Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Hestus – AI Copilot for CAD (hestus.co)
224 points by kevinsane 68 days ago | hide | past | favorite | 87 comments
Hello! We’re Kevin and Sohrab from Hestus (https://www.hestus.co). We're working on an AI copilot for CAD. Today we're releasing a simple sketch helper for Fusion 360 and would love your feedback. Here’s a quick demo: https://www.youtube.com/watch?v=L9n_eY-fM_E.

Why we’re doing this: Mechanical engineers excel at generating initial design concepts but get bogged down translating ideas into final designs due to tedious, repetitive tasks. Our goal is to automate these mundane processes, allowing engineers to focus on the creative aspects of design.

Having worked at multiple hardware companies—from medical devices to space launch vehicles—we know how often “trivial” components such as manufacturing rigging, get brushed under the table in scheduling conversations. These tasks aren’t necessarily complex, but they take time and still require the rigor of production components. From finding the perfect fastener to making sure mounting holes align, we aim to simplify and accelerate the design process from the complex to the mundane.

We're tackling this problem similarly to how coding copilots help programmers work faster. Initially, rudimentary coding assistants offered simple suggestions like auto-completing variables. Now, they understand complex tasks, write entire code blocks, and help fix bugs. We're taking this step-by-step approach, starting with a beta that focuses on sketching.

Our sketch helper offers design suggestions, such as applying equality constraints to similarly sized circles or adding tangent constraints between lines and curves. While designers can do these tasks manually, they often require dozens of precise mouse clicks. Our software makes suggestions that you can preview and accept to streamline your workflow. Over time we aim to improve at anticipating your needs and expand beyond sketching to other design aspects like resolving interference issues, auto-generating bills of materials with purchase links, and offering manufacturability suggestions.

How this is different from other solutions: we've heard of complete generative part design solutions, but we don't believe this top down approach is the best way to assist mechanical engineers. Engineers excel at and enjoy designing new concepts—we want to focus on streamlining the most tedious aspects. Crucially, we find that generative solutions often overlook manufacturability, a key aspect of design.

We invite you to try our sketch helper and share your thoughts! If you can think of any additional features that would make it more useful to you, we’d love to hear what they are. Any and all feedback is welcome!




I've been using parametric CAD (Pro/E / Creo, Solidworks, and Onshape) for 30+ years. Automatically adding sketch constraints is nothing new. Just about any 2D sketching tool out there will have some sort of snapping and/or auto-constraint-adding system. What's often more frustrating is fixing things when they get accidentally over-constrained. Solidworks tries to help with this in the form of cycling through what it thinks are the most likely constraints to remove and then re-solving the sketch. It's OK, but that sort of tool could be better.

A number of popular CAD systems use the D-Cubed 2D sketch constraint solver [0]. Siemens owns this and the Parasolid kernel, along with NX. All have been in constant development since the 80's. I really question what major new problems a startup is going to fix in 2D sketching constraints. I'm sure there a bunch of small quality of life things which may be out there, but most of the hard issues are more 3D or spline related, not finding things which could be tangent or equal.

Probably the biggest paradigm shift with constraints that still hasn't really taken off is what Siemens is doing with SolidEdge. It allows for defining the 3D equivalent of constraints between surfaces, holes, edges, etc. and then using direct modeling techniques to modify solids. Perhaps adding more intelligence to that approach would make direct modeling more popular.

Onshape has innovated in the way that it's brought Google Docs-like collaboration and GITHub like versioning, branching and merging to parametric CAD. Nothing else has these capabilities at the moment. To me that has been one of the most innovative changes in the mechanical CAD industry.

Onshape also has FeatureScript, which is the programming language which describes all the parametric features. Right now, none of the LLMs know FeatureScript well enough to be the least bit useful. They hallucinate wildly. I'd be very happy to have a Copilot for FeatureScript.

[0] https://plm.sw.siemens.com/en-US/plm-components/d-cubed/


> I've been using parametric CAD (Pro/E / Creo, Solidworks, and Onshape) for 30+ years.

True - I implemented the first Solidworks Autodimension sketch as a contractor around 2002 based upon earlier work that I'd done for a consultancy client at D-Cubed in the late 1990s. I'm sure it could be improved with AI and a large data set of sketches though.

