How Speckle Powers Open Source Construction for AEC Data

The construction industry has a data problem. Not a shortage of data, quite the opposite. Every building project generates enormous volumes of it: design models, structural calculations, cost estimates, schedules, RFIs, change orders. The problem is that most of this data is trapped. Locked inside proprietary software, fragmented across incompatible file formats, and invisible to the stakeholders who most need to understand it.

Dimitrie Stefanescu saw this problem from the inside. Trained as an architect, working on a masterplan for Brussels Noord, he watched an economist build a sophisticated financial model of a project that was supposed to inform architectural decisions, and realized that nothing in his training had prepared him to bridge that gap.

"The data is already part of the design," he says in his Foundamental University masterclass. "But it was very difficult to surface that back up in the language of an economist, or the language of the mayor, or the language of someone running a different feasibility study."

That frustration became the founding insight of Speckle, today one of the most significant open source construction technology companies in the world, and the subject of one of the most candid masterclasses in Foundamental University's Season 1.

Watch the full masterclass here: Commitment To Community

What Is Open Source Construction?

Open source software has transformed industries from finance to healthcare to consumer technology. In construction, the shift has been slower, but it is now underway, and Speckle sits at the center of it.

Open source construction does not mean free construction software. It means something more precise and more powerful: software whose underlying code is publicly available, auditable, and modifiable by anyone. Dimitrie uses a pointed analogy to explain the distinction.

"Open source is the free recipe for the beer itself. You still have to brew the beer yourself, and you bear the costs of setting up your brewery. But you have the recipe."

For the construction industry, historically dominated by closed proprietary platforms with opaque file formats and vendor lock-in, this represents a fundamental shift in how software relationships work. Dimitrie explains.

"The AEC industry is rather boxed in by vendors who lock people into proprietary software and walled gardens, as well as opaque file formats that are very difficult to parse outside the authoring tools that produce them."

"Speckle, by being open source and transparent, was diametrically opposed to that."

The implications go beyond software licensing. Open source construction means that the data generated by a project can be accessed, analyzed, and shared across tools and stakeholders without requiring everyone to use the same software. A structural engineer, an economist, a project manager, and an architect can all look at the same underlying data through whatever lens their work requires, including pairing proprietary systems with an open source architecture tool where it makes sense.

The Founding of Speckle: From Brussels to UCL

Dimitrie's path to founding Speckle is unusual even by the standards of construction technology. Born in Bucharest, trained as an architect in Romania and at TU Delft, he spent years running computational design workshops across Europe before settling in London at UCL around 2015.

The intellectual foundation was laid during his time as an architect in Brussels, working on the Noord masterplan. The project involved an economist whose financial model was supposed to inform architectural decisions, but the two domains were speaking entirely different languages. Dimitrie recalls.

"As architects, we were not well equipped to create shared understandings between, say, someone who does economics and someone who does structural design."

"I realized: it's not about how a building looks. It's much more about how that building translates into things that other stakeholders understand."

He turned this insight into an application for a Marie Curie Fellowship at UCL, an unusual move for someone with his background, and one that gave him the space to develop the ideas he had been accumulating. At UCL, the original Speckle emerged not from a product roadmap but from a desire to escape the tedium of academic literature review.

"I started coding something on the side, a very simple tool for Grasshopper that just instantly shared your 3D model in seconds. That was all it did."

That modest beginning attracted interest from industry partners involved in the research project, including Foster + Partners, Buro Happold, and McNeel, the company behind Rhino. When Dimitrie made the code publicly available, contributions from McNeel and others accelerated rapidly, and a community of computational designers started forming around the project.

From UCL to Arup to Independence

When the Marie Curie research project ended after three years, so did the funding. Arup, which was already using Speckle internally, offered Dimitrie a position to continue developing the project as an open source initiative. Two years there put the platform through its paces at real enterprise scale.

But the limitations of developing an open source project inside a company that competed with other Speckle users eventually became apparent.

"That's why the company was born," Dimitrie says. "Three years of intellectual and practical incubation at UCL, two more years of putting the project through its paces at Arup, and then we decided: we can go much faster and grow this thing much further if we go independent."

Speckle raised its pre-seed from Foundamental, and the company began in earnest. The community that had grown around the open source project became the foundation for the commercial entity, and the trust built through years of transparent, open development became one of its most valuable assets.

What Speckle Does

Speckle describes itself as an open source data platform for AEC. In practice, this means a set of tools and infrastructure that allow construction and design companies to move data freely between authoring tools, from BIM models to structural software to cost estimation platforms, without losing fidelity or requiring manual re-entry.

The platform covers several distinct use cases. At the simplest level, it enables design reviews in a browser, issue tracking, and tracking design changes over time. At a more sophisticated level, it enables companies to aggregate data from multiple APIs, run business intelligence on project data, and increasingly, train AI models on their own historical design and project data.

