itstransfigure

working on it, raising our Series Seed now at xfgr.ai a proper foundation model will operate purely in the latent space to output a STEP file. there will be no LLM written cadquery/openscad. there is no other way to scale, give us 5-10 years.

we had a lot of people ask us for help with their art projects. try xfgr.ai where we tried super hard to preserve artistic intent, while also still managing to make press-brake simple bend brackets.

sendcutsend can do it as well, reasonably so. but you don't get the file. you have to buy the part from them, and they still own the file, you'll never get it for yourself.

sendcutsend wont give you the file though. Gotcha!

y'all are super harsh, and while I'm bearish on the crowded space of anything using LLM+cadquery/openscad, I congratulate samanthajeanneb on launching. that takes guts and a lot of hard work.

Hey jeremyherman at xfgr.ai we don't pretend about anything, WYSIWYG, and no LLM, only math. You asked me 10 months ago when we would have something to look at, give our MVP a shot! Working on seed round now, to train a POC of the actual foundation model I really intend to scale.

As someone that started in Pro/E 20+ years ago, used Solidworks before Dassault bought them, touched V4, became an expert at CATIA V5, and dabbled in NX and Fusion, I don't see LLMs ever understanding how to make automotive or aerospace parts.

Why? because the things you have to do in your spec tree to make an automotive stamping are so un-holy that you'll never find this outside of an OEMs vault.

to get around that I'm working on Transfigure's foundation model that goes straight to the STEP graph where spec trees don't matter.

Would love to hear from any automotive body/interior/exterior/wiring/piping folks out there. -Bobby

Give it a shot at xfgr.ai, and make some weird stuff. In particular try making some chainzz https://xfgr.ai/chainzz.png

Water / laser, simple single or double bend sheet metal brackets are do-able, technically 3-4 flange brackets are also do-able but we're working on the quality pipeline to understand what blocks these from completing (some don't, if that happens try again). Will add CNC and 3D print type shapes soon.

First person to sign up for the $200 refill and the 20th person to sign up for $10 refill gets αLPHA reloads for life*

---------

There is no LLM here, no groveling or complaining to the chatbot, no dependence on claude/chatgpt/gemini to write your openscad or cadquery code, none of that.

We're took the hard path to get AI to write 3D STEP directly because that is where things need to go if we want AI CAD to be efficient, and stop cocooning amateur hour free python CAD kernels or $30,000 CAD licenses with LLMs.

I just closed our pre-seed in May, and will be raising the Seed in July, to recruit and hire the team that will architect the first useful foundation model that writes engineer and DFM approved STEP graphs from images. We will train this foundation model with a bonkers Series A in 2027.

*terms subject to change

Also our pre-seed closes this month, if you're a pre-seed VC we've love to talk to you. We're over half way with angels, and wouldn't mind giving pro-rata to the right investor. But time is running out!

What's up everyone, I'm getting ready to launch the Public Alpha of Transfigure next week. It's our AI that makes Physical Stuff. https://www.instagram.com/p/DX-P47BB9zv/

Image to STEP is the pipeline that makes things actually easy for anyone to make parts. Everyone can draw. So anyone should be able to get a part in hand just as easily.

Anthropic - Autodesk is interesting but damn that's a lot of tokens and a lot of typing. The only winners there are the power company and hyperscalers (in 30 years)

At Transfigure we're just in early stages with an Alpha (simple sheet metal inference for now) but our pipeline will grow to handle CNC, turned, laser, injection molding, casting, stamping, you name it. Even Revit BIMs someday. Because if you can draw it, Transfigure wants to help you physically make it!

3 Years to AI that designs itself. Skynet, but in a good way.

not a great answer but the true answer would be only God knows. Personally my goal is 1 month.

posting random "artistic" snapshots of various aspects of Transfigure (instagram: itstransfigure) if you care to follow.

without saying too much, the competitive advantage is in training data and I think that will be the deciding factor as to whether or not a company is successful in (Transfiguring? rezzing? summoning?) a real part from a sketch, that can be made into physical stuff, without looking like shapenet cartoons.

as a founder working on a direct competitor to this product, I am happy to see there's so many opinions and interest. this means we are on to something. I am totally behind the idea, so I am super excited for Pranav to get this in the wild.

however, the output of the current SGS-1 model leaves a lot to be desired in terms of usability and manufacturability.

as sir liam powell so explained, the STEP generated is indeed a STEP file in the fact that it will load on your CAD of choice (CATIA, NX) but it's not even close to being useful for the designer/engineer/fabricator much less a factory or even a 3d printer.

Most importantly: -Surfaces are not G1, G2, or even G0 (un-CAE-able and un-manufacturable) -simple shapes are hyper overcomplicated (a fillet turns into 10++ surfaces mashed together) -and most unfortunately in one part I generated, the output was a collection of separate STEP child parts (meaning, when importing, your single "part" is actually a product containing 10000+ child parts)

looking forward to V2, keep it coming

shameless plug: my entire career has been in CAD and that is why I started Transfigure. to eliminate my own job as a CAD monkey: and get on with the engineering. already having anticipated the limitations which SGS-1 is dealing with, Transfigure's AI architecture is approaching the problem from the perspective of a mechanical engineer whose customers rely on clean data, ready for simulation and manufacturing. because if I submitted data generated by SGS-1 to our fab shop... I would be fired.

yo@xfgr.ai