> Solidworks tries to help with this in the form of cycling through what it thinks are the most likely constraints to remove and then re-solving the sketch. It's OK, but that sort of tool could be better.

I agree it could be better. The behavior with under-constrained sketches depends on D-Cubed's DCM, and I seem to recall they were rather floppy. It seems kind of ridiculous to make users jump through the hoops of making sketches fully constrained once they've added the constraints they care about.


I am not an expert but surely an under constrained sketch is not a completed sketch.

Does "fully constrained" mean the simplest set of constraints that yields a shape (volume/hypervolume)? Or something rather more complicated? A simple yes or no, with a pointer to a paper will do!


To add to delhanty's reply, a "degree of freedom" can be thought of as a dimension of the drawing that can be changed or "stretched" or moved without violating a constraint (this is slightly inaccurate, but it's a good start). In a CAD program, a fully constrained drawing can't be freely stretched or dragged around; the program won't let you and the drawing will feel "rigid".

It's very intuitive if you play around with a CAD program for a bit. There is a free (GPLv3) 2D and 3D CAD program called Solvespace (https://solvespace.com/) that is probably easiest one to obtain and learn. There are detailed tutorials on the website, and you could probably download it and finish the first tutorial in an hour.


In addition to Solvespace there is the nascent (but surprisingly polished):

https://dune3d.org/

the Github page of which has the following footnote:

>I ended up directly using solvespace's solver instead of the suggested wrapper code since it didn't expose all of the features I needed. I also had to patch the solver to make it sufficiently fast for the kinds of equations I was generating by symbolically solving equations where applicable. ↩

Which really impressed me because it was the first graphical and interactive 3D program I tried which felt sort of comfortable and understandable (which is why I mostly use OpenSCAD and similar programmatic approaches).


In Fusion 360, you can drag underconstrained parts of the sketch around with the mouse. So the “degree of freedom” is still specified, but by accidental placement of how you drew the sketch. Fully-constrained sketches show a black outline and can’t be changed without explicitly editing a constraint or dimension.

For a quick example, see slide 10 in https://docs.google.com/presentation/d/1TiVoKF7mLlF360Oxy8hN... This is just a quick doc I made to show a friend how to use Fusion 360.


"fully constrained" means that there are no degrees of freedom left so that there are only a finite number of valid solutions, which are then consequently disconnected in the space of possible solutions.

DCM then chooses one valid solution that it believes is "close" to initial supplied positions/directions/radii etc for the geometry.


i love onshape and the capabilities featurescript offers. I wish they had a hobbyist pricing that only provided minor QOL offerings like the ability to make some things private. As long as it remains free for public use though I am very happy using it.

I was wondering if AI might help with writing FeatureScript. I’ve only scratched the surface but have leaned heavily on what other people have written for documentation.


Your experience, I do not doubt. However, I believe you misrepresent what the authors are claiming. They do not claim that constraints are new, rather that setting all the constraints manually can be tedious, and tools that intelligently select sets of constraint options can be useful at the practical level.


Many of the existing tools will automatically add constraints as you go. Certainly the tangent, horizontal and vertical constraints can be done while sketching with a tiny amount of care. Some systems will look for other circles and try to set equal diameters. Most systems have shortcut keys which can be used to quickly set missing constraints - taking a similar number of clicks as the UI shown in the demo video. I hope there are cases where their tool is faster or smarter, but it wasn’t evident from the demo.


Agreed, sync modelling in Solid Edge and NX can be really powerful, especially when requirements change unexpectedly.


Was a Mechanical Design Engineer for about a decade. Personally, the CAD part was the easy and fun part (Used Solidworks, Creo, Inventor, Catia v5 extensively. Professionally trained in all four as well). The “tedious” part was SAP (part number and material management), drafting, ECN/R/O, BOM handling, etc. I would be more excited for “AI” for PLM like Windchill, TeamCenter, or Cadedge (or whatever SAP PLM packages you used)


Echo this. As an EE fun part is circuit design and the tedious part is SAP and agile. This is where AI needs to be. Parts creation from data sheets, and automatic sync between SAP and Altium/cadence.


