Urban Forecast

Antonio Canero explains PRYZM's browser-based AI design tool that integrates GIS data and automates BIM workflows for architects.

40 min Listen to the original
AEC TechnologyArtificial IntelligenceBIMSaaSReal EstateArchitectureGIS

Summary

Antonio Canero introduces PRYZM as a browser-based SaaS platform that integrates GIS site data with BIM authoring, aiming to assist rather than replace architects. He describes the AI as a data reader that automates site analysis and repetitive documentation tasks while keeping the human in the loop for decision-making. Canero addresses the competitive landscape, noting that while tools like ChatGPT Astra can generate 3D models, PRYZM's engine is required for accurate, constraint-aware modeling. He discusses the erosion of Revit's dominance due to accessibility and cost, positioning PRYZM as a complementary or alternative tool for early-stage design. The episode concludes with a demonstration of the platform's interface, pricing model, and interoperability limitations.

Episode sections

  1. 0:00 – 2:19

    Introduction and PRYZM Overview

    Antonio Canero introduces PRYZM as a browser-based AI-assisted design tool for architects and developers. He emphasizes that the software reads data and proposes options but does not create or draw, aiming to make life easier rather than replace architects.

  2. 2:19 – 6:49

    Data Integration and Site Analysis

    PRYZM connects GIS site data with BIM authoring in the web. It starts projects from the site, pulling publicly accessible cadastral data via APIs to provide context like terrain, sun impact, and local building constraints such as setbacks and height limits.

  3. 6:49 – 12:11

    AI Role and Human-in-the-Loop Workflow

    Canero clarifies that AI acts as a reader that connects user intent with the modeling engine. He states AI is not a decision-maker but enables faster decisions by automating data review, while the architect retains control to accept, decline, or comment on proposals.

  4. 12:11 – 14:06

    Impact of ChatGPT Astra on Workflow

    Canero expresses being impressed and concerned by ChatGPT Astra's ability to generate 3D models from images. He notes that while Astra is a strong enabler for understanding data quickly, PRYZM's engine is required to build accurate models, and he advocates for slowing down to strategize rather than panic.

  5. 14:06 – 25:45

    Future of BIM and Revit's Role

    The speakers discuss whether Revit will become merely a system of record or if AI will erode its dominance. Canero agrees Revit is a commodity in the AEC sector but believes accessibility and cost will drive a shift toward browser-based tools and potentially proprietary tools built by large companies.

  6. 25:45 – 28:03

    PRYZM Business Model and Pricing

    Canero confirms PRYZM is a SaaS, not open source, though it uses open standards like IFC. The current price is $49 per user per month, with discounts for volume. Enterprise clients can eventually embed the software or create white-label versions.

  7. 28:03 – 33:17

    Live Demo in Barcelona

    Canero demonstrates PRYZM using a parcel in Barcelona, a pilot city with rich data. He shows how to select a parcel, view cadastral data, analyze sun impact, and define building envelopes and massing directly in the browser without switching applications.

  8. 33:17 – 34:40

    PRYZM Interface and AI Automation

    Antonio demonstrates PRYZM's interface, showing how users can modify envelope levels and access site data. He explains that the platform uses a natural language engine to automate tasks like creating walls and windows without consuming tokens for basic operations.

  9. 34:40 – 35:00

    Model Selection and Token Usage

    The speaker notes that users can choose their own AI models by providing an API key, which allows them to consume tokens and pay for usage beyond the initial free tier. This gives users control over the underlying AI engine.

  10. 35:00 – 36:58

    Design Iteration and Live Documentation

    Antonio shows how to automate window creation on walls and modify their type (e.g., to timber casement) via chat commands. He highlights that the platform generates live documentation, including elevations and plans, in real-time as the design changes.

  11. 36:58 – 38:07

    Software Interoperability and Development Stage

    Antonio states that PRYZM can currently import from Rhino and Revit but does not yet natively create Revit files, noting that interoperability is a work in progress. He mentions that while the tool can handle detail design, it is currently more suited for earlier stages.

  12. 38:07 – 39:30

    Impact on Architect Workflow

    The speaker argues that AI will take over unglamorous tasks like redrawing, tagging, and dimensioning. This shift allows architects to make decisions earlier in the process by discovering constraints and context issues faster, moving the heavy lifting to the machine.

  13. 39:30 – 40:05

    Accessing the Software

    Antonio directs listeners to try the software by requesting a demo via LinkedIn or visiting pryzm.so to click the demo button. He confirms the company is in an early stage where direct contact initiates the trial process.

Claims worth verifying

PRYZM connects GIS site data with BIM authoring in the web, starting projects from the site.

Unidentified speaker (speaker_0) ·2:42

How to verify: Test PRYZM platform to confirm GIS integration and web-based BIM authoring capabilities. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Primary source

Site data is sourced from publicly accessible APIs, with varying richness by region.

Unidentified speaker (speaker_0) ·3:40

How to verify: Check PRYZM documentation for data sources and compare with public cadastral records in pilot cities. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM provides site analysis including sun impact, temperature, and contextual building data.

Unidentified speaker (speaker_0) ·4:56

How to verify: Use PRYZM demo to verify availability of sun impact, temperature, and contextual building data features. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

AI is not a decision-maker; it enables faster decisions by automating data review.

Unidentified speaker (speaker_0) ·6:59

How to verify: Review PRYZM platform documentation for AI role and decision-making capabilities. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

ChatGPT Astra can generate 3D models from images quickly.

Unidentified speaker (speaker_0) ·12:47

How to verify: Test ChatGPT Astra to verify ability to generate 3D models from images. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Revit is considered a commodity in the AEC sector.

Unidentified speaker (speaker_0) ·22:51

How to verify: Review industry reports or surveys on BIM software adoption and market share in AEC sector. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM uses open standards like IFC for interoperability but operates as a proprietary SaaS.

