How Speckle Used Open Source to Crack AEC: Inside Dimitrie Stefanescu’s Playbook
Speckle started as a side project inside a UCL research grant and grew into a venture-backed company with a $12.5M Series A, used by Arup, Mott Macdonald, Aurecon and Jacobs. Founder Dimitrie Stefanescu argues open source was the only way to break into an industry locked into proprietary file formats. Here is what the journey teaches AEC executives, ConTech founders and investors about trust, data ownership and the open core playbook.
Most construction software stories follow a familiar script. A founder spots a gap, raises money, builds a closed product, charges a license fee, and spends years convincing skeptical buyers to switch. Speckle did almost none of that.
The company gave its core platform away, made the code public, and let competitors of its own customers self-host the same software. In the process, it quietly got into Arup, Mott Macdonald, Royal HaskoningDHV and Jacobs before raising a $12.5 million Series A led by Addition in October 2024.
In a long conversation on the Bricks & Bytes podcast, founder and CEO Dimitrie Stefanescu walked through the full arc. The research grant origins. The slack group that hit its message limit. The pivot to an open core model that some community members initially called evil. And the broader argument that AEC data, the actual designs firms produce every day, should not be held hostage by the authoring tool that created it.
For AEC execs, the question is whether your firm controls its own design data or rents it back from a vendor. For founders, Speckle is a real template for selling into a notoriously conservative industry. For investors, it shows that open source plus community can be a viable go-to-market in a sector most assumed required heavy enterprise sales.
From recovering architect to open source founder
The data problem an architect noticed first
Stefanescu calls himself a recovering architect. He studied in Bucharest and Rotterdam, taught computational design across Europe, and landed at UCL in London in 2015. The pivot moment came earlier, in Brussels, while working on a master plan north of the city. An economist was building an economic model alongside the design, and the architects had no real way to share their data with him.
In modern AEC, every project involves a dozen stakeholders with different software and file formats. The economist needs floor areas. The structural engineer needs loads. The sustainability consultant needs material quantities. Today, getting clean data between them means exports, re-exports, manual cleanup and frequent errors. Stefanescu’s bet was that the data already existed inside every floor plan and BIM model. It just had to be liberated from the authoring tool that produced it.
Why open source was a feature, not charity
How transparency became a moat in a closed industry
Speckle went open source for an unromantic reason. UCL had political tangles around who owned the IP, and making the code public sidestepped that. But Stefanescu also believed an industry boxed in by proprietary file formats would respond to a tool that did the opposite. He was right. Within a few years, Speckle was being used inside Jacobs, one of the largest engineering firms in the world. Stefanescu only found out months later, when a random forum user turned out to be the firm’s CTO.
Open source means the free recipe for the beer itself. You still have to brew the beer yourself.Dimitrie Stefanescu, founder and CEO, Speckle
Open source is not a synonym for free. It is a synonym for transparent and portable. If you self-host Speckle, you owe the company nothing. If you want managed hosting and enterprise features, you pay. The threat of leaving keeps the company honest. The convenience of staying keeps the revenue coming.
The open core pivot, and the moment the community got angry
Building proprietary features on top of free code
Around 18 months into the company’s life, Speckle started building proprietary features on top of its open source core. Stefanescu cites the funny phenomenon where a closed company open sourcing something is celebrated, while an open source company closing anything is treated as betrayal. The way Speckle handled it is worth studying. From the start, the team told the community commercial features were coming. When they arrived, the message was already set. Most reactions were of course you guys need to do this, you are a company. The trust paid off.
Hear the full Speckle founder story
Dimitrie’s complete masterclass, recorded with Foundamental, covers fundraising, hiring, the open core pivot and what he would do differently. Free to watch.
Watch the MasterclassWhat this means for AEC executives
Stop renting your designs back from your vendor
If you run a construction or engineering firm, Speckle raises an uncomfortable question. Who owns the data inside your BIM models? Legally, you do. Practically, it sits inside a proprietary file format only one vendor can read fluently. If that vendor raises prices or gets acquired, your historical project data is suddenly a liability. This is the same problem covered in why 95% of construction data goes unused. Stefanescu’s enterprise customers are already streaming their historical models into AI training pipelines to build firm-specific LLMs on their own design history. That use case is only possible because the data was liberated from its authoring tool first.
