124: ‘Interoperability is a Trap’, with Dimitrie Stefanescu and Matteo Cominetti

A conversation with Dimitrie Stefanescu and Matteo Cominetti of Speckle.

124: ‘Interoperability is a Trap’, with Dimitrie Stefanescu and Matteo Cominetti

Dimitrie Stefanescu and Matteo Cominetti of Speckle join the podcast to talk about Speckle: an open-source data platform for the architecture, engineering, and construction (AEC) industry, which aims to make it easier for people to connect their tools and share data. We also discuss how the platform's community-built connectors enable users to connect to Rhino, Revit, AutoCAD Civil 3D, Excel, and many more applications via the cloud.

The conversation also covers how Dimitrie, Matteo and the team are exploring the integration of machine learning models and automation workflows into the platform, their belief that open-source infrastructure is necessary for equitable distribution of value within the AEC industry, and how the industry should compete on the designs produced, not on the tools used to produce them.

Connect with Evan

Watch this on YouTube:

124: ‘Interoperability is a Trap’, with Dimitrie Stefanescu and Matteo Cominetti
Dimitrie Stefanescu and Matteo Cominetti of Speckle join the podcast to talk about Speckle: an open-source data platform for the architecture, engineering, a…

Episode Transcript

124: ‘Interoperability is a Trap’, with Dimitrie Stefanescu and Matteo Cominetti


Evan Troxel: [00:00:00] Welcome to the Troxell podcast. I'm Evan Troxel. This is the podcast where I have a conversation with guests from the architectural community and beyond to talk about the co-evolution of architecture and technology. In this episode, I welcome Dimitri Stefanescu and Matteo Cominetti of Speckle. Dimitrie is an architect and programmer by trade. And co-founded speckle as an open source project in late 2015 while a research fellow at UCL.

Matteo is Speckle's co-founder and he was previously involved in driving automation and digital design implementation efforts at Arup, Foster and Partners, WeWork and Case. Together they started the company that is now accelerating Speckle's mission to transform the AEC industry.

In this episode, we discuss Speckle, an open source data platform for the architecture, engineering, and construction industry, which aims to make it easier for people to connect their tools [00:01:00] and share data. The platform's community-built connectors enable users to connect to Rhino, Revit, AutoCAD, Civil3d, and Excel, and many more applications via the cloud. We also talk about how Dimitrie, Matteo and the team are also exploring the integration of machine learning models and automation workflows into the platform. Their belief that open source infrastructure is necessary for equitable distribution of value within the AEC industry, and that the industry should compete on the designs produced, not on the tools used to produce them.

So, without further ado, I bring you my conversation with Dimitrie Stefanescu and Matteo Cominetti.

Matteo and Dimitri,e welcome to the TRXL podcast. It's great to have you here today.

Dimitrie Stefanescu: Same. Thanks for having us.

Matteo Cominetti: Thanks, Evan. It's a pleasure to be here.

Evan Troxel: Yeah, [00:02:00] this is gonna be fun. The Speckle team, the co-founders are here, and how many years has interoperability been the buzzword for all things AEC/tech?

Dimitrie Stefanescu: Quite a



Evan Troxel: say quite a few. I remember that being a specific keyword that many people on my team, in the digital practice at our firm were searching for at Autodesk University, because this was. The problem that we were all trying to solve was how to get files and data between apps for so long. And I think, uh, it was Scott Davidson from the McNeel team who came onto the show and he coined the, the term like, file formats are terrible, right? Like file files. In general, we've definitely seen a huge move away from files and instead. It's all APIs. It's all about streaming data back and forth, and you guys are right at the center of that. So let's [00:03:00] go back and talk about how Speckle started and why this is something that you've committed yourselves to, uh, taking on as the, the challenge.

Dimitrie Stefanescu: Mm. Okay. Yeah. So there's gonna be an, uh, let's say an answer to that question that we cannot give fully. Like there's a, there's a little backstory too, you know, a bit of cloak and dagger and mystery.

But the official version, uh, goes like this. Um, Essentially, I was in UCL, a Marie Curie fellow after a stint in aec, like being a normal architect, and I was tasked with like looking at why is the, the industry so lagging so much behind essentially many other industries who are profiting from say the quote unquote digital revolution.

And then obviously interoperability came, came up as you said, it's, it's an obvious, uh, place to look at. And with speckle kind of, it was an [00:04:00] in, when I started it off as a research project, basically it was literally started on the assumption that actually maybe it's not necessarily interoperability, that's the problem.

It's actually pe getting people to talk with data. Right. Interoperability is the process through which they do that, of course. But Speckle tried to provide essentially an alternative answer to essentially making files work together, like being able to import the proper Rhino file into Revit, um, or going through and endless DBS I c exports.

And, uh, I was super lucky to find the best co-founder ever in my first contributor to speckle,

essentially Matteo. And I'll

hand over to his side of the story now.

Matteo Cominetti: yeah. Was lucky to come across Becca, I think in the first place. I was actually at the time working on something very similar, but as, uh, many EC firms, it was quite a corporate environment, so we were not allowed to share any, the tools we were developing. Um, [00:05:00] even, you know, having a very nice open source project out there that we could have contributed to the, I mean, I wasn't allowed, so I decided to do that on my own, on my free time. And I met with imagery at a pub, you know, and, you know, that was a bit the start of spec, uh, you know, as we know it. I think, I think the, the first step was a dynamo connector. Is that right? Imagery. and then from there

Dimitrie Stefanescu: Yes, indeed.