Unidentified speaker (speaker_0) ·25:45

How to verify: Check PRYZM documentation for IFC support and business model details. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Primary source

Barcelona is a pilot city for PRYZM with rich data availability.

Unidentified speaker (speaker_0) ·28:27

How to verify: Check PRYZM documentation or press releases for pilot city information and data availability in Barcelona. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM uses a natural language engine that can automate processes like creating walls and windows without consuming tokens for basic tasks.

Unidentified speaker (speaker_0) ·34:00

How to verify: Visit pryzm.so and attempt to use the AI chat features for basic design commands to see if token usage is deducted or if it is a free tier feature. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Users can bring their own API key to PRYZM to choose which AI model they want to work with and pay for token consumption.

Unidentified speaker (speaker_0) ·34:40

How to verify: Check the PRYZM settings or documentation on pryzm.so for an option to input an external API key (e.g., OpenAI, Anthropic) and review the pricing page for token-based billing structures. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Primary source

PRYZM can import files from Rhino and Revit.

Unidentified speaker (speaker_0) ·37:27

How to verify: Test the import functionality in the PRYZM platform or review the official feature list on pryzm.so for supported file formats. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM cannot yet natively create Revit files, though it can bring elements into PRYZM.

Unidentified speaker (speaker_0) ·37:37

How to verify: Attempt to export a design from PRYZM to a .rvt (Revit) file format; if the option is missing or results in a generic format, the claim is verified. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM generates live elevations and plans on demand.

Unidentified speaker (speaker_0) ·36:21

How to verify: Use the PRYZM platform to generate an elevation view from a 3D model and check if it updates automatically when the 3D model is modified. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM is currently more suited for stages 1-3 of design, with stage 4 (detail design) being a future goal.

Unidentified speaker (speaker_0) ·37:01

How to verify: Review PRYZM's marketing materials or case studies on pryzm.so to see which project phases are highlighted as supported versus future roadmap items. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Antonio Canero previously worked in the UK.

Unidentified speaker (speaker_0) ·37:01

How to verify: Check Antonio Canero's LinkedIn profile or professional biography for employment history in the UK. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM can automate the creation of windows on walls every two meters.

Unidentified speaker (speaker_0) ·35:10

How to verify: Demonstrate this specific command in the PRYZM platform during a trial to see if the AI places windows at the specified interval. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM can modify window types, such as changing them to timber casement, via chat commands.

Unidentified speaker (speaker_0) ·35:36

How to verify: Use the AI chat to request a change in window material and style, then verify if the 3D model updates to reflect timber casement windows. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

PRYZM handles unglamorous work like redrawing, automatic tagging, and dimensioning.

