The Robot Body Now Iterates Overnight — But Only the Parts You Can Print

A rubber-like part takes repeated stress on a desktop printer's build platform and snaps back into shape. That's Formlabs' Flexible 80A Resin V2, and I posted the clip because it looked like a small material update sitting on top of a much larger shift.

Robotics progress usually gets narrated as brains. Better foundation models, longer planning horizons, sharper perception. I've written plenty of that myself. But a good share of the bottleneck sits somewhere less photogenic: the body. Grippers that hold an egg without cracking it. Joints that flex a hundred thousand times. Skins that give the way tissue gives. For years, testing one of those shapes meant tooling, molds and weeks of lead time before you learned whether your idea was any good.

[BEELD NODIG] flexible printed part flexing and snapping back

What actually changed in the resin

The specifics matter here, because "new material" is a claim that gets made every quarter. Against V1, Formlabs puts V2 at roughly twice the tear strength, twice the elongation at break, and four times the rebound, with a tack-free matte black finish and better aging behaviour. Distributor data sheets (MatterHackers, GoEngineer, current listings) place it at Shore A 68–83 depending on post-cure, tear strength up to 28 kN/m, elongation at break up to 230%, and ultimate tensile strength around 10.9 MPa. Formlabs positions it against 80A cast urethanes and hard rubbers.

Of those figures, rebound is the one that tracks what you see in the clip. Rebound is the snap-back number. V1 parts came back slowly, which is fine for a gasket and useless for a finger pad that has to reset between grasps. Four times faster return is the difference between a demo material and something you'd put on a joint. I'd hold the exact multiples loosely — they're the manufacturer's own comparison against the manufacturer's own predecessor — but the direction is consistent across the third-party listings, and the mechanical band is specific enough to design against.

The other half of the story is curing, and it's the part I underplayed in the post. Printing was never the whole cycle; wash-and-cure ate the morning. Formlabs' next-generation Form Cure, shown at Formnext in November 2025 and recapped in the company's year-in-review of 11 December 2025, claims up to 6x faster post-curing, with general-purpose resins cured in under 60 seconds and validated presets for more than 45 materials. Form 4 delivers most parts in under two hours at up to 100 mm/hr, per Digital Engineering 24/7 (February 2025). Print plus cure is now genuinely same-day.

And engineers do switch when the gap closes. Formlabs materials PM Adam Lang told VoxelMatters (14 November 2025) that Tough 1500 V2 saw twice the adoption rate of V1 since its spring 2025 release. That's one company's data on its own product, so treat it as directional. But it matches what the research literature has already settled on: a fabrication review in Advanced Robotics Research (8 September 2025) treats 3D printing as the default route for soft grippers and actuators, and a hybrid gripper paper in Electronics (17 February 2026) makes the argument in the authors' own words — printed components make geometry changes straightforward and cheap, so one form factor gets re-adapted per application instead of re-tooled.

[BEELD NODIG] soft printed gripper handling a delicate object

The distinction the post didn't make

Here's where I'd sharpen my own claim. The body isn't catching up. Prototyping the body has caught up. Call it the Print-or-Procure Split: everything about a robot that is geometry, compliance, grip surface or skin has entered software's iteration cycle. Everything that is actuator, battery, magnet or bearing has not, and won't via desktop SLA.

That second column is where the real wall sits. An IEEE Spectrum investigation (October 2025) put Agility's Digit at roughly 90 minutes of runtime, operating in 30-minute intervals at Amazon, and Figure 02 at two to three hours — against industrial customers who expect 95–99% uptime. McKinsey (17 April 2026) notes that China holds around 90% of permanent magnet processing, 40% of precision bearings, 35% of motors and 30% of power electronics. Money isn't the constraint either: the same McKinsey piece, citing CB Insights, has robotics VC funding more than tripling from 2023 to 2025 to $40.7B a year, and Yole Group (November 2025) counts $9.8B cumulative into humanoids since 2017 across 60-plus companies, with UBTECH and Figure taking $4.3B of it. Goldman Sachs, for its part, still models only about 1.38 million humanoid shipments globally by 2035.

So the honest read: your CAD loop is no longer your bottleneck. Your bill of materials is.

Why that split is useful to you

If you build hardware, the split tells you where to spend attention. Anything on the print side should now be redesigned weekly, because a bad idea costs you an evening. My own archive keeps turning up the same pattern from other directions — open-source robotic hands with printable tactile sensors, e-Flesh sensors that read deformation, printed products that go soft where you want comfort and firm where you want support, zone by zone. Those are all print-side wins, and they compound.

Anything on the procure side deserves the opposite treatment: fewer variants, longer commitments, and honesty about lead times you don't control.

What I'll be watching is whether the split holds. Printed actuation — pneumatics, soft muscles, tendon routing — is the one place it could break, because that would move power out of the procurement column and into the overnight one. Until then, the interesting question for your next build isn't how fast you can iterate. It's which half of your robot you're allowed to iterate on at all.

[BEELD NODIG] humanoid robot hand assembly, mixed printed and machined parts