Matteo Cominetti: continued adding pieces to it. but I think, you know, you touched on something important, which is the interoperability bit, and I think we're trying to move away from that concept. Uh, I think interoperability is a bit of a trap. It's a word with, you know, lots of meaning. Very, not that we're trying to change it to be more connectivity, and I think that's what we're trying to enable at the moment, as.

Evan Troxel: the cool thing about what you guys have done is you've con, you're basically connecting everything and you, you talked about connecting, right? Being the basis of the, the shift in thinking.

I think that has a, words matter, [00:06:00] the words that people use matter. And so this idea of connection is so important because there are so many little silos. And there's silos for, for good reasons, right? There's, there's. Design applications, there's documentation applications, there's right, like Excel, there's there, you know, there's, there's things where you need spreadsheets.

You need these tools for what they're really good at, and you want to use the best pieces of all of those tools to do the things that they're the best at. And until the day comes, when there's the one ring to rule them all, which hopefully I, I would hate to put all my eggs in one basket. Like really the idea is, Using the best tool for the job.

Right? We've all heard that, that cliche, that saying over the years. Right? And so that's really what you guys

are enabling. Give us an idea of the timeline here of when Speckle kind of started so that we, and then let's start to kind of piece together the. The timeline of what you've tackled along the way.

Because to me, I mean this is, this is an old problem in [00:07:00] aec, but it's not a really old problem. I mean it, I guess it's always been an old problem, but a lot of people have tried to tackle this and I think they get really specific about connecting this one thing to that one thing. And you guys are trying to connect all the

Dimitrie Stefanescu: Almost all the things. Yes. Timeline-wise, I think, yeah, there's, there's, there's a limit to our ambition, of course. Um, timeline-wise, I think the first line of code that had speckle in it as a, as a word, was written sometime maybe late 2015. It was a bit of a different beast then, and it went through several iterations and revolutions and reinventions, uh, across the way.

Uh, one funny. Part of that, of this early timeline of speckle was that flux, uh, came out of stealth. Just as, let's say the incipient version of the speckle we know now was, was being born and Flux came out, and I think,

Evan Troxel: Mm.

Dimitrie Stefanescu: I don't remember. I

looked at Matteo. We [00:08:00] said

like, should we shut this down? Should we just like jump on flux and, you know, Actually there was the decision, okay, no, let's keep on doing this because it's, uh, it's worthwhile, you know, uh, flux doesn't do what we're doing.

It's very close conceptually, but, um, There's also the whole kind of different political agenda that we had by being open source that we want to keep alive, right? Then um, I think in 2020, Mattel and myself, we were kicking each other in, uh, in the proverbial place and we decided to, you know, go out and see if there's a way to kind of move faster.

Like, cuz we. Possibly we're getting a bit bored to try and solve this problem from inside the industry, from within an AC company or another.

Evan Troxel: Hmm.

Dimitrie Stefanescu: And that's when we first raised, and

Matteo Cominetti: Yeah, at the time we


Dimitrie Stefanescu: to say. That's when the company was born.

Matteo Cominetti: At the time, we, we were quite naive. We

had no clue [00:09:00] what, uh, raising funds, uh, was what I got to market strategy is, you know, um, uh, you know,

What that pitch deck was. So it was a bit, you know, almost, uh, unintentional that we started raising, spoke with investors, you know, uh, we found some amazing partners along the way that, you know, understood open source and they gave us the convention that, you know, VC money is not bad per se.

You know, you just need to find the right partners and be aligned. Mission. And, you know, they gave us the right, you know, resources to,

to start Speckle as company.

Evan Troxel: It's, it's interesting that you brought flux back up. I haven't heard that in a long time and it's like long forgotten past of connecting some apps in a e C I'm sure it's near and dear still to some people. I, you do hear come up in different presentations of kind of going back and

talking about the history. How did you guys feel when that went away? I mean, you guys obviously outlasted them. You've played the game of [00:10:00] Survivor better than they did. I mean, like you said, they have different motives for sure. Um, and I'm sure that had a lot to do with it of what they did or did not find as a response from the a EEC industry. Kind of talk through, through that part of it, because. Did you feel like this weight lifted off your shoulders, or was it more like then you doubled down? Or how, how did that whole thing play out on in

Dimitrie Stefanescu: I think we were not very well positioned at the time to double down and to accelerate. Um, my first reaction was, thank God I didn't stop working on speckle, because otherwise my PhD would've been. Dropped out. Like I wouldn't have essentially access to the data that I needed to do my research. Um, second reaction was, oh look, there's a bit of a panic inside the AC world, inside the cool kids club of computational designers, so to say.

And um, that's when we tried doubling down. But actually we didn't have the resources to scale [00:11:00] up development the extent that. We would've liked. it was also a bit of a disappointment. And when we started the company we. Looked quite deeply or as, as, as best as we could, obviously into what went wrong with Flux.

What did they do? How did people react to their pricing strategy? How did people react to them not being open source or not offering kind of control over data residency and so on. but at the same time, flux was amazing for us because it literally paved the way in terms of mind share across aec, right.

People expected a tool like Flux to exist. Now, uh, they were not content anymore with swapping files and, you know, here's, here's a Google Drive link. Here's an attachment, here's my PDF attached to this email, et cetera.