Unidentified speaker (speaker_0) ·38:13

How to verify: Test the platform's documentation generation features to see if it automatically tags elements and adds dimensions without manual input. For recent events, verify 2026-08-09 through 2026-11-09. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify PRYZM's current pricing and enterprise discount structures directly on pryzm.so or via sales contact.
  • Check the specific GIS data availability and API endpoints for target markets to assess data richness limitations.
  • Review the interoperability roadmap for native Revit export capabilities, as current limitations may hinder workflow integration.
  • Analyze competitor offerings like ChatGPT Astra and other AI-BIM tools to assess the durability of PRYZM's technical moat.
  • Assess the regulatory and liability implications of AI-generated design proposals in local jurisdictions.
  • Investigate PRYZM's customer acquisition costs and retention rates given the direct sales and demo-based go-to-market strategy.
  • Review the financial health and funding status of PRYZM to evaluate long-term viability as a SaaS provider.
Full transcriptexpand
0:00
Unidentified speaker (speaker_0): What's gonna be the future of AI and I'm scared about how things are moving so quick. At the same time I'm excited and positive.
0:08
Unidentified speaker (speaker_1): Antonio Canaro is a platform architect leading the architecture and development of Prism, shaping the technology behind the platform.
0:16
Oliver Lowrie: To what extent is AI involved in the kind of authoring of the project?
0:22
Unidentified speaker (speaker_0): It reads data and proposes. It never creates. it never draws. We are not replacing architects at all. I think the goal is to make life easier. What we have is like we have the plan view to the left and then the 3D to the side with real heights.
0:40
Oliver Lowrie: What changes about the way that the architect works?
0:48
Oliver Lowrie: Antonio, welcome to Urban Forecast Podcast.
0:52
Unidentified speaker (UNKNOWN): Thank
0:53
Unidentified speaker (speaker_0): you very much, Oliver. Thank you.
0:55
Oliver Lowrie: Um so you uh were originally an architectural technologist, uh and now you run Prism, um, which is uh the software that we're gonna be talking out about today. But do you remember the the kind of the moment where you're there, presumably kind of late at night in an architectural studio do drawing stuff.
1:15
Oliver Lowrie: and you were like, you know what, there must be a better way of doing this. I need to build a software company. Do you can you remember that moment?
1:22
Unidentified speaker (speaker_0): Uh I think uh it has happened slowly and it was not like an inflection point or a big one. But I think it's a dream for anyone that likes technology, like design, um, I like 3D um visuals. to build its own technology. And it started
1:42
Unidentified speaker (speaker_0): a bit like a personal dream that thanks to technology nowadays it became real.
1:51
Unidentified speaker (speaker_0): And I always have a vision which is
1:55
Unidentified speaker (speaker_0): make
1:57
Unidentified speaker (speaker_0): authoring tools in the browser accessible to everyone
2:02
Unidentified speaker (UNKNOWN): and
2:02
Unidentified speaker (speaker_0): not to to a few people. So that's how it started, and it's how it's it's going now.
2:11
Oliver Lowrie: Um and so in a kind of in a nutshell, what is it that PRISM does?
2:19
Unidentified speaker (speaker_0): Prism is a design tool for architects, uh builders, uh landowners. land owners, or future landowners, anyone that is in the construction sector
2:33
Unidentified speaker (speaker_0): that works in the browser. That's kind of like the quick definition. But if we go a little bit more in detail,
2:42
Unidentified speaker (speaker_0): something that Prison does that I find quite valuable is that it connects GIS, so site data. with bin
2:52
Unidentified speaker (speaker_0): authoring in the web. So when you start your project, you
2:59
Unidentified speaker (speaker_0): start always from the site,
3:02
Unidentified speaker (speaker_0): and then without going to any other platform, you can start designing and detailing it all in the same platform, all in the browser. So it's really collaborative, anyone can jump in. and check the design.
3:18
Unidentified speaker (speaker_0): It's easily accessible and easy to share as well.
3:23
Unidentified speaker (UNKNOWN): And
3:23
Oliver Lowrie: where's it getting the site data information from? How
3:28
Unidentified speaker (UNKNOWN): is
3:28
Oliver Lowrie: it?
3:28
Unidentified speaker (speaker_0): Good question. At the moment we are working with a few pilots uh countries. Everything is machine readable data. It's data that
3:40
Unidentified speaker (speaker_0): AI can get from APIs. In prison we whenever you choose a parcel, we tell you the cadastral information where the data come from. Everything is publicly accessible. And not in every country the same data is accessible. There are
4:00
Unidentified speaker (speaker_0): some areas or some regions where we have really rich data, and there are some others where you can get the basics, which is cadastral,
4:10
Unidentified speaker (speaker_0): number,
4:13
Unidentified speaker (speaker_0): the surface, perimeter,
4:17
Unidentified speaker (speaker_0): and pretty much that's all. Then you can go into some rich regions where you can get what's the maximum surface that you can build. maximum height, setbacks, etc. And then from there you can move into into the health frame tool.
4:33
Oliver Lowrie: So it's got the planning sort of information as well as the actual existing buildings and topography on the site.
4:41
Unidentified speaker (speaker_0): Yes, uh good point. So I didn't cover this, but of course uh whenever you go into the site view, you can interrogate the model in 3D or in plan view. You can get all the context buildings, terrain,
4:56
Unidentified speaker (speaker_0): water
4:58
Unidentified speaker (speaker_0): sources, roads, pedestrian, and you can get also sun impact, so how the sun moves during the day. So you could get all this uh easy access information plus the the site data.
5:15
Oliver Lowrie: And then so you can start designing straight away in the browser, I'm the architect, or maybe I'm not the architect, maybe there's a developer and they just start kind of building their own building on the site. And is it is it telling the developer or the architect, no, that's not compliant with the local restrictions? I mean I think in the UK that would be pretty hard thing to do
5:35
Oliver Lowrie: because the planning law is written in quite a obtuse way.
5:39
Unidentified speaker (speaker_0): Uh so again it's depending on the country. Uh we didn't go that far yet. I think the more data becomes accessible uh
5:48
Unidentified speaker (speaker_0): to us,
5:50
Unidentified speaker (speaker_0): publicly available, the easier will be for us to define what can you build and what not. But in some pilot uh cities we are able to define precisely the setbacks of a plot. or of a uh of a parcel. We can precisely say if what you are designing is going out of