I've been a Mechanical Engineer for a similar length of time, using Solidworks and Creo, and I concur. I've released many thousands of parts over the course of my career, and I don't think of the modeling itself as a major time sink (provided you have some decent training, and are using a thoughtful approach and not fighting the software). If the AI copilot could actually do all the modeling based on prompts, that would be pretty interesting, especially if it could create a feature try that I could tweak.

Automatic drafting is something that I think might be a good target for doing some AI research on. Your prompt exists in the form of the solid model, and it's feature tree and parameters. The output is the lay out views and annotations. It's okay if the output isn't perfect, as I would expect to be reviewing it and doing some tweaks.

Helping with all of the ERP processes involved with releasing and maintaining engineering documentation would be a HUGE time saver. If I could ask a copilot program to start a change request, and give it some basic descriptions of what I'm changing, it would be massively helpful if it could start pulling relevant files, and auto-filling the right requests/forms/whatever to do that process.


Hey! We're building this. It's a workflow driven automated documentation system for engineers based on your document templates. We started in energy and have expanded into engineering services. Would love your feedback.

fast-draft.ai

Please, please send me an email or dm if you are interested.


We are on the same page. We absolutely want to help out with most of those tasks as well. We especially want to help with part selection and BOM creation. However, material/inventory management I will admit, wasn't on our roadmap. Although absolutely important, we felt that there were many solutions already existing in that space. If I am wrong, I would love to chat more! Please ping us on the contact us page, and I'll try to get back to you!


What do you do now? Did you switch to software?


Yup, had the opportunity to switch to full time software a couple years back so I jumped at the opportunity.


I'm looking to do the same. What did you jump to, and how did you make the switch?


I’ve been debating writing a book about this to enable hardware to jump to software. I work on a bunch of infrastructure related code (make sure everything is running correctly). As for the switch, it takes a bit of luck and hard work. Luck wise, my employer has an opening and I happen to fit the need. Hard work wise, you still need to prove your worth. I was coding about 50/50 in my previous role so it was easy to sell my case. I went on some interviews lately and it’s important to know the game. Being able to Leetcode and system design is for the better or worse the de facto standard if you want to land a SWE role.


I've used CAD nearly every day for my entire engineering career, and this feels like a logical step. My day is filled with a tedium of micromanaging Solidworks or NX to simply do what I need. It takes very little time for me to envision a solution, but it takes ages of clunking through UI to actualize it. I'll be on board once you have SW support, I'd be interested in implementing it immediately.

My questions: 1) What's your CAD background? 2) What led you to realize that this was a problem that needed solving?


Sohrab and I have been in deep tech for most of our careers. For me personally, most of my background focused on system integration and design. Working closely with the mechanical engineers, I encountered multiple just annoying paper cuts that I felt could be solve with software. We hope that we can create software where it is just easier and faster to prototype and test things in real life. This release was our attempt of a first baby step in that overarching goal


That's neat.

Autodesk Fusion already knows how to apply those kinds of constraints. This is an easier way to input them. The general problem is to express "what am I pointing at" in cluttered situations. Just figuring out how to do that was a huge problem in CAD. For decades, you had to have multiple views on screen just to select. The UI for doing it in one view was really tough.

It still doesn't translate well to fat-fingered devices such as tablets and phones. There are lots of construction people who could really use a tablet that has the design, knows where it is in space, and shows what's supposed to go there.


I wonder if the prevalence of coding LLMs will bring renewed attention to code-based modeling systems like OpenSCAD. It seems much easier to get them to generate code vs. translating into GUI interactions or modifying some other internal state directly.


I’m sure there will be some value, especially as the LLM systems become capable of doing more of the work, but as a professional robotics engineer with a focus on mechanical engineering, OpenSCAD just cannot perform appropriately to serve as a primary CAD platform for professional mechanical design. Parametric GUI based design has so many advantages, a key one being that you can click on a feature and change it easily.


I think both approaches have merit, I find the ability to have proper libraries/functions (such as "place M4x8 socket head countersunk bolt here") is really nice, though I'm faster with the click&change approach for exploratory design.

However the thing holding OpenSCAD back is the fact it is CSG (basically booleans on primitives) which is just not good enough for non-trivial parts. More interesting tool is cadquery[1] which uses the OpenCASCADE b-rep kernel. Still not as powerful as commercial offerings sadly, but at least on the right path to get there.