Evan Troxel: Yeah. The, the whole idea of what Flux opened up in the Mindshare, like you said, it, it did set like a new expectation in the industry, or [00:12:00] at least gave people an idea of what's possible and that then drove expectations for that kind of data trading that ha would happen in the, I I want to get back to what you said about. Kind of realigning people to be able to talk to data. Can you talk through that and how that was your approach rather than interoperability? Because it, that gives another mindset shift that I think is in the same vein of thinking where, where there's. You have to retrain people's brains to think about this in a different way. And if that is kind of a guiding principle for you, how have you, how do you see that approach, that problem, and how have you tried to address

Dimitrie Stefanescu: Good question. Um, it's, it's funny, interoperability ultimately is a technical problem, right? And. Kind of what we're trying to solve for is communication. So when people talk to each other, they build shared context. When different disciplines, [00:13:00] different silos interact, they create trust, they build a shared context that they can operate on and, you know, collaborate efficiently to deliver amazing architectural and infrastructural projects.

Um, now when the industry kind of. Forced architects and engineers to focus just on interoperability, which is purely a technical problem, like that degraded relationships in between the silos on a, on a human level, on a social level. And then by thinking about, you know, exchanging data as a communicative process right through which we want to enable these different stakeholders to have a better working relationship together.

We're actually bringing them back to what. It should have all been from, from the start. In a sense, it's not, it's not about how do I process your file? It's about what information do I get from your file, from your, from your design, right? So it is funny, it is [00:14:00] a mind shift that I think we're still, we're still pushing the industry to move in that direction

Evan Troxel: those are the most difficult things to do. I think, the one way to say it is behavioral change,

right? Like there's, but, but I don't think people like to hear it that way. People don't like to change their behavior. They think I'm fine. What do you mean? I need to sh change my behavior? But it is, It's deeply rooted and those are the hardest things to change in the industry. and so I mean, kudos to you guys for taking on a shift. and, and helping people make, start to make that happen because that's, it is, those are the hardest problems I think in the industry to try to tackle.

Dimitrie Stefanescu: Yes, agreed. It's also, it's not a difficult shift to make once. Let's say people get converted, actually, Hey, actually, fruitful collaboration can exist. We don't need to stew each other over every information exchange that we operate, right? So it, it becomes, you know, the whole atmosphere, defenses, and [00:15:00] it's, it's, it's much better.

And you realize, okay, there's actually joy in

working on a BIM model together. We don't have to be at each other's throats about who made the last change and so on. We can just go check and see why and how.

Evan Troxel: It's interesting to think about. I don't think I've ever told this story on the podcast, but I, I used to have a training website for digital tools and it, it, this was a long time ago, so it really focused around like Form

Z was my main modeling tool and people listening to this podcast don't even know what that is anymore.

Right. So

Dimitrie Stefanescu: I used it in uni. Yeah.

Evan Troxel: but, um, I had what I called a, a, uh, file interoperability cheat sheet. I had this cheat sheet and it was a PDF based, you know, and, and it was like, if you're going from this app to that app, here are the file formats you would use for different situations, and it would kind of talk you through all of the settings.

And I would update and republish this pdf. All the [00:16:00] time to make it more and more current. Right. And so how ingrained is that file format? Like if you're going from Form Z to lumion, you use, I don't remember what it was. It was probably like step or some weird format. Right. Or if you're going from Form Z to Rhino, you would use stepper, you know?

And if you're going from. Any, you

know, to 3DS Max, you might

use the

light wave format or whate. You know, it's, it was crazy to think. And then in each one of those export dialogue boxes, you had to check a different set

of things, right? If you're going out OBJ file format, you've gotta flip this one direct, you've gotta flip one of the coordinate systems. And can you do it with materials or, or do you do it without materials?

Like remember like how crazy it was And now we're talking about something. I mean, it's, it's the same but different, right? We're talking about sharing data and information that is kind of the, [00:17:00] you almost wanna say the metadata behind the things that describe the geometry, that describe the attributes and the parameters and publish

those, or stream those as it may be. Um, and I, I think about it kind of like, the music industry, right? We used to. Share songs we used to share. Some people, you know, everybody was on Napster back in the day and downloading these files and, and now there's streaming service is out there and and the information just gets pushed out to you

as you on demand, right?

As the user neat wants it. Here it comes and, and to me that is kind of a really interesting shift with how we're dealing with data in our design applications. And this streaming happens from various players, various teammates. On the design team to get us the most current data. And, and I think that another shift that we've seen is, remember in Revit we always used to

publish our files [00:18:00] like once a week, right? Friday

Dimitrie Stefanescu: to Central

Evan Troxel: the arch, the newest architectural model. Sync to Central and, and share it on the

FTP site for a long time it was like that, right? And then all of the consultants would download the model and, and and look at what the changes were or try to find the changes if they weren't using any kind of change tracking system.

And, and you can just kind of, it's interesting to go back through the timeline and see how much things actually have changed in this, but also kind of what we take for granted nowadays. Once. The light is seen like

Dmitri, what you were just saying, right? Like it's not a hard behavioral shift. And once people see it, they actually see the value in it.

But actually getting them to pay attention that there is a better way, I think is the hard part. Right? it's a, it's very interesting to see, but now we're streaming data back and forth on demand and that I think is a, it's a sign of the times, right? We're seeing this in many adjacent or other, you know, things that are going on in our lives, but we don't necessarily [00:19:00] think of it the same


Matteo Cominetti: Yeah, I, I just wanted to say that I think, you know, the, the streaming of data, you know, uh, that's just the start for us. You know, what we've noticed is, you know, as soon as this data is. It's actually accessible in a central place. It's not just going from A to B. There's so much more that, that you can unlock, right?