6:10
Unidentified speaker (speaker_0): the area where you should build, and if you are designing more square meters than what's compliance on the footprint, we will be able to highlight it. That's the maximum that we have gone. The same with height or with number of uh levels of of the building.
6:28
Oliver Lowrie: Um and I think uh so we originally got in contact because I saw a demo of something that you were doing, and I was like, Oh, that's cool. Um, to what extent is AI involved in the kind of authoring of the project? And I I don't want to use necessarily the word design, but uh to what extent is AI involved about in building the model?
6:49
Unidentified speaker (speaker_0): So the AI what does is like it reads data and proposes. It never creates, it never draws.
6:59
Unidentified speaker (speaker_0): The engine is the one doing all the maths and the one creating in Prism.
7:06
Unidentified speaker (UNKNOWN): So
7:07
Unidentified speaker (speaker_0): the AI you could see it as the reader. From the moment where you talk with Prism AI chat, the AI is reading what you are proposing and then connecting with the engine. And if, for example, something that we are investigating now, maybe we can speak a bit further in detail later, with Astra,
7:27
Unidentified speaker (speaker_0): now the AI what does is like it reads any data that you provide it, so if it's uh floor plans, elevations or images of the building.
7:39
Unidentified speaker (speaker_0): and it's uh gather the data and bring specifically what's there for the engine to recreate in this specific use case. So and where the architects live, or the engineer, or the designer, or the landowner.
7:58
Unidentified speaker (speaker_0): So I think what is important, and I think you are really into that, is like uh we are not replacing architects at all. I think the goal is to make life easier. So the architect will be then able to review the proposal, to the review what the eye is proposing, and accept or decline, or comment, or follow up.
8:18
Unidentified speaker (speaker_0): It's uh it's basically enabling the architect to get data quicker
8:24
Unidentified speaker (UNKNOWN): and
8:24
Unidentified speaker (speaker_0): be bit quicker.
8:25
Oliver Lowrie: So how how on what basis is it for so it's reading the AI is reading the topography and it's saying okay I think you need a retaining wall here or or I think you know the like to what extent is it like I I don't love the idea of humans just being there to accept the proposals of the AI. I feel
8:43
Unidentified speaker (UNKNOWN): like
8:43
Oliver Lowrie: that's the the analogy which I'm stealing off somebody that I can't remember. I was from a guy from Benoit who said AI could be uh a centaur or a reverse centaur. A centaur is a has a the brain of a human and the body of a horse, so it's very powerful and strong, but it still thinks like a human. And that's that's the way that
9:03
Oliver Lowrie: you could optimally imagine the relationship between AI and architect.
9:08
Unidentified speaker (UNKNOWN): The
9:09
Oliver Lowrie: reverse centaur is what is what I think you might describe as um well, it has a the head of all. a horse, the brain of a horse, and the the legs of a human, so not particularly fast, but thinks with a small brain. And that you know, the the idea that the human is just there to sort of validate, go, yep, that looks right, or yep, that seems like that's compliant. Um,
9:29
Oliver Lowrie: that doesn't fill me with inspiration. Um so ha why is you know, to what extent is the AI like proposing and what is it proposing from? what data set is it using and how does that compare to the experience of uh like a designer that traditionally would have done the design?
9:48
Unidentified speaker (speaker_0): Um at the stage where we are, we we are not uh at the point where AI can propose you to create a retained wall there, at least in prison, I don't know elsewhere. What we are doing is pretty much automating the process, the tiring and
10:08
Unidentified speaker (speaker_0): not fun and automatic process of
10:13
Unidentified speaker (speaker_0): reviewing any data that's accessible and can make you take decisions. So definitely it's not only about approving or validating what AI is telling you. For example, in the use case of recreation, as we are seeing now a lot out there. it's uh it works like that. So the AI is gonna propose you, okay, this is what I'm seeing in your images. Uh,
10:33
Unidentified speaker (speaker_0): do you want to build it? Um Prism will be able to recreate fairly detailed uh beam model that after you can continue to work with, no?
10:43
Unidentified speaker (speaker_0): But if we go into um the example that you said, uh Prism is not there yet at the moment.
10:51
Unidentified speaker (speaker_0): The AI is just reading data
10:55
Unidentified speaker (speaker_0): enabling the designer or the architects to take quicker and faster decisions. It will not be the decision maker never. And for now, at least in Prism, it's just
11:06
Unidentified speaker (speaker_0): enabling you to move faster and to take faster decisions. For example, if
11:13
Unidentified speaker (speaker_0): you want to understand the site, you can ask what are the site constraints that you can tell me about and the architect will be able to know them. It will not be still
11:28
Unidentified speaker (speaker_0): possible to ask a prison AI chat to
11:35
Unidentified speaker (speaker_0): go much further than that at this stage. However, things are moving quite fast. And the more data that we have available out there, machine readable, um with uh coming from a nice source of truth, the more we will be able to do.
11:51
Oliver Lowrie: When you first saw the demos of the new ChatGPT Astra, um basically building buildings and kind of well, certainly recreating buildings. Not accurately necessarily, but like from bits of video and photograph and creating 3D models in in Blender. Did
12:11
Oliver Lowrie: it fill you with dread because it looks like it's going like you know, it you may have had an advantage two weeks ago that now is being eroded by like something that's proprietary and you can get for twenty 20 bucks a month? Did that scare you?
12:29
Unidentified speaker (speaker_0): Um honestly, I think I'm I'm scared about how things are moving so quick. Um but I'm at the same time I'm excited and positive. So honestly, when I saw first time to answer your question Astra
12:47
Unidentified speaker (speaker_0): uh demos, I was first uh super impressed. I guess like everyone, and also concerned. I was like, well, what next? And even that is
12:57
Unidentified speaker (speaker_0): strong, what's able to do really quickly. Then I started to investigate what that will mean for Prism, and actually I built already the engine, and pretty much it was a big enabler for Prism because we could get the advantage of the AI. model to understand data really
13:17
Unidentified speaker (speaker_0): quick and the engine of Prism to build. But I think it's a a really important moment nowadays to slow down, to uh step back and
