Welcome to the Onshape forum! Ask questions and join in the discussions about everything Onshape.
First time visiting? Here are some places to start:- Looking for a certain topic? Check out the categories filter or use Search (upper right).
- Need support? Ask a question to our Community Support category.
- Please submit support tickets for bugs but you can request improvements in the Product Feedback category.
- Be respectful, on topic and if you see a problem, Flag it.
If you would like to contact our Community Manager personally, feel free to send a private message or an email.
LLM powered Autocomplete
S1mon
Member Posts: 4,116 PRO
As I commented on LinkedIn, "I'm excited to see this step, but isn't this where Copilot was 4 years ago? I'm writing whole features using Codex/ChatGPT with local clones of the standard library and FeatureScript documentation for its reference. As Michael Pascoe commented, we need integration with the latest agents, ideally through MCP alongside the current IDE."
BTW, I tried to sign up for this AI event post-facto and the web form rejected my email. The same one connected to my paid Onshape account. Hmmm.
Simon Gatrall | Product Development, Engineering, Design, Onshape | Ex- IDEO, PCH, Unagi, Carbon | LinkedIn
Comments
I think this link should work for anyone?
https://www.ptc.com/en/resources/corporate/webcast/ai-in-focus-webinar/thank-you
Here's an example of a major SaaS based design tool with MCP integration:
https://www.figma.com/blog/the-figma-canvas-is-now-open-to-agents/
Here's an example of chat driven CAD programming (Grasshopper in Rhino3D):
Simon Gatrall | Product Development, Engineering, Design, Onshape | Ex- IDEO, PCH, Unagi, Carbon | LinkedIn
Ok. Finally watching the video. Perhaps the LinkedIn post would have been much better if it didn't only mention the LLM autocomplete which is the most basic part of the whole thing.
Thanks @eric_pesty for sharing that link.
(Make sure you scroll down to the Onshape AI video, and then jump to about ~11 minutes in where @cody_armstrong starts giving the demos)
We've got AI Quickrender rendering coming (sorry RenderCAD, looks like you might be getting Sherlocked.)
We've got some drawing annotation replication coming (i.e. tell the AI to look at a view on one drawing for part A, and apply that to the second drawing for part B) (sorry https://www.drafterinc.com):
And most important to me at least, AI Agents including one tuned on FeatureScript, and a general MCP connection:
Simon Gatrall | Product Development, Engineering, Design, Onshape | Ex- IDEO, PCH, Unagi, Carbon | LinkedIn
And so we begin to make ourselves redundant….?
Maybe we’ll be CAD project managers and have to oversee a whole team of AI staff to get the job done?
So far, my experience is that with the limited training data available (mostly very proprietary), AI CAD tools that can actually understand things like materials, tolerances, assembly constraints, etc. are very very primitive. There are a lot of demos of things where you can make very basic shapes, but compared to writing, programming and math, AI ability in mechanical engineering is a lot further away.
That said, the demo of the "model assistant" in the video linked from PTC/Onshape is interesting. Most of those future demos in the video seem very scripted and perhaps smoke and mirror (fine for directional stuff), but I believe they're at the level of stuff that is possible. The idea of, I need to make a bunch of simple changes to this part. I could theoretically give that task to a junior engineer while I do something more complicated, or I could use this agent. Yeah, sucks to be a junior engineer at that point, but the reality is that if we don't all get good at using better and better tools, we will lose out to people and companies that are, You can't sit around and use files and hand drills while others are using 5-axis CNC machines and expect to compete in the same space.
Simon Gatrall | Product Development, Engineering, Design, Onshape | Ex- IDEO, PCH, Unagi, Carbon | LinkedIn