So, uh, proper collaboration and communication, you know, are instantly available, you know, to get information about your model. You don't have to open it up, you know, use, uh, complex APIs to export quantities or information about it. You cannot use modern technologies, APIs. SDK is to get information about your model. You can actually parse that information in tools like Power bi, that's really being able to, to access the information and get it out when you want it, and how you want it, where you want it. And this is what we're trying to, to, to enable users.

Evan Troxel: Uh,

great point. I think, uh, it would be, it would be great to actually kind of set the. [00:20:00] Set the actual idea behind what Speckle is doing right now and, and because maybe even people out there like me think about speckle doing one thing, and that is sending data from this to that. But like you just said, there's like this central repository and you can do so many other things.

So lay the groundwork for the audience and tell them exactly what your vision is for what Speckle



Matteo Cominetti: Well, I'll

start with our mission.

Dimitrie Stefanescu: Yeah, that we have a very ambitious vision. Here. Go mate. Sorry.

Matteo Cominetti: Yeah, sorry. There's some lag again. I, I would start with our mission and you know, it, it's to enable

everyone to, to access, consume, uh, 3D information. We want to democratize three-dimensional information. This means making it easy to access, consume parts, uh, understand and communicate in general speaking, 3D is a bit our motto at the moment, and we are doing it in, let's say, across three different, [00:21:00] uh, Pillars.

One is interoperability, or as I said before, connectivity. We prefer calling it collaboration. Uh, this happens online once you have all the data in the central repository. At that phase, at the object level, people can query and actual access informations about every individual object. Uh, you're enabling, you know, collaboration across the different stakeholders, across the different designers, across the different people involved on a project. But you can also then enable, uh, automation and workflows that automatically, you know, Do something on top of your data. And here we're taking inspiration on, uh, tools that developers already used to, like CI and cd, being able to, you know, create dashboards, being able to check the quality of your building and being able to, you know, just run constant logic on what you've done. Uh, this is the third, let's say, main direction, which we're pushing, respectful, and

I'm sure Dim has something

to add

Dimitrie Stefanescu: [00:22:00] Not really. I think, uh, globally, why we, why our vision is ambitious is because we're trying to kind of. Own the data layer within architecture, engineering, and construction. And why we're trying to do that is because we see data being captured by private clouds offered as a subscription. And these are designs that you make and you produce.

This is your intellectual property, and currently there's no alternative open source solution that would give you a way out. From example, from an endless lifetime subscription to Autodesk or to Trimble, or to Bentley and Speckle is positioned, and this is kind of what drives us there. Also, as developers, we believe in creating value across the board rather than just for a few stakeholders.

So that's the political, let's say, addition to

Evan Troxel: It. One thing that you said there that I don't want [00:23:00] anyone to get hung up on, so I'm gonna give you a chance to just change your words a little bit. You said own and I don't. And I know you don't mean that you as speckle want to own the data. Like you just

talked about how there is intellectual property being ex exchanged.

Yeah. Your users. And so you're, you're creating and enabling the collaboration to happen, but that data you, you're not interested

in owning that data.

Dimitrie Stefanescu: Absolutely not. No. It is your property, and essentially you can control where it exists and how it exists, and for how long

Evan Troxel: Great. So how are you doing that? And you talk about this kind of, I don't know if you want to call it a common data environment, but a common location where the data, like you're saying that the data's kind of sitting in the middle and then all of these things plug in and plug out of, of that. Repository. And so talk about how you're tackling this problem. I mean, I know you guys have so many pieces [00:24:00] to this puzzle going on, so talk through exactly what people like, set the expectation of what it's like to use speckle, and then maybe we can talk about some use cases after that.

Dimitrie Stefanescu: Okay. Absolutely. Out of the box to pick up speckle, it's very easy. You, we offer a speckle server that is hosted and maintained by us. You can subscribe and that's it. You're, you're off to the races. Now, how we resolve that problem of data, residency ownership and so on, is that, Each company or each team, or each project, each consortium can deploy their own speckled servers without interacting with us in any way.

So if you have your own requirements that are a bit more special for large infrastructure projects, for example, you can deploy your own speckle server without going through through us at all. We can deploy it and. Whichever cloud provider you want or in whatever, on whatever infrastructure you want, we've taken great care to make [00:25:00] sure Speckle is very agnostic that is

Evan Troxel: Just real quick to to jump in right there, there are a lot of projects out there that have this kind of privacy

requirement, right? There's GSA projects, things, yeah. Where they, it is absolutely not allowed to go into, you know, be hosted on some, on maybe, Whatever Autodesk Cloud is called today. Um, but also, you know any other one either, right?

And so they have to stay on premises. So you're saying you can actually spool up your own instance of a speckle server running on premises that then your various teammates and collaborators and consultants can all publish

to and pull from.

Dimitrie Stefanescu: Absolutely. and

we've seen, uh, speckle used in so-called air gap rooms where essentially people leave their phones outside,

but within, they've got the infrastructure required to essentially design model, and that includes a speckle server. So,

Evan Troxel: Great. [00:26:00] So let's talk about some of the connections that happen here. So, uh, I know that you have this, I, I've, I've downloaded and installed the, the speckle manager on my. Mac and I, I loved, like, that was a cool, cool thing to see happen, that you're not leaving us Mac users out because I do use Rhino on my Mac.

Right. And I do need to talk to other people, so. Cool. Thank you so much. But like, give people, and I, I know you're, you're pretty much, uh, agnostic to hardware and, and obviously, uh, application subscriptions and things like that. Like you said, one of your goals is to allow people to access the data without having potentially a subscription to. Any particular application or a license of, doesn't have to be a subscription,

right? It could be a whatever. So the idea is though, that you, people can see the information in 3D geometric form just through the web browser as well, right? So maybe paint a picture of what that is, what, what's

going on there.