13:31
Unidentified speaker (speaker_0): understand where we want to put the effort in, because at least for Prism,
13:36
Unidentified speaker (speaker_0): uh
13:37
Unidentified speaker (UNKNOWN): the
13:37
Unidentified speaker (speaker_0): decisions that we are gonna take now. are gonna impact our roadmap in the future quite a lot because AI could be doing anything that we are putting a lot of energy on really quickly, extremely cheap in the upcoming month. So and I think that's also the same for architects or engineers
13:57
Unidentified speaker (speaker_0): or designers. We need to probably reverse engineer our the way we are working and Um I think it's the correct moment to start thinking about it.
14:06
Oliver Lowrie: I had a um did a podcast with another Antonio, another Spanish Antonio from that open company,
14:11
Unidentified speaker (UNKNOWN): uh
14:12
Oliver Lowrie: who I think you know, and um he was sort of saying, Well, look, this is you know, this is the nuclear bomb moment for architecture, and this is way before Astra. Because people are going to be able to build their own BIM tools, and I think w it that you
14:27
Unidentified speaker (UNKNOWN): know
14:27
Oliver Lowrie: the new chat GPT model moves almost past that where you're
14:31
Unidentified speaker (UNKNOWN): like,
14:31
Oliver Lowrie: oh wow, you know, to what extent do we really
14:34
Unidentified speaker (UNKNOWN): in
14:34
Oliver Lowrie: the long run need BIM tools? Like, do we, or is it just like you know, Revit sort of becomes the cassette? I'm not saying it's out of date, I'm saying it becomes like the system of record. The Revit file itself becomes what other things. it's the it's the interoperable uh element,
14:55
Oliver Lowrie: but like basically you can create Revit models in lots of different ways. So you don't need Revit per se. I did something with Astra
15:03
Oliver Lowrie: where I was like, oh I wonder if it can do this. And I I've basically lived in my house for a couple of years and I've I'd always been meaning to put in an application to go up one story. Never quite got around to it. So I was like I'll just
15:15
Unidentified speaker (UNKNOWN): I'll see
15:15
Oliver Lowrie: if Astra can do
15:16
Unidentified speaker (UNKNOWN): it.
15:16
Oliver Lowrie: Um and so I put in my dress, it looked up planning doc planning out a planning application that somebody had submitted on that property a few years ago, downloaded the drawings, built a model in Blender in about like five minutes of the whole existing property, and then built the upwards extension. Um and I probably prompted it like four or five times to
15:36
Oliver Lowrie: get to that point. It was amazing. But then and it was like it built it like in Blender without me kind of being able to see it.
15:43
Unidentified speaker (UNKNOWN): And
15:43
Oliver Lowrie: I was like, well that's weird. Like how you know how do you even where even is this thing? And it was like, oh I've just you know I've got a copy of Blender that I just use. I was like what? So I'm not even you know this is all happening in the darkness somewhere. And I was like show me. So it's like okay download Blender. So I was like downloaded Blender and I I was it like ta-da, there you go. Went into Blender. I can't use Blender I don I've never never
16:03
Oliver Lowrie: used it in my life. So I couldn't even rotate anything. I was completely useless. So I went back into ChatGPT and I was like, well,
16:09
Unidentified speaker (UNKNOWN): I
16:09
Oliver Lowrie: don't actually want it in 3D. I need this as the you
16:11
Unidentified speaker (UNKNOWN): know,
16:12
Oliver Lowrie: this is a planning application. I need 2D plans, sections, elevations, et cetera, et cetera.
16:17
Unidentified speaker (UNKNOWN): And
16:17
Oliver Lowrie: it was like, oh, say no more. Ta-da! And it just made them. And I don't even know what, again, I don't know what software it used because it must have been some open source 2D CAD software that it could import the. Blender thing for. And I was like, wow, I did all of that. And I've got these sheets. They're I put I they're on LinkedIn if you want to see them, because I posted about it over the weekend. They're not perfect, but they're not far off. And that was all
16:37
Oliver Lowrie: like kind of one shot-ish. Um and
16:40
Unidentified speaker (UNKNOWN): I
16:40
Oliver Lowrie: was like, wow, like for people like me that I I don't use Revit on a daily basis, I
16:44
Unidentified speaker (UNKNOWN): don't
16:44
Oliver Lowrie: use Archad, I
16:45
Unidentified speaker (UNKNOWN): don't use
16:45
Oliver Lowrie: Revit, I don't, you know, I
16:46
Unidentified speaker (UNKNOWN): just
16:46
Oliver Lowrie: about use Vectorworks because that's what I can still use from about you know 20 years ago.
16:51
Oliver Lowrie: We're now like being upskilled back into the conversation where we can, you know, like it's sort of dangerous that you get other directors that are like aged out of the software back in going, I can I can use Rhino
17:01
Unidentified speaker (UNKNOWN): now.
17:02
Oliver Lowrie: Um, but because you can just prompt it. But also there's this other thing of like,
17:06
Unidentified speaker (UNKNOWN): in
17:06
Oliver Lowrie: the future, do people really end up needing to touch the software? You know, it it like you don't even need to necessarily see it. That was the crazy thing. I we did
17:15
Oliver Lowrie: you know, I said we like I and an AI was able to produce this stuff without having the software, and that feels strange and meaningful for the architectural industry. I don't know what the meaning is yet. You know, I'm still like trying to work it out.
17:32
Unidentified speaker (speaker_0): Yeah.
17:34
Unidentified speaker (speaker_0): Um, so
17:37
Unidentified speaker (UNKNOWN): do
17:37
Unidentified speaker (speaker_0): do you want me to understand answer any specific point or
17:40
Oliver Lowrie: yeah yeah yeah just go with it.
17:42
Unidentified speaker (UNKNOWN): I
17:42
Oliver Lowrie: mean what do you think? Like that's I like it it's all very new and I w I I want to know your opinion on what you think the kind of the future is.
17:49
Unidentified speaker (speaker_0): Um yeah I saw your post and I was impressed together with some other post uh I at the time I didn't test um Astra myself, I have been testing it all all the over the weekend and I was it equally impressed what was doing in my own application. So I was like, wow, before I was using
18:09
Unidentified speaker (speaker_0): some AI models, I have been using from Anthropic, Fabul, Optus, and they were far from being able to do what Astra is able to do now. So I was extremely impressed.
18:23
Unidentified speaker (speaker_0): In prison, you could potentially model as well. without needing to do the classic UI interface job, which relates to what you are saying in Blender.
18:37
Unidentified speaker (speaker_0): I think
18:39
Unidentified speaker (speaker_0): you will always need to at least use natural language to modify the model, right? Yeah. But that might be it. It could end up just
