's a lot which could come in the next, you know, several

Evan Troxel: Yeah, I think about that as kind of designing from the inside out, which is like through, you start with the experience, right? And it's, and and you are explaining or imagining as you in physical space are walking through something and you're pointing at things and you're like, okay, and here this is where this and this, and I go through this tight space and then it opens.

And you could see that how that could lead to, I mean, based on your description that you just gave, I could see how that could lead to the creation of spatial. Elements, right Bounds or you know, transparent levels of transparency or whatever, that would then translate into models that translates into drawings, that translates into the physical built environment could be an extremely powerful way to get ideas built.

Right? I mean, to, to have that translation be a much less manual process to make it happen.

George Guida: Yeah. And the, the, the debate in this context is about the potential loss during those moments of translation.

There's a representational shift, and so when going from a thought, our imagination, we have to put that into text.

There's a loss and there's a

Evan Troxel: Totally language fail is, is, I mean, it fails us in describing things so often. And then there's the layers of interpretation of the person receiving that on the other end, which also leads to potentially a lot more laws. Right. And so language is not a perfect medium to, to do the things that we're talking about for sure.

And so, like you're saying, if there's a more direct connection from thought to output, that could, that could be really incredible.

George Guida: Let's build it out.

Evan Troxel: I like the, the pregnant pause there. It was like, yep. Uh, this has been a fun conversation, George. I mean, is there anything else that you wanna, you, you wanted to talk about today that, that we could, uh, extend this conversation even further? Or do you feel like we've exhausted this, this round? Where are we?

George Guida: Yeah, I'm, um, I think we've covered good ground and it's been a great conversation. If, if anyone is of course, interested in Una, uh, happy to share more information and we're only getting warmed up here, so, uh, looking forward to bringing you guys along this, this journey with us.

Evan Troxel: This has, this has really been great for me too. I feel like some of the provocations that you brought, I mean, I like how you're looking at it from multiple angles and not just really driving from a, um, I mean, this is actually one of my criticisms of, you know, you've mentioned the report on AI that the a i a released this year.

That was, I think just really, um, benign. Like, it, it was just like, and here's the state of the, you know, most architects are just watching from the sidelines. It wasn't, it wasn't like, here's where we're going, here's what we need to do about it. It just didn't really hit me like that. Um, and so I feel like because you're have this kind of, uh, this, this.

Perspective of, okay, here, here's one way to look at it. Here's a completely different way to look at it. I like the balanced nature that you're bringing to this. I'm glad to hear that you're on that team that is gonna help this industry kind of focus on, on where we're going because, um, it's hard to focus with, with the extreme, um, fire hoses of information that are constantly blasting us with all the changes and the game changers and the, this changes things forever and this new model and constantly being bombarded with that, it's like, it's uh, it's, it's, it's overwhelming but it's also just kind of like, oh my God, if it's just gonna change every three days, I might as well not even pay attention.

So, I mean, that's the reality of, of practice practicing architects is they've got deadlines and it's hard to keep up. And, and so I appreciate that there is some guiding force that is. Happening and that you are involved on both sides of the Atlantic as well, so that you can, uh, bring those perspectives to it as well.

And, and note the differences in the way that people are operating, but also try to bring all that together into, um, a cohesive thought. And you've got academia and you've got practice and you're, you've, you're doing it all. I hope you don't burn out too fast.

George Guida: I appreciate it.

I'm

Evan Troxel: the energy you're bringing.

George Guida: excited to be with you and yeah, thanks for the conversation.

Evan Troxel: Absolutely. Alright, and I'll have links to everything that George was talking about in the show notes for this episode. Thanks, George. It's been fun.

George Guida: Likewise. Bye everyone.