Dimitrie Stefanescu: Yeah, usually the first time someone picks up [00:27:00] speckle, they open their favorite connector, be it Rhino Mac, which is actually the, no, not the most polished one, so apologies of it. Um, they

they the

Evan Troxel: I, I also


that. We'll


Dimitrie Stefanescu: I, yes, I I know you did. I, I felt,

you know, we had to come clean. Um, so yeah, they pick up their favorite connectors, say the rev connector, and they just select, let's say a specific view or a specific level or even the entire project.

They click send, they wait for a couple of, of, uh, of seconds, and then ta-da, they can actually go. To the speckled server and see online in the, in their browser, they can see their 3D model with all the metadata attached to it, right? So each door, each wall, each component has whatever information attached to it that we can extract using the APIs that we use.

Um, what follows from there from that moment of excitement is that that link gets copy pasted in Microsoft Teams . So here,[00:28:00]

This is

Evan Troxel: it

goes to everyone all

Dimitrie Stefanescu: and, and it goes to everyone all at once. And then here, like depending on the, on the, let's say access settings on that specific project. They get to see it or they don't like always the project owner can choose who they share with, uh, and, and how much access that they have.

beyond that, the next step is that people usually take is like, um, Hey Joe, why don't you try receiving this inside Rhino so you can coordinate your complex facade design with my latest arch BIM architectural model. And then Joe picks up. The speckle connector hits receive and then da da, he gets a representation of what came from the Revit model, for example,

that then he can start aligning his either grasshopper scripts or facade designs too.

And it doesn't stop there in large enterprises, actually, uh, where we are actually seeing the developers figuring out, oh [00:29:00] look, I've got all this data. Like, I've got all the fire ratings on all these walls and doors. I can finally run those automation, those checks that I'm kind of running manually can now run them automatically with a Python script, but that's kind of a much later stage of adoption.

I'll hand over to

Matteo Cominetti: Just to add on top of what you were saying, I think this

is what the user sees, but what is happening behind the scenes is quite, uh, interesting. We are not just, you know, first of all, we're not dealing with files, as you mentioned at the beginning of the podcast. We're working with data. We are working with the, in this case, the Rhino API or whatever the host application is, extracting the geometrical information metadata. Raw data and sending it to the speckled server. What this means is then it's possible to programmatically query, you know, any of these information without, you know, needing a specific software license without, you know, having to pull the entire model that you can just get the pieces that you need. The second very [00:30:00] important thing is that every times you hit send or publish, you know, every time your model goes from the host application to spec. We are creating a version of it. So it's always possible to go back in time and see what are the changes. And one actually very exciting features that we are about to release is being able to see visually this, you know, versions of the same model after you've published. So, uh, being able to see, you know, who's changed what and when, and how it's gonna be much easier. And this is really opening the door, you know, for many things that we. It can then enable on the, on the website of things, uh, such as, you know, maybe issue tracking, class detection. There's so much that, uh, we've, you know, constantly being requested by our, our lovely community members in terms of, you know, we love to do this in a web native environment.

You know, when can you do it for us?

Evan Troxel: And you talked about Mateo, like Power bi, being able to pull the information into there where I think one of the important things [00:31:00] to say about that and, and you, I. Love for you to elaborate on this is a lot of times people have a tool, they've already got it right? And so giving them access to maybe a different kind of information than they're used to seeing or be able to grab insights from this architectural data and look at it in a different way, a query it in in different ways.

I mean, I remember Nathan Miller from Proving Ground years ago at au. Having a Power BI dashboard of a Revit model. Right. And, and querying it, like asking it a question using, I think back then it was Cortana, right Now that's long gone ancient history in Microsoft terms, but, but I could immediately see the ability to, you could actually just use natural language. You could speak. A question into it potentially, and it would show you that information. And for many people on a team who are not the super nerds on the project, right? This [00:32:00] opens up doors, and this is why we're seeing chat g p t take off, right? Because the UI. It's something that everybody already does is I can ask a question, right?

I can just simply ask a question in normal terms and, and it understands me. That's what we're getting to, I think by allowing this data to be read by whatever can read it, and a lot of times that being a tool that they already have, that they're maybe not using or maybe using for something else. But hopefully already know how to use.

That's, uh, an amazing thing. I'm sure that's just like a light bulb goes off for those users

Matteo Cominetti: Yeah,

absolutely. Uh, what I always say in the circumstances is the, the motto of, you know, a company I used to work with called Case, which you, you might remember, was, you know, bill. Links equals data. And you know, I think people are maybe starting to realize just now a bit better. What that really means is, you know, [00:33:00] uh, the power of data, the power of being able to, to,


know, to. Get information out of it quickly and parse it with different tools. And, uh, just the past week, I think we saw two super exciting demos, uh,

created by Speckle community

members. They created integrations of spec and, uh, some AI chat, uh, based AI platforms. So we had an

example of, uh, Chat GPT, no, it was

a Pandas AI integration where again, you could ask questions and you would, you know, answers with information coming from the model. We had another one also super exciting. It was, uh, using, uh, long AI and, uh, the speckle comments. So within the 3d, uh, viewer environment that speckle offers, you can, you know, add comments. Usually it's users post comments. And in this case, we would have an AI for applying to your own comments and answering any questions you had on the model, on the selected objects, and also on any attached, uh, [00:34:00] PDFs or, um, Clients that were attached on the comment themselves, so you can really see how, uh, you know, having data that is more accessible, you know, is extremely powerful.