18:52
Unidentified speaker (speaker_0): everything within the context of a chat.
18:56
Unidentified speaker (speaker_0): I was in some of my posts, I asked the people what do they think where
19:03
Unidentified speaker (speaker_0): AI is gonna, what's gonna be the future of AI and and the AIC software industry towards uh the interaction from the user with the model
19:14
Unidentified speaker (UNKNOWN): and
19:14
Unidentified speaker (speaker_0): some people were saying that probably it will be only natural language. But in my experience, because I use uh I have been using Revit and other tools for many many years as well as Prism now, I always love to be able to select an element, it's quicker, to rotate, to even sometimes move it
19:34
Unidentified speaker (speaker_0): with my own hands. I mean when I say with my own hands it's with the mouse.
19:37
Oliver Lowrie: Yeah.
19:38
Unidentified speaker (speaker_0): Not always just describe what I want to do. Sometimes it's actually way slower to describe, okay, I want to move this wall 10 millimeters up to the left because it's really what I want to do. However, the direction that I'm seeing now in the industry, which is extremely interesting from my point of view,
19:58
Unidentified speaker (speaker_0): is the intent.
20:00
Unidentified speaker (speaker_0): So what we are trying to do in prison. and I think many other software companies are gonna go in this direction, is like the user, the designer, or the architect will define the intent, what he intends to achieve
20:15
Unidentified speaker (speaker_0): on any different level. So, for example, if you want to rather than just move a wall or
20:23
Unidentified speaker (speaker_0): move a door, what you want is the living room to be temporary. 10% bigger without impacting the bedrooms, but potentially impacting the entrance and the kitchen. You define the intent and
20:39
Unidentified speaker (speaker_0): the AI, so the AI model together with your engine, with the software engine, will reshape Um and then you could move along with the design working together with the agent.
20:51
Unidentified speaker (speaker_0): because you will not always be agree, uh you will need to redefine and you could always reshape yourself on demand.
21:00
Unidentified speaker (UNKNOWN): It
21:00
Unidentified speaker (speaker_0): makes sense.
21:01
Oliver Lowrie: Yeah. So your Prism sort of set up for a hybrid way of working.
21:07
Unidentified speaker (speaker_0): Yes. I think we are gonna need to potentially learn how to work with agents uh more and more,
21:13
Unidentified speaker (UNKNOWN): uh
21:13
Unidentified speaker (speaker_0): rely on them. on things that they can do really, really good, and of course being always in the loop and always uh with the control.
21:23
Oliver Lowrie: And what do you think? So uh obviously it was last week, well not obviously, uh last week was the Autodesk University where they announced that the uh Revit essentially will eventually be sort of replaced by former.
21:39
Oliver Lowrie: um, which I suppose is the same, you know, that's the kind of the same direction of travel.
21:45
Oliver Lowrie: But it's interesting, sort of working with more people that are building AI things for the AEC space, that Revit's um
21:57
Unidentified speaker (UNKNOWN): I don't want
21:58
Oliver Lowrie: to say like Monopoly or Stranglehold because those seem like overly negative, but like there is a ubiquitous. ubiquity to Revit that means that I think a lot of people that are building AI stuff, and I don't know where you stand on this, are building it like basically for Revit because they're like, Well, it's you know, it's the most interoperable way of so Revit has this kind of legacy
22:17
Oliver Lowrie: theme of like being the way that people want to end up communicating. So it kind of it keeps its relevance, even though the tools around it might change. But I do think
22:29
Oliver Lowrie: you know what yourself and the open comp that open company would probably believe that the this the that the ability of larger businesses to um develop their own software might start to erode the licenses that Autodesk is able to charge for, um, maybe. But they're still gonna have this ubiquity of of information
22:48
Oliver Lowrie: transfer. Do you agree with that?
22:51
Unidentified speaker (speaker_0): Yes, yes. I think uh Revit is the commodity in the AC sector um and habits are really strong to
23:01
Unidentified speaker (speaker_0): to change. Uh me my I myself I love Revit. Uh I love uh the what's enable people to do. But of course it's a tool as everyone speaks uh from that has a lot of years and the rapidly technology is enabling um
23:21
Unidentified speaker (speaker_0): many many people that before they were not able to be part of the designing game of the software of the upcoming years
23:31
Unidentified speaker (speaker_0): and
23:32
Unidentified speaker (speaker_0): I think um
23:35
Unidentified speaker (speaker_0): even if uh
23:37
Unidentified speaker (speaker_0): most of people are using Revit
23:40
Unidentified speaker (speaker_0): uh nowadays because of different reasons, this will likely change in the near future, b mainly because of accessibility to software and prices. So as Antonio from that open company mentioned, uh he thinks Things that probably very
23:59
Unidentified speaker (speaker_0): soon many big companies will likely develop their own authoring tools. I believe the same, and well, authoring or coordination tools. Prism
24:12
Unidentified speaker (speaker_0): goal was to enable people to work in the browser and out or in the browser with a low cost. So, for example, here in Portugal where I am, not everyone can pay a Revit license, and not to speak about other regions in the world. And I think it will be so good if
24:31
Unidentified speaker (speaker_0): everyone can visualize and interact with a 3D Beam environment in their practice.
24:39
Unidentified speaker (speaker_0): However, I also believe that many big companies they will build their own tools and they will uh create their own answering tools really real really quick thanks to that open stack like uh that open company stack. Um
24:56
Oliver Lowrie: what so there's uh obviously Antonio other Antonio believes in um open I was gonna say open weight, he believes in open source software and so his kind of business model uh um is
25:08
Unidentified speaker (UNKNOWN): that
25:08
Oliver Lowrie: he's not really charging for the well he's charging kind of minimal for the software licensing and he's actually those kind of open source licenses, he's making money through other things, consultancy, etc. etc. Are you your thesis, if I'm paraphrasing it correctly, is that you is that the the big the big paradigm shift for you for Prism
25:28
Oliver Lowrie: is that you are authoring in the browser, so you're not having to download the software onto your computer. Um
25:36
Oliver Lowrie: but are you are you still doing a kind of monthly license? How does it work in terms of is is it open source and um how does one pay for
25:43
Unidentified speaker (UNKNOWN): it?