Dimitrie explains.

"This goes all the way from simply doing design reviews in a browser to vacuuming up data from various APIs and actually training AI models on top of your company's historical data."

"It's not just about helping design in the now. It's also about helping design in the future, keeping track of historical learnings, and making those available both to human beings and to AI."

The platform's modularity is key to how it serves both small teams and enterprise clients. Dimitrie describes it as a set of Lego blocks that can be assembled into standard configurations off the shelf, but also reconfigured for the specific standards, workflows, and requirements of a major capital project.

"When you're building an airport or a major capital project, it becomes its own micro-universe. That's where being a modular platform really helps."

In many firms, Speckle also plays the role of an open source architecture tool that bridges disciplines, connecting authoring environments and helping teams standardize data without forcing a single vendor stack.

Open Source as a Business Model

One of the most instructive parts of Dimitrie's masterclass is his account of how Speckle has navigated the tension between open source principles and commercial sustainability.

The company operates on an open core model: roughly 75% of development time has historically gone into the open source core, with a proprietary layer of features built on top for enterprise customers. The transition toward more active commercialization was handled with deliberate transparency.

"We were transparent with our community from the start, saying that at some point this would happen, and adamant in pushing the message that open source is a free recipe, not a free product."

When the commercial features launched, the reaction from most of the community was pragmatic.

"Most people's reaction was: of course you need to do this, you're a company, we totally understand, let's talk about how I can get access to these features."

Some pushback came, particularly from users who had assumed the platform would always be free. But Dimitrie is clear about what made the transition manageable:

"The key to navigating it was being transparent with our community and setting expectations early. Honesty and trust do pay off."

The commercial model also created a healthier dynamic with customers than a pure SaaS approach would have.

"As a customer, you can always hold us accountable: if you're unhappy with Speckle, you can take your toys, self-host Speckle, and never send a single dollar to the company. This creates a very good, trust-based relationship with our community and our customers."

Community as Infrastructure

One of the most distinctive aspects of Speckle's growth is how much of it has been driven by community rather than sales. The early Slack group that hit the 10,000 message limit and had to migrate to a Discourse forum. The contributions from McNeel that accelerated the platform's early development. The random forum user who turned out to be the CTO of Jacobs, one of the world's largest engineering firms, and who had self-hosted Speckle six months before anyone at the company knew he existed.

Dimitrie says.

"Trust is really important to us, towards our community, towards our clients, and the trust we receive back."

"Speckle being open source and having this bottom-up driven adoption really helped us infiltrate AEC enterprises, both large and small, in a way that a top-down approach would have made into a completely different game."

He describes the community as being like a village that helps raise a child.

"It comes with all the joys of a family, but also all the responsibilities of one. That needs to be handled carefully, articulated and shaped thoughtfully. It goes back to trust. Communication and transparency are key to having this village co-found the company with you."

Hiring Honestly: The Anti-Sell

Dimitrie's masterclass is unusually candid about mistakes, and one of the most concrete concerns hiring. In the early days, the team was overselling the company in interviews: describing work-life balance as amazing when people were actually working days, nights, and weekends; presenting a mature startup culture that did not yet exist.

He reflects.

"That attracted the wrong kind of people."

The fix was what he calls anti-selling: a deliberate moment in every interview process where the team says plainly what Speckle is hard, where it falls short, and what it actually asks of people.

"Depending on the role, we just say: Speckle actually sucks from the following points of view. And then people self-select."

The candidates who stay after that conversation are, by definition, the ones who want exactly what Speckle actually is.

"Being truthful to yourself as a company and as an individual is really important. And communicating it well is key to having a good marathon. This is not a sprint. It's a series of sprints that keep coming every year, every quarter. You cannot last through them if you're not in tune with who you are."

AI in Open Source Construction: An Honest Assessment

Dimitrie's view on AI in AEC is more nuanced than most. He is genuinely excited about the potential and clear-eyed about the current limitations.

On the positive side, AI is already delivering value in parsing large volumes of RFIs, scanning PDF documents, and amplifying day-to-day administrative work. The challenge is that general AI models do not yet understand construction at a deep domain level.

"They have context around AEC but don't really know the heart and soul of it yet."

The more fundamental constraint is one of reliability.

"Buildings need to stand up and not kill people. That's mantra number one. And using AI systems that are right 95% of the time is not feasible in that context, because the 5% of the time they're wrong will still skyrocket your insurance premiums."

Speckle's approach to this problem is to focus on making company-specific data available for AI training rather than building centralized models trained on everyone's data.

"We're helping companies take all their historical projects and models, put them into Speckle, and then stream that data down and start training specific AI LLMs on top of their own data. We believe a core asset of each of these companies is their designs and their data itself."

Velocity Over Consensus: The Cultural Lesson

As Speckle grew from an open source research project into a commercial company with 27 full-time team members, Dimitrie found that the democratic, consensus-based culture that had served the early community well was creating friction at speed.