Being able to passe it with. power bi, you know, business Intelligent tools, uh, brings extreme value to our users and our Power BI connector is one of the most successful ones I would say, which I didn't expect. And then tools, like, you know, AI is, they're just making it easier for us to access and consume this information. Really, you know, taking a lot of the effort of, uh, tracking dropping components, uh, and passing the data.

Evan Troxel: The, the whole idea of, you know, and this seems so basic, but I don't know that, that, maybe not the audience for this show, but the wider audience, the people that those people deal with in their firms. For those people to be able to just say, show me all the fire rated walls on this project. That would be a big deal, right?

Because humans are so [00:35:00] good at pattern recognition, right? During the qa qc process. That could be an invaluable tool show. Like how, how I could just say how many classrooms are in the

school, right? And, and just to double check that we're actually meeting the program could be such a big deal. It's those kinds of basic things like that are overlooked.

I think when we talk about kind of the density of the information that is encoded into these models and actually being able to filter is that's what our, our eyes, that's what our senses are doing every day, right? We're filtering out 99% of the information so that we can actually deal with whatever's right in front of us. But that information is also in our models and for us to quickly with natural language and our eyeballs, say, say something and then see something, I think is just the opportunities there for, I mean, Quality control one thing, but just the making sure that we're actually on [00:36:00] track. That to me is just such a huge, huge thing that, that you are

enabling people to be able to do. It's really fun to watch the technology advance to the point where, Simpl because if somebody just asked me this question like, what's the, what's one of the biggest problems in aec? And it's like, it's not easy enough

for more people to get involved, right? The stuff that you're talking about, I mean, this is, this is why it is so hard to adopt tools in a, this is one of the big reasons why we are behind. I mean, you, you talk about coming at it from this, this collaboration and connection point of view, but. I mean, we all know how complex it is. We know that the projects are only getting more and more complex. We also know that innovation is happening at a pace that will outstrip adoption any day of the week and is only getting bigger, right?

The gap is only getting bigger. So to see this kind of excitement around the large language

model, uh, Chat GPT style, uh,

ais out there [00:37:00] to reduce, be able to reduce large data sets down to simple queries. Think that directly applies to the work that, that you all are doing and could really help the adoption of technologies like what we're talking


Dimitrie Stefanescu: Yeah, agreed. 100%. And here, kind of what we're most proud of is that it's not us that built that integration. It's our community that built it. And this kind of like literally gives us a sense of, okay, we're getting there. We're making it Because

Evan Troxel: mm-hmm.

Dimitrie Stefanescu: part of the reason Mateo and I like dedicated our lives pretty much to speckle.

Is because we wanted to scratch our own itches, which was back in the day, you know, use clunky APIs coming from one vendor or another just to get some very basic metadata was always not a trivial task. Right? And now, and that's what we are building. We're essentially, we're building that infrastructure that we never had in [00:38:00] our, let's say, AC careers.

So to see our community doing this, It's literally thrilling. It means we're doing something really good.


Evan Troxel: Very cool. So give us an idea of the kinds of connectors that are available

in Speckle cuz it's, it's a big,

big diagram.


Dimitrie Stefanescu: It's a big

Evan Troxel: gotten a

lot bigger over time.

Dimitrie Stefanescu: Yes. We went through a phase in which we kind of expanded to see what sticks, because there's a bit of a discovery process, which are the workflows that are most required, that are most painful within aec. what are the problems there? So we have quite a few of them now.

We're going through a phase to which we're kind of. Okay. We know which these workflows are and we're focusing on a core set of connectors. Uh, so. We are connecting essentially to Rhino Revit, the usual culprits, AutoCAD civil 3d, um, acad in a very, uh, early beta stage [00:39:00] currently.

Evan Troxel: Mm.

Dimitrie Stefanescu: Um, there's Excel of course, because this world runs on Excel and people like to, to dismiss Excel, but we think no, there's a lot of embedded knowledge in Excel that needs to be.

You know, given a place to manifest and interact with other places of ac. So let's go down that route. Uh, blender is a hugely successful connector as well, because it's open source visualization workflows are important for those eyeballs that you mentioned. Um,

Evan Troxel: Mm-hmm.

Dimitrie Stefanescu: and kind of that's a picture of the constellation.

We also

integrate with Q G I

Matteo Cominetti: We also have some structural connectors. I think you were forgetting about the Tela, a bunch of CSI products.

Um, we also have some, uh, Bentley, uh, connectors that were actually community contributions from a, so kudos to them. Uh, and I think those are, most of them, the, it's 25, so it's easy to let some slip times.[00:40:00]

Evan Troxel: Wow, that's a good, good number. Easy number to remember, right? Fantastic. So, Let's go down the open source road and, and you, you mentioned Dimitri, that the con contributions have kind of opened the door for this integration with chat G P T, but just. The idea of being open source, you kind of laid this out earlier in the conversation and the value that that can deliver to users potentially not being locked into a lifetime subscription, I think is what you said, Mateo dur, you know, paraphrasing what you said, but, but give us an idea of why that's important to you and, and, how that shapes your decision making. Uh, because. People have solved some of these problems. Maybe they're not looking at this holistically, but, and by people I mean the, the vendors that you talked about, right? They do try to solve problems for connecting their app to other [00:41:00] apps. Like we, we've seen the tremendous amount of work that McNeil has gone through for Rhino inside Revit, right?

For in, for example, um, or now we're seeing.

Form coming out from Autodesk and, and this cloud of connectors and all these things. So you're not alone right, when it comes to this, but you, I don't, you tell me, are you the only one out there that's doing this in an open source manner and, and like, why is that something that you've built your company on?