25:45
Unidentified speaker (speaker_0): So it's not an open source um
25:48
Unidentified speaker (speaker_0): project. Uh is connected with open standards, so we are IFC
25:55
Unidentified speaker (speaker_0): natively creating elements and therefore they are uh IFC
26:01
Unidentified speaker (speaker_0): readable. They can with we can import and export IFC
26:07
Unidentified speaker (speaker_0): and we are focused on open standards so the schedules could be open easily on other tools and so on. But
26:19
Unidentified speaker (speaker_0): Prism is
26:21
Unidentified speaker (speaker_0): It's not an open source, it's connected with open standards, but it's not an open source project. The business model is a SaaS for now.
26:31
Unidentified speaker (speaker_0): And I think this might shift in the near future as everything is moving so rapidly and quickly. We are planning to provide a low well low price depending for whom, right? At the moment we are um 49 uh dollars the monthly. subscription
26:51
Unidentified speaker (speaker_0): per user. If you are more than X,
26:55
Unidentified speaker (speaker_0): the price will decrease.
26:58
Unidentified speaker (speaker_0): And then we have
27:00
Unidentified speaker (speaker_0): this is the enterprise. If you want to, for example, create your white label prism for your company, if you want to embed
27:13
Unidentified speaker (speaker_0): the software. within your own tools or bring it into your platform,
27:21
Unidentified speaker (speaker_0): you can do that, or you will be able to do that. So there are kind of two streams: like if you are a small architectural studio or a solo architect and want to use it, I think the best way for now will be to pay the subscription SaaS. However, if you are a larger company, it doesn't need to be too large. I think
27:41
Unidentified speaker (speaker_0): the future is gonna go in this direction. You want to create your own authoring tool, you want to use Prism Stack, then we will enable you to create your own authoring environment within your own platform.
27:56
Unidentified speaker (speaker_0): If that makes sense. So let me give you a hint of how Prism works.
28:03
Unidentified speaker (speaker_0): and
28:06
Unidentified speaker (speaker_0): hopefully shows something interesting. So first the user will
28:13
Unidentified speaker (speaker_0): start with the project location. So we are gonna be looking this time Barcelona which is one of the pilot By lots cities of that I have been working on
28:27
Unidentified speaker (speaker_0): Um
28:28
Oliver Lowrie: Barcelona seems to have lots of mapping information. It's really and also lots of people developing like every time I go onto some cool software like Urban Eyes, for example, it's always Barcelona that they start you in because I think either they've somebody's gathered lots and lots of data on Barcelona or there's loads and loads of people from Barcelona that or who live in Barcelona that think that the AI software is gonna do something
28:48
Oliver Lowrie: interesting.
28:49
Unidentified speaker (speaker_0): Barcelona has a lot of rich data. I was impressed when I started to look into uh compared with other cities, yeah. It's pretty cool.
28:56
Oliver Lowrie: That's it now.
28:57
Unidentified speaker (speaker_0): So at the moment, what we have is like uh we have uh the plan view to the left, um and then the 3D to the side with the real heights uh LOD 100, so it's uh pretty much uh a real footprint, uh real height, but we are not considering root. roofs, etc. for now. This
29:17
Unidentified speaker (speaker_0): might come later. So what we can do is like we can um select a parcel
29:25
Unidentified speaker (speaker_0): that we want to work from.
29:28
Unidentified speaker (UNKNOWN): So
29:28
Unidentified speaker (speaker_0): imagine that we are an architect or a land developer, you are looking for options uh for a new project. If you are in a really early stage.
29:41
Unidentified speaker (speaker_0): And then we can
29:44
Unidentified speaker (speaker_0): start to review the data that we have, cadastral data.
29:49
Unidentified speaker (speaker_0): So, first thing that we can see is of course the context of the buildings. We can then go into a site analysis
30:00
Unidentified speaker (speaker_0): see how the sun impacts the site and how the other contextual buildings
30:06
Unidentified speaker (speaker_0): interact with our site. We have also some other analytical models
30:14
Unidentified speaker (speaker_0): that they are not all completely finished,
30:17
Unidentified speaker (UNKNOWN): but
30:18
Unidentified speaker (speaker_0): you can see the impact of sun and temperature on your site.
30:25
Unidentified speaker (speaker_0): And then we can start reviewing the data that we have access. As I mentioned before, not every parcel has the same data, even in the same city, is something that we are working on.
30:34
Unidentified speaker (UNKNOWN): But
30:34
Unidentified speaker (speaker_0): where you can see the cadastral reference, so you could easily have to go to the cadastral, the address, the area registry, the area front ring. Then you can start getting,
30:46
Unidentified speaker (speaker_0): for example, the maximum height. This is the maximum height for this parcel.
30:52
Unidentified speaker (speaker_0): In this parcel, we get also the setback, so it tells you if we enable the envelope where you can build.
31:09
Unidentified speaker (speaker_0): Parcels, So the perimeter, bonding box, envelope, etc.
31:16
Unidentified speaker (speaker_0): We can also interrogate the site with the surrounding context. So if we go again to site analysis, you can see okay, tell me about what's around the site. And then you can start analyzing what's around. So this is uh office building is blue, then you
31:36
Unidentified speaker (speaker_0): can see retail food, education. hotel, etc.
31:42
Unidentified speaker (UNKNOWN): So
31:42
Unidentified speaker (speaker_0): this is for contextual information only.
31:45
Unidentified speaker (speaker_0): Um and then uh we can start the massing directly here. So uh the UI needs to improve a little bit on this regard, but uh let's give it a try. So you can draw a profile, so you can say, okay, my building, I want it to
32:05
Unidentified speaker (speaker_0): look like the footprint. Of course, I know where I can build in this specific case, but in other cases it might not be like that.
32:16
Unidentified speaker (speaker_0): Um
32:19
Unidentified speaker (speaker_0): the drawing tool where you can define curves, etc. So first thing that we have done is define the footprint.
32:27
Unidentified speaker (speaker_0): and then
32:29
Unidentified speaker (speaker_0): you can define the levels. So, I think the maximum here are six
32:36
Unidentified speaker (speaker_0): here.
32:38
Unidentified speaker (UNKNOWN): And
32:38
Unidentified speaker (speaker_0): now we have the envelope which is gonna tell you which is the square meters that you have per floor. You can then manipulate the data, it will update on demand, um, quite basic. but good enough if you want to do a quick uh site planning. Uh let me remove this aside.