[1] https://cadquery.readthedocs.io/en/latest/


I've been using build123d (based on cadquery) quite a bit and I really like it. I still use solidworks for simulations, drawings, and most fillet operations - because it turns out that doing fillets right is very non-trivial.

Code-cad is great because you get exactly what you put in, but sometimes it can feel like starting with a bucket of logic gates instead of a microcontroller.

I've had "test LLM's with code-cad" on my todo list for a while, as I think it has the possibility of greatly accelerating the production of the additional part libraries that I'd need to able to make more regular use of it.


Been trying to install cadquery on a mac (M1) but ouch, not easy, and not working yet.


FWIW conda-forge packages cadquery and it's dependencies, including on macOS-arm. So micromamba/mamba/conda can easily be used to install it. build123d is being worked on.


Great, thanks! I'll give it a try. Brew -> pip -> cadquery was a dead end for me.


Got it running, thanks again!


I've been wondering the same thing. Unfortunately, I've found that GPT-4o is much less good at OpenSCAD than it is at Python.

It feels totally untrained on the step-by-step logic you need to build things in a programatic way. This is likely a problem with not having enough training data.


There is a version of OpenSCAD which adds support for Python:

https://pythonscad.org/

That said, the bigger problem is that OpenSCAD is easy for things which are tractable to being described mathematically, and that the boundary for what can be described is one's fluency with mathematics --- once one gets beyond those things which can be described by rectangles, cylinders, and spheres, it gets quite difficult, and getting elements aligned often requires trigonometry (I have one backburner project which needs for me to get up-to-speed on conic sections).

There are some folks who are able to write code to generate point clouds or polygons which describe surfaces (see recent discussions on the OpenSCAD mailing list), but not many, and there isn't much such code, and when it does exist, it tends to be quite special purpose and hard to apply to different shapes.

It would help if there was a Bézier curve primitive in OpenSCAD, or better some support for a NURBS surface as a core primitive so folks wouldn't constantly be rolling their own options.


Surely all of GitHub is “enough training data”


Well, there's a lot less OpenSCAD on GitHub than Python. At first glance maybe a few orders of magnitude less.

But the bigger problem is (probably) not the lack of code, but the lack of good tutorials and well-documented libraries. And since there's no package manager (https://github.com/openscad/openscad/issues/3479) there's still no agreed-upon way to reuse code between projects.

All this is conjecture, of course. LLMs are mysterious beasts.


OpenSCAD is frequently written on the same level as M4 macro language, or assembly code with no comments. If you're bad at mental arithmetic and bad at keeping track complicated internal state, like LLMs are, you will have a bad time.

Look at the one linked below, for instance. This is from the official OpenSCAD repo, examples folder. There is no way for even an experienced programmer to understand and successfully modify that code without sitting down with a pencil and paper and doing arithmetic and geometry for half an hour.

https://github.com/openscad/openscad/blob/master/examples%2F...


The low amount of training data is definitely one issue, but I also think back to the SCAD I've written and it's not got very much in the way of English prompt-friendly text in it to "guide" the LLM towards using it.

Arguably, this might mean I don't comment the code "enough", but in other languages, the variable names and overall structure seem to carry a lot of that for the human programmer and also in a way that guides an LLM towards effectively using it. I don't know much about how LLM's work internally either, but I picked a random SCAD file of mine and I don't know how it'd be found to create a PCB spacer with generous clearance for an M3 screw.

  $fn = 90;
  
  outer = 4.5 + 1.6;
  inner = 4.5;
  h = 13.5;
  
  //round = true;
  //square = true;
  hex = true;
  
  if (square)
    difference() {
      cube(size = [outer, outer, h]);
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          
      translate([(outer-inner)/2, (outer-inner)/2,-0.05]) cube(size = [inner, inner, h+1]);
    }

  if (round)
    translate([0,15,0])
      difference() {
        cylinder(d = outer, h = h);
        translate([0,0,-0.05]) cylinder(d = inner, h = h+1);
      }

  if (hex) translate([0,30,0])
    difference() {
      cylinder(d = outer*1.25, h = h, $fn=6);
      translate([0,0,-0.05]) cylinder(d = inner*1.25, h = h+1, $fn=6);
    }


that's right; and its pretty good at this, see here a few examples : https://git.osr-plastic.org/osr-plastic/osr-ai/src/branch/ma... (check the generated scad files). For some reason, chat-gpto looses relations, after 5+ parts within an assembly.