Dimitrie Stefanescu: Yes, we are the only ones doing that in an open source manner. So when you know smart people ask themselves the question, what is my office's data strategy for the future, right? Then they don't have many alternatives, um, to look at. And we are the only open source ones, open source one out there. Um, and why we've built our company.

On, on top of open source. It's because we can, we, there are commercial open source models that exist. There are open source companies out there [00:42:00] developing open source software that have crazy valuations and make crazy amounts of money. Um, open source is not free as in free beer. It's, it's, it's, this is Matteo's metaphor.

Now, uh, open source means you have the recipe to make. How on, on making bread, right? But if you want to make bread, you need to pretty much come with your own ingredients, purchase them, assemble them, and create bread. So essentially, kind of a for-profit goal as a company is not at all at odds with, with creating equal value within the industry.

Matteo Cominetti: Um, something to

Evan Troxel: Mm-hmm.

Matteo Cominetti: Yeah, I would like to say that, you know, um,

one key reason for which spec is open source, uh, is that, you know, we are building infrastructure at the end of the day, and. Infrastructure. Imagine roads and bridges. You know, connecting digital tools and data silos only really [00:43:00] works if you know you can freely, you know, Utilize it.

You know, there's no proprietary database that is being, you know, widely successful. There's no proprietary, you know, um, SDKs or APIs or frameworks that are becoming, you know, widely adopted, successful. So we believe that infrastructure has to be open source for others to be able to, to build on. And this, you know, I think the AC industry has been plagued by proprietary tools, by large vendor monopolies. And the result is that, It's been stagnant for years, right? We feel like our mission is dual, is, you know, bringing spec as a product, as a platform to the world and helping, you know, others build on top of us, but also, you know, bringing new, cool technologies that others can just take out and run away and use and even monetize on top.

We have a very permissive open source license ache too, and so we've already seen people. Kind of ripping off parts of Speckle, building on top of them, re utilizing them. And for us it's great to [00:44:00] see because it means, you know, they're know we are doing something useful. That goes beyond from building a company, building a product. We've seen our viewer, you know, which is now better and more powerful than ever being used in all sorts of applications. Same goes for, you know, some of our connectors, logic. We see now this bundling in some of our connectors, it breaks some of our licenses, but it's okay, you know, uh, at the end of the day, they're doing it, you know, to, to everybody advance and, uh, you know, we are fine with that.

Evan Troxel: It's interesting to think about how the industry has been held back by this idea of the intellectual property. You guys both talked about how you worked at firms where you weren't allowed to share, and this became kind of your nights and weekends project to build this open source. Product, but also community, right?

Of contributors who are making, helping to make this happen. And I think it was Ian Kio from a high par who said, design shouldn't always have to start with a blank page, right? There's all of this [00:45:00] knowledge and wisdom and the industry and their approach obviously is encoding that information into a high par function so that it can be reused and shared, right?

Like the whole idea of sharing that. That function, uh, that makes a thing with, so with a designer who maybe doesn't have the experience of designing one of a building with those components in it is fantastic because they're not starting from scratch every time. We don't always have to redraw the bathrooms.

We don't always have to redraw the operating room from scratch, right? We can start with an operating room and we can use best practices encoded into these functions from that firm's point of view. To jumpstart the, the design process on a project. This idea where you said people are ripping speckle off and it's totally cool, is a, is an amazing point of view to witness you say, because we don't see that very often in [00:46:00] aec, but it, we also can't deny that it happens all the time when an employee leaves a company. And goes to work for another company, they take that knowledge with them. They, it's not like the experience starts

over. Right? And so I think it's important to acknowledge that this happens all the time on the human level. How every CAD detail library out there is, has like so many

shared details that from another firm, right? And it, it happens all the time. And to think that that doesn't happen is ridiculous. And so, To just acknowledge that and say, you know what? We're gonna actually capitalize on that and help raise all boats. That's really kind of a missing piece to the industry that I, it, it has been a missing piece for so long, you're enabling the raising of the boats to happen by saying, you know, if you're on the right track, if you're making something that somebody wants to quote unquote rip off, right?

Because it, it's useful, it's valuable, but it's also going to lead to [00:47:00] additional. Innovation by something maybe you never even foresaw. I mean, I think that's, to me one of the most exciting things about sharing information and this, you know, the whole trademark and patent system and you know, like there's no. Well, there are some expiration dates, but the idea of, of things being like totally protected, it does stifle innovation, right? So there are big trade-offs in these decisions that are made that aren't necessarily fully foreseen when those decisions are made. Um, and, and to be able to pivot on that, I think, and be agile around those, you know, making changes to those for the benefit of all is something that we should take, be taking a closer look at.

But it sounds like you guys have kind of. You, you've adopted that very early on to say, this is open source for a reason. You can take it, you can fork it, you can do something else, but you're not starting from

scratch again.

Dimitrie Stefanescu: Yeah, absolutely. Because we feel, to add to what you say, we feel like when [00:48:00] the AC industry competes with each other on various software tools, it's like competing on who has the better pencil. Like the competition should be on

actually the designs produced by those pencil rather than on who has, yeah.

Evan Troxel: Right,

Absolutely. Well, what, what are we forgetting to talk about here? I would love to hear where your vision is for speckle and maybe give us a hint about what the current challenges and future challenges that you're interested in attacking are.

Dimitrie Stefanescu: Well, current challenges and future challenges, they all are intermixed, kind of in the head of founders. Um, so current challenges, we're doing a big product revamp. We started out as a very geeky product, um, that appeals to a very wide audience of computational designers. But then this Kiki product, uh, fails these computational designers when they wanna spread it.