What ConTech founders should steal from this playbook
Anti-selling, community, and iterating on business model
Stefanescu offered three pieces of negative advice worth more than most positive advice. First, anti-sell to candidates. Speckle used to oversell itself as a mature, work-life-balanced startup. It was not. Now every interview includes a section on what sucks about Speckle. Fit goes up. Second, iterate on business model harder than you iterate on product. Pricing and packaging are where most founders are too scared to touch. Third, any API surface you expose is a feature, even if undocumented. Taking it away will annoy someone. Plan for it.
- Set the open core expectation early. Tell your community on day one that commercial features will arrive. The reaction is determined by what you said years before.
- Anti-sell in hiring. Surface the real difficulties of working at your company. The right candidates self-select after that conversation.
- Treat business model as a product. Pricing, packaging and unit economics deserve as many iteration cycles as the feature roadmap.
Open source versus closed source AEC platforms
How the two models compare in practice
For executives weighing whether to back open source AEC infrastructure or stick with traditional vendors, the trade-offs look something like this.
| Dimension | Open source (e.g. Speckle) | Closed source incumbents |
|---|---|---|
| Data portability | Self-host, export, walk away anytime | Locked to vendor file formats and APIs |
| Pricing leverage | Customer holds the exit option, keeps vendor honest | Vendor sets terms, switching costs are very high |
| Customization | Source code is public, internal teams can extend | Limited to what the vendor exposes via API |
| Total cost of ownership | Free core, paid managed and enterprise tiers | Per-seat licensing, often growing year over year |
| AI readiness | Native access to structured object data for training | Depends entirely on vendor’s AI roadmap |
The bigger picture for AEC investors
Why open source plus community is a venture-scale GTM
Five years ago, most ConTech investors would have told you open source was a poor fit for AEC. Speckle’s Series A, alongside the broader investor interest in vendor-agnostic data infrastructure flagged in a16z’s AEC investment thesis, suggests that view has shifted. The reason is bottom-up adoption. Speckle did not need to win an enterprise RFP to get into Jacobs. A single technical user did. That motion is hard for closed source companies to replicate. They cannot offer the free download that starts the loop.
Speckle is an open source data hub for the architecture, engineering and construction industry. It connects to authoring tools like Revit, Rhino, Grasshopper, AutoCAD, ArchiCAD and Blender, breaks 3D models into structured objects rather than files, and lets teams version, share, automate and analyze that data across software. (Source)
Speckle raised a $12.5 million Series A in October 2024, led by Addition. Previous investors Foundamental, Frontline Ventures, Matrix Partners and Foreword Ventures all participated, along with dbt Labs founder Tristan Handy. (Source)
Open core means a core product is released as open source and can be self-hosted free of charge, while the company sells proprietary features, managed hosting and enterprise support on top. Speckle followed this pattern, and its leadership argues the transparency builds the trust that closed competitors cannot match. (Source)
Public customers include Arup, Mott Macdonald, Aurecon, Royal HaskoningDHV, Bollinger+Grohmann, Jacobs, Herzog & de Meuron, Gensler, Multiconsult, Stantec and Suffolk Construction. The company describes thousands of individual engineers and developers as active users across these firms. (Source)
AI models trained on AEC data need structured access to historical designs. Most BIM data sits inside proprietary file formats that are difficult to parse at scale. Speckle decomposes models into structured objects, which means firms can stream their own design history into training pipelines and build internal LLMs on top of their own intellectual property rather than handing data to a third party. (Source)
Stefanescu studied architecture in Bucharest and Rotterdam, taught computational design across Europe, and arrived in London in 2015 as a Marie Curie Fellow at UCL. He started Speckle as a side project during that fellowship, continued development at engineering firm Arup as a Senior Programmer, and co-founded the company with Matteo Cominetti in 2020. (Source)
Speckle’s $12.5M Series A and its enterprise traction at firms like Jacobs and Arup is the clearest evidence yet that the model works in AEC. The key is that open source is not a pricing strategy, it is a distribution strategy. The commercial product sits on top, and the trust earned through transparency makes the enterprise upsell easier, not harder. (Source)
Related Articles
Data Analytics is Revolutionising Construction Management
The VC Firm Behind Facebook Just Told Construction: Your Software Is Next
Why Pilots Fail Between the Starting Line and Scale