You could model the internal state like code that a LLM can work with.


Nice work!!

I tried to run it, but got an API Error:

Due to some internal changes made to the Fusion API, the Add-In: 'sketch_helper' from 'Hestus, Inc.' cannot be loaded. You need to install a new version of the Add-In that is compatible with this version of Fusion.

Can you just put it up on the Autodesk app store?

I struggle with this model, because of the capture by Autodesk and the other CAD providers. If you read the Autodesk EULA, I think it prohibits use of any open-source software in these add-ins. And I don't think anyone can use this software outside of Fusion 360. And if this add-in becomes popular, Autodesk can trivially release something that has the same functionality, built into Fusion 360 by default. And, as you are no doubt painfully aware, the Fusion API can be limiting.

If you can have your LLM ingest a non-parametric CAD model and spit out a parametric model with a beautiful, complete, editable feature tree full of Extrudes, fully defined parametric sketches with these nice constraints, all tied to the sketch origin, now that's something.

I think Autodesk BIM is $5000/mo/user, Fusion is $500/mo/user. I have thought that means the money is in architectural/structural.

As a side note - one other tip I have, for all CAD users everywhere: avoid Tangent relationship wherever possible! Stick to vertical/horizontal on the lines and arc endpoints, and you will be golden. The Fusion sketch solver is badly compromised, it can't do more than two or so simple successive Tangent relations without bugging out. And, my experience with Solidworks is the same, not sure if this is still true.

Curious if you dealt much with the Fusion constraint model, and have any insights into why it works so poorly, or even how it works? Many times, you click on a line, and it turns from blue to black, and back to blue again.


> The Fusion sketch solver is badly compromised, it can't do more than two or so simple successive Tangent relations without bugging out. And, my experience with Solidworks is the same, not sure if this is still true.

In my experience of building a sketcher at D-Cubed for a consultancy client (1995-2000) on top of D-Cubed's DCM, this is because DCM's curves (which are unbounded BTW) are not directed so that there are lots of erroneous solutions to attempting to constrain G1 chains of tangent bounded curves. For example the Apollonius Problem [1][2] of 3 tangent circles/lines has 2^3 = 8 solutions. IMO if John Owen had chosen directed curves for DCM then dragging configurations of tangent circles would be more stable.

I'll end with a quote from the Preface of Julian Lowell Coolidge's 1916 "A Treatise on the Circle and the Sphere" [3]:

> Among the cartesian theorems there is a sharp sub-division between those where the radius is looked upon as essentially signless and those where a positive or negative radius is allowed.

[1] https://mathworld.wolfram.com/ApolloniusProblem.html

[2] https://observablehq.com/@d3/apollonius-problem#

[3] https://en.wikipedia.org/wiki/A_Treatise_on_the_Circle_and_t...


Thank you for the great response, that does mirror my experience, those tangent curves want to flip/change direction.

I can't believe that all of Fusion, and many other CAM software, is built on such a shaky "sketchy" foundation. It seems that in general 2D sketch constraint models are far from solved in computer science, this seems ridiculous as such a basic and elementary problem. It should be so obvious that the tangent doesn't want to go back on itself to create zero thickness geometry.

Wonder if the 2D sketch experience is much better in Onshape, NX, Catia, etc?


> Wonder if the 2D sketch experience is much better in Onshape, NX, Catia, etc?

Well those systems and pretty much all of the mechanical CAD industry is built upon D-Cubed's DCM, so I'd expect the behaviour to be the same.

If I was creating a new 3D CAD system from scratch now I'd probably license Parasolid eventually, but I'd pass on DCM.

I don't want to dump on D-Cubed and John Owen though. I have huge respect for him and the company he created. This is technical criticism with the benefit of hindsight.


> If you read the Autodesk EULA, I think it prohibits use of any open-source software in these add-ins.

That would have surprised me, as most add-ins I am using are open source and the most popular Fusion apps state that they are open source in their store listings and shouldn't have passed reviews if that were the case.