Across to a wider team. So [00:49:00] we've taken that on board and we're, we're doing, going through like kind of a very heavy product, repackaging as we call it, internally, just to make the concepts easier to be adopted by a wider audience. Uh, future challenges that we see is essentially being able to capture and.

You know, equitably distribute value within the AC industry. When it comes to data, uh, there's a wider challenge around automation that we're trying to slowly start having an opinion on. These are places where things like high par or Pollination Cloud are already operating in that space, and we want to enable them to operate with an even bigger impact.

And we're looking on how to do that.

Evan Troxel: Mm-hmm.

Dimitrie Stefanescu: Those are top of mind challenges right now.

Matteo Cominetti: Yeah. And just to say I think we are, you know, a bit flirting with ai. You know, we definitely, uh, see spec as a data platform. You know, [00:50:00] that's where you want to host all your, say historical projects, you know, and being able to run queries, you know, see patterns, you know, get suggestions on. We haven't yet kicked off anything officially internally, but we know that spec the perfect. Platform for running machine learnings, you know, oris, you know, or um, anything data related. Uh, so we are still exploring, you know, playing around with things at the moment, but you see spec playing a key role. And as I was saying before, seeing community members, they build tools on top of what you've done.

It's the best form of validation, right?

Dimitrie Stefanescu: Yeah, and as Mateo was talking, I, I realized kind of what our, our biggest challenge that me and him, we like to complain about often is the opportunity is so huge within the AC space for innovation, for disruption, for bringing these

machine learning models and putting them in the hands of everybody from enabling these collaborative workflows, developing new connectors, focusing on extra workflows, uh, improving our viewer [00:51:00] doing that.

So, Our biggest challenge is doing all those things well and fast enough. This is where we

are really trying to find the best way in which we can move even faster than we are currently moving.

Evan Troxel: And what you said to kick that section off, uh, about making this more approachable to an even wider audience, I think gets back to that idea where I was speaking about it's, it's too difficult for so many people to get into this, and so really, what, what would this look like if it were easier as kind of a guiding. Principle, I think that a lot of software companies could work from. Uh, and I think they are, I think we are seeing that, we're seeing, you know, the, the smartphone changed everything, right? There's an app for that. The

UI matters. It so it, it matters so much if for adoption of a tool and for you to get adoption of. [00:52:00] Speckle out there. It really has to be so intuitive to use. I think that was, it's one of the things that. Apple or Steve Jobs or whoever was so good at, was like, you already know how to use this. Like, there's no learning curve, right? Like that, that was really kind of a, a game changer when it was like, no, the tool is more powerful and it is simpler. I think that is, that's that special secret magic sauce that you know. Is, is really the, the convergence of those things is what really, what enables something to take off. And so I I hope for the best with what you guys are doing because

it's a, it's a tall

Matteo Cominetti: we, The bar in the AC industry is quite low. If you look, you know, the interfaces of most of the tools that people are have to use on a daily basis are not, you know, the best. So, so a lot of the feedback we get to be, you know, from users is I, oh, I love the, you know, spec, UIs, the interface, the clear, but we, you know, we instead want to set our bars at very high end, as you say.

You know, trying to make them as, as user-friendly and, [00:53:00] and intuitive as possible. So definitely a big challenge nonetheless.

Evan Troxel: Tell everybody where they can find what you're working on, follow along and contribute. I think an important piece of this, because you are open source, I'm sure there's a lot of people listening to this who would be interested in not just using, but also maybe they have an idea or they wanna see what's going on under the hood. And so I'll include links to all these in the show notes so that people can easily click on them. But I would also love for you to say it, say them to them.

Dimitrie Stefanescu: Absolutely. So probably the best place for a person new to Speckle to look at what's going on is our community forum, which is speckled community. I know it's a strange url, but it exists. Uh, then our. Our, our GitHub organization is Speckle Systems, and then obviously our website, which is Speckle Systems.

Uh, probably the most fun part is, is our community forum, though

Matteo Cominetti: if you

want to [00:54:00] try Speckle, you haven't tried it before, you can get a free account, unlimited resources, that you can create unlimited projects at Speckle.xyz. So you get

an account if you haven't yet.

Evan Troxel: Yeah, I think everybody needs to take it for a spin, and because you have 25 different connectors, it won't be difficult to find one to play with, and then that opens the door to the next one. So, Then the dominoes start to fall, which is, which is fantastic. So thank you both for taking the time to have this conversation. Get the word out about what you're doing in the open source community in a e c, uh, I, an incredible undertaking. And, and thank you for all of your work. And I'm so happy that you did find funding to actually be able to scale up and bring this to a larger community than you could. Nights and weekends are just toiling away with a small team and really being able to expand the team.

That's, it's a big deal. And it's, it's great to see investors who are willing to [00:55:00] take a risk and work in the AEC sphere if they weren't already. I, I don't know the makeup of your, of your funding, but it seems like, uh, a worthy problem worth funding. And so, you know, I'm rooting for you guys. This has been a great



Dimitrie Stefanescu: Even the same.

Thanks for having us

Matteo Cominetti: Yeah, and thanks also to all our supporters. You know, we wouldn't have made it without our community, our contributors, you know, this is really a shared effort that we're,

Dimitrie Stefanescu: Especially the crazies who joined at the beginning

Hall and Claire, and so, yes.

Evan Troxel: great. All right guys.





Dimitrie Stefanescu: Thank you.