[demeyer1]

We have a few million unique users, which is small within the broader context of 3D creators and consumers - so small N, but I'll share a couple observations at that volume.

Adoption

STLs remain the most popular, today, on our platform. 3MF has been growing, but only in the past quarter did our users tell us in meaningful volumes that we need to support it - which we did, happily. The push to 3MF seems to be gaining momentum.

Expected benefits vs current state (again, small N and biases in my data)

1. Color/texture is a benefit, though creators include it to varying degrees based on their use case and the intended consumption. We support printers, general 3D designers/sculptors, and mechanical engineers - so our audience is broader than primarily/solely 3D printing. As such, adoption varies considerably depending on the user and use case.

2. Often, the model has more metadata (though rarely includes color, photorealistic renders, etc) and a smaller storage footprint. Benefits creators, consumer and platforms. Though again, adoption lags on the more advanced capabilities inherent to the format.

3. Support for assembly like models is a huge benefit! However, the average 3D printing oriented creator still prefers collections of STLs. In user interviews, I often hear the rationale for this is that there are (perceived) tooling and general workflow changes that seem time consuming - so the usual.. change/learning takes time.

Timing

Creators will drive adoption, as if often the case. We are seeing this increase every month on our platform (thangs.com), but the relative share of uploads still skews towards STL and other formats.

Platforms will need to adapt and take advantage of the format benefits to help consumers see the value.

Over the mid term, I am really excited about 3MF. The fact that there is a spirited discussion on HN about it, if nothing else, strikes me as a great sign!

[ddingus]

>Support for assembly like models is a huge benefit! However, the average 3D printing oriented creator still prefers collections of STLs. In user interviews, I often hear the rationale for this is that there are (perceived) tooling and general workflow changes that seem time consuming - so the usual.. change/learning takes time.

Yes! STL actually does this, and I've employed it for years, but often find software isn't expecting / designed for it, and or others are not aware and use the bundle of files workflow and it's easiest to just go with that flow.

Model metadata is exciting to me, and it's for reasons similar to those found in other manufacturing contexts. JT is a great example, and in the 90's VRML actually got used for model + metadata representation by SDRC / FORD to communicate tolerances and other data along with the geometry.

Things are very slow to change. Decades for some of this stuff. We were capable of real paperless in the 90's, and still... Not there, but more there than not these days.

[demeyer1]

Well said. What I suspect we are seeing is the beginning of a trend where the rate of change begins to accelerate.

If nothing else, the growth in new 3D creators (be they mechanical, design/sculpt, AR/VR/MR/XR, printing/additive) will bring in new individuals less beholden to the way things have been done in the past. This is happening today.

If we look at those with some of the highest 'change cost', we can look to CAD and PLM within manufacturing. Most of the tooling in that category (design clients through to PLM) hasn't even come close to keeping pace with modern conventions. I'm as excited as you, even within that relatively small, change averse context.

[ddingus]

And, where that tooling does provide for modern conventions, it often goes unused. Same inertia in play that we have been discussing.

I have only a quibble or two with 3mf.

Curved triangle representation is one, and lack of source topology is the other.

At some point, we need to generate real curves, and frankly I would be pleased with second degree curves. (The analytic ones, arc, comic, hyperbole, parabola)

When describing small features, and when coupled with machine process requirements, chordial deviation becomes a real issue. We either end up overloading path planner subsystem with a crazy number of too small line segments, or accept a cutoff of 16 ish and deal with undersized radii.

As far as I understand, we do not yet have those features in the output file specs just yet, and it sure would be nice if we did.

Frankly, arc and potentially second degree curves in general, fitting makes some sense at this point just to make use of firmware that supports it.

Software I am involved with has headed down this road to both resolve small features correctly so that they are printed correctly, and to enable feature discrimination so that tool paths can be generated with higher order formulas intended to improve on both machine performance and part quality, consistency.

[ddingus]

Watching the new types of creators is super interesting!

I am fluent and trained many groups on high end tools, like NX. Sadly, I made a ton showing people how to do higher order things with the base toolset. (Almost nothing one cannot build to high fidelity these days)

I feel there is a lot of false value out there in those tools too. We would all benefit from the base toolset being extended and deepened some so broader adoption is on an incentive path, not so discouraged for license cost reasons as it is today.

And I get why. The hit in revenue would be profound and the skills needed to really use those tools remains a barrier, meaning there is no realistic volume strategy.

Still...

In my current role, I get the opportunity to use a lot of new tools. Have almost no dependence on the traditional CAD/PLM stacks!

I can use that stuff, and will at times because it makes the most sense.

But, in the additive sense, one can create with Python, or OpenCAD, Blender (!?!) Etc.

Voxels are becoming a thing, as are various mesh tools that are nor chained back to mega geo kernels...

I suspect as these people mature into the manufacturing and design scene, we will see another wave of changes on par with what NURBS and solid models did compared to wireframe.

And that clash? (With the traditional geometry stacks)

Might be epic!

In a good way.

One particularly interesting aspect, or potential I should say, is the non NURBS ways can be employed both in concurrent and parallel fashion!

Something like Parasolid can be employed in concurrent fashion, and that depends on model history being present or not and how it is shaped. We see that today when users of the high end tools really put the features to use.

However, there are limits and a a largely sequential compute path requirement on any given branch.

When one sets that aside to favor other ways of representing models, parallelism enters the game and the possibilities are very well aligned with additive.

One last ramble, because I live this stuff: Hybrid systems may well be able to incorporate both,,making the big tools bigger and that may be one outcome from that clash I mentioned above.

However it all goes, and I may well have it all wrong, we are headed into new and very interesting times.

[demeyer1]

Couldn't agree more with your points, and I agree that this clash is going to happen.. and that will be a great thing if it does!

If you are interested in connecting - I would be happy to trade observations over a video meeting and coffee/beer. [e-mail]

Disclaimer: I don't know if we are allowed to post this on HN, off the official posts, but I'm also hiring. We do 3D search, are growing rapidly, and have some excellent investors: Sequoia, Google Ventures, Tiger, etc.

[ddingus]

Looks like a pretty cool project you got going there. I wish you luck, and success! I always thought that was a high-value space, but many of the tools are just thick. Expensive kind of painful.

I'm not looking at new opps right now, (have a startup of my own ramping up) but we may find a chat worth it. I'll try and connect with you in the near future. :D