You’re reading Engineering Paper, and here’s the latest design and simulation software news.
Last week I covered updates from 3DExperience World 2026, Dassault’s Systèmes’ user conference that took place in Houston, Texas (next year it’ll be in Nashville, Tennessee).
A few follow-ups and updates are in order. Firstly, I promised more info on the AI features that were announced for Solidworks. I spoke with Manish Kumar, Solidworks CEO, and Craig Therrien, senior product manager for Solidworks, to learn more. I’ll publish those details on Engineering.com soon.
Meanwhile, Kumar also answered my questions about Dassault’s trio of virtual companions: Aura, Leo, and Marie. These AI agents have distinct specialties and personalities, but I wrote last week that I didn’t understand why there are three of them, nor how users would choose between them. As it turns out, there kind of aren’t, and users kind of won’t.
“It’s not three different intelligences. It’s the same intelligence. What is different is the context, persona and constraints,” Kumar told me.
Those personas are split by scope and agreeableness. Aura is “very agreeable,” according to Kumar, “and she is going to explore everything” without worrying about whether it’s realistic. Leo is more assertive, and acts like an engineer—“he only gives you ideas which are manufacturable.” Marie, the scientist, is more assertive yet. She “has to give ideas which are scientific in nature.”
Since they’re based on the same underlying AI, each of these personas is essentially a pre-defined prompt that specifies the nature of their response. Something like, “Answer from the perspective of an engineer concerned with manufacturability.” I imagine it’s more sophisticated than that behind the scenes, but that’s the basic idea.
So how do users choose between Aura, Leo, and Marie? For one thing, each persona will have a visible icon, so you always know which one is responding. But Kumar explained that you don’t really need to worry about picking one over another.
“When you start a chat with Leo, or Aura, whatever, if it needs to get the answer from the other persona, it will switch automatically. You as a user, don’t have to do it,” said Kumar. “Having said that, anytime you use a multi-agentic model, knowing what agents are working for you is important.”
Each of the virtual companions will also have what Dassault calls competencies. For Leo, the engineer, competencies might include mechanical design, sheet metal design, systems engineering, etc. Each competency is further composed of what Dassault calls skills. For example, the mechanical design competency could be divided into skills for conceptual sketching, detailed design, drawing creation, and the like.
“As these companions are released, we are also going to release more and more competencies, and we are going to release each competency with a certain number of skills,” Kumar said. He added that the skills will be continually refined and new skills added over time.
My final question on the virtual companions was how they’d be licensed. Kumar told me Dassault is planning a consumption-based model that will vary depending on the task. Asking Leo to generate a steel structure to store a water tank will cost more than asking him to generate a drawing.
The licensing model will involve three distinct types of credits that Dassault is calling units of knowledge, units of know-how, and units of work. Units of knowledge account for licensing and royalty fees for standards, textbooks, and other training data; units of know-how account for the cost of extracting and indexing process and design data; and units of work account for computational cost.
“The combination of those three is going to define how much you will consume, and we will convert it into something which is simple to understand,” Kumar said.
That consumption-based licensing model will be live in July, which is when Dassault plans to release Leo. Marie is scheduled for release in September. Aura is currently live and will switch to the new licensing model when Leo is released.
A correction on 3DExperience confusion
Kumar drew my attention to an error I made in my original coverage of 3DExperience World where I misquoted Dassault Systèmes CEO Pascal Daloz stating they’d fixed six confusing things about the 3DExperience Platform (referencing Peter Brinkhuis’ article from last year, 37 things that confuse me about 3DEXPERIENCE).
Daloz actually said “we fixed things,” not “we fixed six.”
Kumar told me that there have actually been 26 fixes to date, and that all of Brinkhuis’ complaints will be addressed by July, not including the complaints pertaining to matters of preference. For example, number six on Brinkhuis’ list was Dassault’s overuse of capital letters. 3DEXPERIENCE, SOLIDWORKS, SIMULIA, and the rest of the brands will continue to shout their names.
AI on the 3DExperience World show floor
Dassault wasn’t the only one showing off AI at 3DExperience World. The show floor—aka the Playground—was full of exhibitors touting their own AI experiences and offerings.
Engineering.com contributor Erin Winick Anthony roamed the Playground and documented her findings in 5 Fascinating AI applications at 3DExperience World 2026. Check it out to learn how AI is being used in 3D screens, bionic hands, pro skate sharpeners, generative design and more.
Quick hits
- Mastercam has updated its eponymous CAM software to version 2026.R2. Available now, Mastercam 2026.R2 adds GPU-based simulation and a new AI copilot that, in addition to providing product support, can also directly adjust feed rates and spindle speeds. Mastercam separately announced that it has completed the acquisition of MLC CAD Systems’ CAM business.
- Precise Simulation has updated its FEA and CFD simulation software, FEATool Multiphysics, to version 1.18. The release includes a redesigned user interface that Precise Simulation says is 10 times more responsive, on average, than the previous interface.
- Siemens announced that it has acquired Canopus AI, a software developer focused on AI-based wafer and mask metrology and inspection. “By combining the computational lithography and manufacturing physics simulation capability in our Calibre portfolio with Canopus AI’s advanced metrology and inspection technologies, we are creating a differentiated, end-to-end EDA digital thread that improves the fidelity of printed wafer patterns, accelerates yield ramp and reduces time-to-volume for advanced nodes,” said Tony Hemmelgarn, CEO of Siemens Digital Industry Software, in the company’s press release. The acquisition closed in January with undisclosed terms.
- IMSI Design has released DesignCAD 2025.1 to improve the stability and compatibility of the 2D and 3D CAD software.
- JuliaHub has launched Dyad AI, an agentic engineering platform for product development. “Most agentic tools stop at producing syntax. Dyad AI engages equations, constraints, and physical laws, integrating simulation, parameterization, performance testing, and automated calibration so agents can co-design systems grounded in real physics,” said JuliaHub CEO Viral Shah in the company’s press release.
- Cadence has launched ChipStack AI Super Agent, a new tool for front-end silicon design and verification. The AI “provides up to 10X productivity improvements for coding designs and testbenches, creating test plans, orchestrating regression testing, debugging and automatically fixing issues,” according to Cadence.
One last link
When was the last time you read a magazine? There’s something nice about flipping through thoughtfully designed pages and perusing interesting stories, especially with a cup of coffee at hand. Try it yourself with Design World’s February 2026 issue (it’s online, so you don’t have to worry about coffee stains).
Got news, tips, comments, or complaints? Send them my way: malba@wtwhmedia.com.
The post A tale of three chatbots: Aura, Leo, and Marie appeared first on Engineering.com.