I took a look at the publisher agreement and it does have this section:

> no Open Source Software is or will be included, incorporated or embedded in, linked to, combined, distributed, or made available with or used in the delivery or provision of Your App in a manner that could cause any Autodesk Assets to be subject to any Open Source License.

There is also a definitions section about this that specifically goes into detail about copyleft licenses, so it seems like they just want to stay very clear of the issue of "copyleft poisoning" and exotic licenses, and ensure that if there is any breach of software license terms that liability is on the add-in author.


> And if this add-in becomes popular, Autodesk can trivially release something that has the same functionality, built into Fusion 360 by default. And, as you are no doubt painfully aware, the Fusion API can be limiting.

This is always the problem with add-ins like this as a business - you're essentially doing free market research for the application vendor.


Yes. We messed up, if you redownload hopefully it will work! I'm sorry about that


Still didn't work for me. I think it's stuck in my system now.

I think Fusion copies the add-in to a different directory, which I need to go find and manually delete. These add-ins should really have a Windows Installer function so that I can run the uninstaller to remove it, or have it run the uninstaller when the add-in loads or something to wipe the old files.

This is part of the problem with these add ins in general. I just installed this thing on my computer, I can't see the python source code, and what did I install and run??? Even if I could see the source, I don't have time to read it. That's why they have the app store, so that it can get reviewed to make sure it's OK. Any reason why you didn't just publish on there?

I should really take the advice of the old guys who say don't trust it if you can't see the source. I'm sure you mean well, but I would feel better/safer if it was open source.

Is there anyone who has made an AI LLM that reviews source code or an installer that you want to run and tells you if it does anything problematic? Would that also work on compiled Python source code?

Curious if you linked to the Python libraries, or copied them in with the add-in? Or were you able to avoid most libraries for this?


I looked up that error, I don't have any experience with compiled python code, but this is interesting:

Script Error

The message you received results from an add-in loaded on your machine that has compiled Python code that will be incompatible with the next update of Fusion or is incompatible with the current version of Fusion. This message is displayed when Fusion upgrades to a newer version of Python, and add-ins compiled against the previous version will no longer work.

If the message says, "The script will not be supported in the next version.", this is a warning telling you Fusion will no longer be able to load this add-in with the next Fusion update. When you get the update and the add-in quits working, you'll need to get an updated version of the add-in from the Autodesk App Store or directly from the app developer. Unfortunately, you can't preemptively get a newer version because the new version of the add-in will not function with the current version of Fusion, you will need to wait until after the update.

If the message says, "The script is not supported in the current version.", this is an error indicating Fusion could not load the add-in. You need to get an updated version of the add-in from the Autodesk App Store or directly from the app developer. App developers are contacted to let them know they will need to update their apps to be compatible with the new Python version, and the updated apps should be available as soon as the Fusion update goes out. If an update is not available, you should contact the app developer. App Developers

If you're a developer of an add-in for Fusion, use Python, and deliver any pre-compiled modules (.pyc files) with your add-in, this update will break your add-in. It will be broken because pre-compiled modules are tied to the version of Python used to compile them. Add-ins compiled with the current version of Python (3.7) will not be compatible with the updated Python (3.9) and will fail to load. This change DOES NOT affect any add-ins that do not deliver any pre-compiled modules and the.py source code is available. For detailed information about what you need to do to update your add-in and have it available for your customers at the time when the Fusion update goes out, see this forum post with the latest information.

Then it links to this, which is talking about Python 3.7 vs. 3.9.7, but that's from 2022: https://forums.autodesk.com/t5/fusion-api-and-scripts/import...

Would you need to release the source code then? Would there be the same problem with C++? How can people develop these sorts of apps without having to release the source code?

I'm on Fusion 2.0.19994 x86_64


I have been using fusion360, at least once a week for 7 years. I don't find I am ever bogged down in tedious, repetitive tasks. The sort of thing in your video is completely second nature to me and the least of my problems.

This said, it sure looks cool.

What would be good would be something more like "I want to make a shape centered on this object on this other complex object, what's the best way to do it with the simplest constraints". With Fusion there are many ways to do anything and I often watch videos whose producers are way less experienced than me but they often have some really good hints, they have come across themselves or from others.

Like when you ask ChatGPT: "What's the best idiomatic way to do XXX in C++", or similar, say in a language I don't use every day, like TypeScript.