32:59
Unidentified speaker (UNKNOWN): Uh
33:00
Unidentified speaker (speaker_0): we can remove the
33:03
Unidentified speaker (speaker_0): the labels for clarity.
33:05
Unidentified speaker (UNKNOWN): And
33:05
Unidentified speaker (speaker_0): whenever we are ready, we can go and do it ourselves.
33:09
Unidentified speaker (UNKNOWN): So
33:09
Unidentified speaker (speaker_0): then is when we go into prison uh authoring or design. as I call
33:16
Unidentified speaker (UNKNOWN): it.
33:17
Unidentified speaker (speaker_0): Now here we have the envelope still which we might want to turn off.
33:22
Unidentified speaker (speaker_0): Let me see.
33:28
Unidentified speaker (speaker_0): And we have the
33:30
Unidentified speaker (speaker_0): levels envelope, right? Here we can still modify
33:34
Unidentified speaker (UNKNOWN): them
33:35
Unidentified speaker (speaker_0): and access to the to the data in the site panel.
33:42
Unidentified speaker (UNKNOWN): So
33:42
Unidentified speaker (speaker_0): we can still reshuffle them around, and then whenever we we want to move forward, we can start working with the Prism AI chat.
33:53
Unidentified speaker (UNKNOWN): So
33:53
Unidentified speaker (speaker_0): this is a way of doing it for a new design. You can start directly here in the Prism prison
34:00
Unidentified speaker (speaker_0): Authoring Layout. and then you can say, Well, create walls on envelope. And again, this is not using any tokens, so this is initially a rack, means that
34:20
Unidentified speaker (speaker_0): there is an uh engine that enables you to kind of like automate processes by using natural language. It's not uh sophisticated so you cannot ask for anything but it will do the job uh up to a certain extent without the need of uh spending tokens. Although you can also.
34:40
Unidentified speaker (speaker_0): So just briefly, we are in terms of proprietary you we are enable the user to choose which model do they want to work with. So you can put your API key and you will consume as many tokens as you you need and and you will be paying for them.
35:00
Unidentified speaker (speaker_0): Then so now we have the walls now imagine that we want to automate the process of window creation so we can say create
35:08
Unidentified speaker (speaker_0): windows
35:10
Unidentified speaker (speaker_0): on all walls every two meters.
35:17
Unidentified speaker (speaker_0): It will always ask you to confirm, and now we have created all the windows. I don't know if you can see because they are with a white frame.
35:29
Unidentified speaker (speaker_0): Next, imagine that we want to change the world the window type so we can say make
35:36
Unidentified speaker (speaker_0): all windows timber. casement
35:44
Unidentified speaker (speaker_0): and it will
35:46
Unidentified speaker (speaker_0): modify them. Now
35:51
Unidentified speaker (speaker_0): this is a really early stage design. You can move this uh uh as forward far as you
35:58
Unidentified speaker (UNKNOWN): need.
36:00
Unidentified speaker (speaker_0): Let me just change the graphics.
36:11
Unidentified speaker (speaker_0): And well, we have uh live documentation so we can of course see everything.
36:19
Unidentified speaker (speaker_0): Um
36:21
Unidentified speaker (speaker_0): in plan view. We have live elevation elevations so
36:29
Unidentified speaker (speaker_0): we can create um
36:32
Unidentified speaker (speaker_0): elevations on demand. This is a really simple building, but still uh
36:40
Unidentified speaker (speaker_0): um
36:41
Unidentified speaker (speaker_0): quite uh nice to to do. This is great, yeah. We can leave it here.
36:46
Oliver Lowrie: Yeah, I mean it's a sort of interactive design modeling tool. What at what stage would you typically get this? to? I don't know if you like know about the kind of the reverse stages, but it looks like you'd maybe use this for like stage
36:55
Unidentified speaker (UNKNOWN): one,
36:55
Oliver Lowrie: two, and then maybe
36:57
Unidentified speaker (UNKNOWN): you know
36:58
Oliver Lowrie: get it.
36:58
Unidentified speaker (speaker_0): You can go into
36:59
Unidentified speaker (UNKNOWN): detail
36:59
Unidentified speaker (speaker_0): design.
37:01
Unidentified speaker (speaker_0): But uh I yeah, we are developing further and further. Uh sorry, I I was working in the UK, but I don't remember. Detail design is a stage four?
37:09
Oliver Lowrie: Yeah.
37:09
Unidentified speaker (UNKNOWN): Well
37:10
Oliver Lowrie: three, four, yeah.
37:11
Unidentified speaker (speaker_0): So I think you could go three uh maybe four uh not yet, yeah. Um
37:22
Unidentified speaker (speaker_0): hopefully soon
37:23
Oliver Lowrie: interoperable back and forward from Revit so you can go in and out of Revit.
37:27
Unidentified speaker (speaker_0): So you can uh there are uh a few formats that you can bring in and bring out. Um so you can bring in Rhino, Revit,
37:37
Unidentified speaker (speaker_0): but they will not name natively uh be created in Revit although you can bring elements natively in prison. So the interoperability is not there yet where I want it to be in the future.
37:51
Unidentified speaker (speaker_0): But yeah it's moving along.
37:54
Oliver Lowrie: Amazing. So I I've just got one more question. Thank you so much for the demo.
38:01
Unidentified speaker (UNKNOWN): If
38:01
Oliver Lowrie: Prism works the way you want it to, what changes about the way that the architect works?
38:07
Unidentified speaker (speaker_0): The main change is gonna be that the
38:13
Unidentified speaker (speaker_0): unglamorous work goes to the machine. So the redrawing, uh automatic tagging, uh even dimensioning, uh checking, maybe a bit further, uh
38:33
Unidentified speaker (speaker_0): All of that might be part of prism
38:38
Unidentified speaker (speaker_0): job. So I think long story short, what's gonna happen is like decisions is are gonna move earlier. So the architect is gonna get a lot of information earlier on. and it's gonna be able to take decisions earlier than now. So
38:58
Unidentified speaker (speaker_0): instead of discovering some side constraint or some
39:08
Unidentified speaker (speaker_0): issues uh with the context, uh it will uh be able to to get them. earlier and
39:17
Unidentified speaker (speaker_0): as well move faster while authoring in the web.
39:23
Oliver Lowrie: Awesome. Well, thank you so much for your time today, Antonio. And if people want to try the software out, how can they get how can they do that?
39:30
Unidentified speaker (speaker_0): So if you want to try it,
39:35
Unidentified speaker (speaker_0): you can ask for a demo on LinkedIn to my site. myself, or you can just go into prison.so
39:42
Unidentified speaker (speaker_0): and click
39:45
Unidentified speaker (speaker_0): the demo button, and basically we will get in touch with you.
39:55
Oliver Lowrie: Amazing. Thanks so much, Antonio03. That was a real treat. Cheers.
39:59
Unidentified speaker (speaker_0): Thank you Oliver.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights