This is Engineering Paper, and here’s the latest design and simulation software news.
Let’s start with an update from Leo AI, the software startup building a mechanical engineering copilot. According to CEO Maor Farid in a LinkedIn post from last week, Leo can now generate full CAD assemblies from a text prompt.
In a video accompanying Farid’s post, the example prompt “Design a hydraulic piston that pushes 50 kg. stroke: 100 mm” along with the passionate instruction to “Prioritize using parts from my inventory!!” shows Leo doing design calculations, searching through both internal and online parts, and putting together a CAD assembly. In an email, Farid told me that Leo’s assemblies have a full part and feature tree and are compatible with Solidworks, Onshape, Catia, and Inventor.
The video skips over some human steps—for instance, Farid noted in his LinkedIn post that “Parts aren’t mated automatically yet – you still need to close the mates manually. We’re fixing that.”
Still, Farid is confident that Leo’s new feature is a significant step forward for AI-assisted engineering.
“This is the first time an AI has closed the loop from idea to manufacturable assembly, grounded in an organization’s own knowledge base,” Farid wrote in the email. “Engineers at Nvidia, HP, and Intel are already using it.”
Hoops AI launches… again
Tech Soft 3D, a developer of engineering software development kits (SDKs), has announced the full release of HOOPS AI, its AI framework for CAD data.
“Built for data scientists and machine learning engineers, HOOPS AI unifies CAD access, dataset preparation, and encoding, providing the fastest path from raw CAD data to production-ready machine learning models,” reads Tech Soft 3D’s press release from November 2025, the first time the company “officially” launched HOOPS AI. I covered it here.
So what’s changed? Tech Soft 3D is calling this the “full release” of HOOPS AI and now describes the previous release as “a successful beta program involving more than 30 companies.”
Along with general availability, HOOPS AI now supports Linux and CAD embeddings, which Tech Soft 3D describes as “a technology that captures the semantic meaning and relationships within complex CAD data, allowing machine learning modules to identify similar parts and understand the relevant design context.”
Quick hits
- CAD developer Kubotek Kosmos has released an update for its model-based definition (MBD) file utilities software. Update 8.0 of the Validate, Revision, Convert, and View programs adds support for saved cutting planes, controlled unclassified information (CUI) markings, and new versions of popular CAD file formats.
- Semiconductor designer Arm has announced the AGI CPU, the first chip it will manufacture itself rather than licensing the architecture to chipmakers like Nvidia. Simulation developer Synopsys was quick to point out that Arm used its EDA tools in the process, with Synopsys’ Ravi Subramanian saying in a press release that “we’re proud that Synopsys design, IP, and advanced verification solutions played a mission-critical role in delivering this innovation.”
- Neuromorphic computing company Innatera has announced Synfire, an open repository for neuromorphic models in the Neuromorphic Intermediate Representation (NIR) format. I bring this up mostly as an excuse to link to this article I wrote a few years ago on neuromorphic computing, a topic which deserves more attention: Is Neuromorphic Computing the Future of AI? A Look at Intel’s New Loihi 2 Chip.
One last link
The third episode of Travel for Engineers has landed. Join Design World’s Paul Heney and Lisa Eitel in Airplanes and airports and face masks, oh my!
Got news, tips, comments, or complaints? Send them my way: malba@wtwhmedia.com.