We would love all the feedback possible! We want to make something that people want to use. Given that, we do want to eventually add features such as recommending fixes and best practices. Regarding a chat interface is something we want to tackle, but may be a little ways away :)


Good luck guys! I watched the video on YT - hopefully will get around to trying the Fusion 360 add-in at some point.

Does the current add-in use AI at all?

What is your plan when, in the event of you getting some traction, Autodesk etc. copy you innovations into the main product?

As per my other comments [1][2], I worked on this area at D-Cubed and Solidworks from 1995 to 2002. Feel free to connect with me via twitter DM @delhanty [3].

[1] https://news.ycombinator.com/item?id=41440016

[2] https://news.ycombinator.com/item?id=41440150

[3] https://twitter.com/delhanty


Great questions. This beta version uses very little, but we hope to incorporate more over time, hoping to get more accurate and better suggestions! If we can get the entire industry producing better products, everyone wins. This is something we feel passionate about, and think there must be a better way. I would love to get in contact with you to chat more!


You might be mildly interested in the fact that "hestus" in Greek can (and will) be read as "shit (on) them!"


καλά πήγε αυτό


My brother in law is in construction and has built and project managed dozens of houses, small apartment blocks etc. He said that if a building job doesn't require planning permission (small extension, temporary structure, certain farm buildings, large sheds, pergolas, verandas, sunrooms etc) then the customer might ask if he can save money by skipping the architect and engineer. A competent builder can generally work that stuff out, but it does still require rudimentary drawings (elevation, plans, end view) and some spacial figuring out. And like a lot of construction guys he has almost zero computer skills, so this shit is literally pencil, paper, squares and rulers.

I always wondered if there's a VR app (or market for one) that can map out an area and let you virtually model your project within it using standard building components (9" solid blocks, cavity blocks, lengths of timber in standard sizes, various sheets materials etc). It should be able to do basic calculations (rafter spans, roof pitch etc). Once you're done, it spits out CAD files, bill of materials etc. Bonus points for looking up local suppliers and offering to order everything for you, or even modifying the design to accommodate what's available.

It's a completely different project and target audience to OP's app, so sorry for the tangent.


I think it's feasible for someone to do this, and quite possibly the software you described, because these things are so standardized. There's a given spacing for studs of a given size, etc.

I've thought about this for decks, because they are so often so poorly done, yet the design space is almost entirely constrained by code that essentially boils down to looking things up in a table.


>He said that if a building job doesn't require planning permission (small extension, temporary structure, certain farm buildings, large sheds, pergolas, verandas, sunrooms etc) then the customer might ask if he can save money by skipping the architect and engineer.

Very interesting. Where are you located? I guess the US.


Ireland!


Interesting! Good to know that - God only knows where I may end up later in life. Curious how assuming all the freedom my brain automatically went to "huh, they're in the US". :=D


There are tools like that, but they tend to be specific to one brand of prefab. [1]

[1] https://impresamodular.com/design-modular-home-plan-using-in...


This looks like pretty much the exact kind of thing I’ve wanted. I’m just a very occasional user of Fusion 360 for personal 3d printing use - and I’m pretty terrible with CAD especially if I haven’t used it in a few weeks.


This doesn't really seem like a copilot, but it's still cool! It reminds me of the plugins for Glyphs [0], which allow font designers to speed up their process.

[0]: https://glyphsapp.com/features#plug-ins


I've no idea what I'm looking at but it looks cool. Could use a bit of an explanation what 'HV lines' are (don't look high voltage to me!) or why 'equating circles' is a good idea for us non-mechanical engineers.


HV probably means horizontal/vertical. You can see the CAD tool snapping to H/V line several times in the demo video and they intentionally make it crooked. Same with circles, if you draw it similar size as previous feature most CAD tools will snap and highlight adding equality constraint.

Honestly the demo is both convoluted and not convincing, but also trivial and I don't see where AI comes in. The HV feature can literally be `abs(angle-90deg) < 1deg`. And randomly adding constraints without deeper understanding is how you end up with poorly constrained geometry that won't work once you change design parameters.


The hard part is not backing yourself into a corner with the feature tree over time. Takes a lot of practice to avoid, and that's something AI could help with. Meanwhile, what their demo does is trivial.