"About a year and a half to two years ago, I needed to make a cultural shift inside the company: research is nice, being kind to each other is nice, taking decisions together is nice, but at a certain point, velocity matters more than consensus."

His broader lesson on decision-making:

"It's much more important to take a decision fast, even if it's wrong, than to agonize over it in circular discussions."

And the area where iteration matters most is not product but business model.

"It's very easy to iterate on product. The most difficult thing to iterate on is business model. This is where the most value is to be gained from iteration, rather than from pure product iteration."

The Future of Open Source Construction

Dimitrie ends his masterclass on a note of genuine optimism about the moment the construction industry is in. The problems are real and the market is enormous. The digital revolution that passed construction by for decades is finally arriving.

"There has never been a time where so much energy, innovation, and dedication is being poured into the foundations of how AEC gets done," he says. "People are understanding they can achieve much more, and be happier in general, rather than suing each other, which is a favourite pastime in AEC, by actually collaborating and reaching shared understandings."

For Speckle, that means continuing to build the data infrastructure that makes collaboration across stakeholders possible, while positioning the platform to capture and make useful the growing volume of AI-ready data that construction projects generate. In other words, doubling down on open source construction practices while ensuring teams can combine them with the right open source architecture tool where needed.

"Providing them the data and the tools to do so is what gets me excited about the future."

Conclusion

Open source construction is not a niche technical trend. It is a structural shift in how the built world manages, shares, and acts on the data that every project generates. Speckle, under Dimitrie Stefanescu's leadership, has been at the forefront of that shift since its earliest days as a side project at UCL.

His Foundamental University masterclass is one of the most honest and detailed accounts available of what it actually takes to build an open source company in a traditional industry: the community building, the commercial navigation, the hiring mistakes, the cultural evolution, and the long-term vision that holds it together.

The full conversation is freely available here.

Q&A

Question: What does “open source construction” mean, and how is it different from “free software” for AEC?

Short answer: Open source construction means the software’s code is public, auditable, and modifiable—so anyone can inspect it, improve it, and adapt it—without implying the product itself is free to run at scale. Dimitrie’s analogy: open source is the free recipe for the beer; you still have to brew it and pay for the brewery. For AEC, this breaks with vendor lock-in and opaque file formats, letting stakeholders access and share project data across different tools. A structural engineer, economist, PM, and architect can all view the same underlying data through their own lenses, even pairing proprietary systems with an open source architecture tool where it makes sense.

Question: Why was Speckle created, and how did it grow from a research project to an independent company?

Short answer: Speckle began with Dimitrie Stefanescu’s realization on the Brussels Noord masterplan that architects weren’t equipped to translate between domains like economics and structural design. At UCL, on a Marie Curie Fellowship, he built a simple Grasshopper tool that shared 3D models instantly, which drew interest and code contributions from firms like Foster + Partners, Buro Happold, and McNeel (Rhino). After the research funding ended, Arup hired him to keep developing Speckle in the open at enterprise scale. The team then spun out independently to avoid conflicts of building an open platform inside a firm that also competed with its users. Backed by Foundamental’s pre-seed, the community that had formed around the open source project became the foundation of the company’s growth and trust.

Question: What does Speckle actually do for AEC teams day to day?

Short answer: Speckle is an open source data platform that moves information between authoring tools—BIM, structural, cost, scheduling—without data loss or manual re-entry. Teams use it for browser-based design reviews, issue tracking, and change tracking. More advanced use includes aggregating data from multiple APIs, running BI on project data, and increasingly training AI models on a firm’s own historical designs. Its modular, “Lego block” architecture supports both quick, off‑the‑shelf setups and custom configurations for complex capital projects, acting as an open source architecture layer that standardizes data and bridges disciplines alongside proprietary tools.

Question: How does Speckle reconcile open source principles with a viable business model?

Short answer: Speckle follows an open core model: roughly 75% of development historically goes into the open source core, with a proprietary enterprise layer on top. The company was transparent early that commercialization would come, framing open source as a free recipe, not a free product. When paid features launched, most community members supported the move. Crucially, customers can self-host Speckle and never pay the company, which creates accountability and a trust-based relationship rather than lock-in typical of pure SaaS.

Question: What is Speckle’s approach to AI in construction, and what limitations does it recognize?

Short answer: Speckle sees immediate AI value in parsing RFIs, scanning PDFs, and automating admin tasks, but notes that general models lack deep AEC domain understanding and that 95% accuracy is unacceptable where safety and liability matter. Instead of centralized, one-size-fits-all models, Speckle focuses on helping firms centralize their historical project data in Speckle and train company-specific LLMs on their own designs—treating each firm’s data as a core asset while keeping reliability and accountability front and center.

Revolutionizing Construction: Speckle's Open Source Impact