Neat! At first glance, I was a little worried this was a "make me a hinge, Hestus!" sort of AI integration. This seems very logical! Coming from CAD from the purely hobbyist 3D-printing side of things, I'm pretty familiar with the tedium of some operations, (mostly because I can only assume I do most things wrong) so it makes a lot of sense to automate the scripts that would otherwise be an annoying task of "click part A... click arc... click the pivot point... click part B..."


That bit about auto-generating BOMs is awesome. Any chance you've thought about supporting the engineer with cost-saving recommendations?

I met a bunch of aerospace engineers a long time ago -- thinking about working on a software product in contract manufacturing or PDM -- and a common theme was the struggle to optimize the design for cost of manufacture.

Seems like something you guys are thinking about if focused on manufacturability.


> Any chance you've thought about supporting the engineer with cost-saving recommendations?

? As someone said: the most expensive is the microcontroller. If you remove it, you save a lot of costs.


Ha, we had similar thoughts. It is a path we want to go down eventually, but one step at a time :)


Some things I’d love to see in AI for CAD that would be above this are 1) Is it manufacturable - where it analyzes different mfg methods, for feasibility and estimated cost. So many engineers design impossible or expensive to manage parts. AI could help make this not the case 2) as others have said, BOM/SAP/etc. 3) make the drawing for a machine shop


There was a plugin for Solidworks that did something similar for machining. You essentially had a side bar that estimated cost in real time as you designed.

Feasibility I think can be done by rules checking / constraints but cost has a lot to do with the specific supplier capabilities in several dimensions.

As a machinist , yes, let's have AI do drawings, please. Getting rid of professional drafters was a huge mistake. Modern GD&T is complicated - which brings up an inherent problem. Our system of drafting a6GD&T doesn't have an explicit mathematical model - it's just based on practice. Therefore, you need some form of systemic approximation. There's one from Arizona University that NIST seems to prefer.

The other problem with automatic drawings is design intent. Consider that you may have several ways to dimension a hole, but one of them may more clearly express design intent. Pretend we drill a hole in a cube block. For the centerline of the hole, we have two different surfaces to dimension from in each direction. Depending on how the component assembles, it can make a difference given the assumption that the machining is done correctly.

If the dimensions go left - down, when the part is in the machine, it should be setup and programmed with the references on top and left. Correctly done, the tolerancing constrains not only the size of the component but also the best practice machining strategy. Tolerancing for manufacturability isn't something I hear discussed but should be. The other question is - can that tolerance be directly measured, or will it need to be calculated on the shop floor?


This is cool but I'd like to see some focus for civil engineers doing land development. 98% of them use autocad, and are stuck with it, but also they only really use no more than 100 different commands - it's just the scale of everything that's difficult and really matters.


If any founders are interested I'd be happy to go into more detail.


Feel free to get in touch with me at chad@constructural.ai -- I'm a civil engineer turned software engineer myself, focusing specifically on this space.


Yes! would love to chat, please ping us on the contact page and I'll try to get back to you


I expected something like the blender text to shape plug-in (via GPT)


I'd suggest re-doing the YouTube video so it's legible/understandable when viewing on mobile. Full desktop UI demos don't really work on mobile; need to zoom and highlight etc.


Is anyone doing something in the space for architecture mockups? I was talking with a friend who is an architect and he would love a tool that could iterate designs.


Yes! I'd love to better understand what your friend has in mind. Feel free to get in touch at: chad@constructural.ai


It's mostly creating variations of a design allowing the architect to iterate on the design with the customer to reach a final design. Today they have to create some limited set of design proposals using a lot of time and iterating is hard because any substantial big differences requires a lot of man-hours.

As an example when designing a single-family home it took several weeks to provide updated draft drawings. It would be useful for the architect and the customer that variations can be easily generated narrowing down the final design faster.


Could not your friend (or you) use Copilot or ChatGPT for such code?


This demo is lame and it isn't really providing much value. Hopefully it is just the beginning, but it currently isn't impressive.


The plugin does not load with Fusion 2.0.19994


Hi, we messed up. If you try downloading again, hopefully it should work. Sorry for the trouble


Works now, thanks.


Hi Luastoned, Can we get your contact info so that we can help diagnose the issue?


Cadpilot




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: