Author Archives:

The Bit Box – An Electronic Twisty Puzzle

I’m happy to announce the Bit Box puzzle! This has been in the works for 9 years on and off. This is the 5th version, and it works!

It was inspired by the electronic game “Lights Out”, as well as “Merlin” and of course the original Rubik’s Cube.
Each cubie has lights on the faces behind glass tiles. Upon the turn of a face, the cubies on that face toggle between on and off. I think the video will help make the idea clear.

Photographing these bright LEDs with digital cameras is challenging, so please excuse the clamping.

iMaterialise did a great job on the print.

Can a fully ‘on’puzzle be solved into a fully ‘off’ state?  –  yes

What does a checkerboard pattern look like?  –  unchanged




Making the Bit Box


Version 0 – 2007

The Bit Box was first conceived in 2007.  At that time, I modified xcube to simulate the idea.

It’s convenient that I used version 0.0 in the title bar when I did so!


During this time, I called this puzzle various names:

Binary Cube

Bit Box

Twist Off or Twiss’d Off

In the years from 2007 to 2013, I spent most of my puzzling time thinking about other projects like the Pentultimate, Petaminx, Radiolarian, etc.  But I was always thinking and planning how one might be made.  I had already considered the mechanical approach to making this puzzle, but the details were daunting.

It was clear that the usual approach of 3D printing everything could not apply here because of the electronics.  I would need precise metal components to make the contacts work.

I needed electrical contacts with a built-in fillet, so that the contacts would not cause lockups.  After a few months considering this problem between projects, I decided that brass brads could be used for the electrical contacts.  They have round heads with rounded edges, and conduct electricity perfectly.


Version 1 – 2014


The first version of the BitBox was designed in April 2014.  In most obvious ways, this design is almost identical to the final working puzzle.

The main, defining problem in this version of the puzzle was that the edges and corners were too hard to assemble.  In the edges, the brads had to turn a corner while being inserted:


In some cases, I was able to get it to work, but it took a very, very long time.  I was able to assemble one part every night in about three hours.,

At this point I started to worry.


The corners were even worse.  One brad in the middle and three around it had to be inserted without touching each other.  It took a long arduous fight to get each brad in.  Paving the way for each brad destroyed multiple brads.


I was still able to make a few successful parts, but I had to use a lot of insulating tape, and again, each part took a long evening.



Version 2 – 2014

The second version was designed to make inserting the brads easier

Tracks were added to the brad holes to guide the brads around corners and into the interior of the part.

They were added to the edge, but ended up too tight to use: (click for a closer view)


They were also added to the corner: (click for a closer view)


The corner was completely unusable.  Brads couldn’t even be forced in.



Version 3 – 2014

In the third version, the edges were finally usable.  Here’s a section view showing the brad tracks.


Brads could be forced in without too much effort.  I experimented with using the tracks to make electrical connections by forcing a wire in alongside, but it wasn’t a success.

20160205_234936  20160205_234948


The corners were also improved.  The tracks were wider and included holes to help nylon powder escape.

20160205_235002   20160205_235010

In spite of the improvement, getting the brads in was still too labor intensive, with so much force required that they constantly bent.

20160205_235017  20160205_235049

Even so, I was able to begin experimenting with assembly.  I tested the fit of the battery pack in the core, which worked well.


picture 005

I also mocked up the planned electrical design on a breadboard to make sure that I could safely and successfully power all the LEDs.

picture 138

I added the brads into the core, and began adding their electrical connections.


picture 141  picture 143




The entire puzzle was too loose.  Even if I tightened the centers as much as possible, they still weren’t strong enough at holding in the edges to make a strong electrical contact.  The more I tried to make the edge contacts tighter, the more the edges lifted up the centers from their contacts.

I tried to make the corners fit closer to the core by switching from brass brads on the centers to nickel thumbtacks.  In addition, the thumbtacks were clamped flat to make the centers even tighter.

20150928_222923 20150930_212718 20151115_010138


Even that didn’t solve the problem, so I also added small springs to the brads on the edges to try to create some pressure on the connection and bridge the gap when it was too loose.

20140629_000748 20140629_001134

Using these techniques, I was able to get some parts to light up.  Unfortunately, the more parts I added, the less likely it was that they would all make good contact.  A very frustrating process of trying to make this work began at this point.




I made the decision at this point to redesign the core in order to make the centers, and thus the whole puzzle, tighter.


Version 4 – 2015

The new core sinks the brads for the center connections by about 1mm relative to all the other terminals.



I also fixed a few other small issues at the same time.

The new corners were broken into two pieces so that the brads could be inserted more easily.


Here is a section view showing the tracks inside.


Brads could be inserted into them separately to clear out any clogged powder.  Then, the brads could finally be inserted into the bottom part, and then threaded into the top part, before simply pressing the two together.


Soldering was a major time sink in past iterations, so I tried using fishing line crimps to connect wires.  These saltwater line crimps are the perfect size to slide over the tail of a brad.

In the end, I still had to solder, but at least the mechanical connection was much faster.


First, I mark the terminals on the core with black and red sharpie for negative and positive connections.  Some brads are inserted:


And then I begin connecting like polarity terminals with wire:

20150907_223949 20150907_233726

The wires spiral around from the center, going through all like polarity terminals.

20150908_002548_001 20150908_010707 20150908_010714 20150908_012734

Almost done:


And here’s a finished half.  I tested this core for some time to be sure that the connections were good.  I found a disconcerting fact.  The thickness of the core was 1mm.  Of course this was an optimization to keep material cost low.  However, I found that the core could flex while working with it, and the flexing was making the connections a little flaky.  I think this was mostly a problem between the core and the moving parts, but it became increasingly worrisome.


Nonetheless, I proceeded to try to get it to work.  It was so promising, but just fell short of being usable once corners were added.

20150908_233742 20150915_214849

20160206_21515120150916_224713 20150918_215643 20150921_221429 20150921_223324 20150922_205312  20150922_205339    20151118_223102 20151118_223732 20151118_224817

The parts were all designed to mate with rounded glass square tiles from Sun and Moon Crafting Kits.


I noticed that many times, the contacts weren’t working when they were very very close to touching.  Sometimes, so much so that it looked like they were touching to the naked eye.  This would cause the lights to dim or flicker, or even just be off when they should have been on.

I recognized that oxidization could be a factor, so I tried conducting oil called DeoxIt.


20151220_202528 20151220_202541s


DeoxIt dissolves minor oxidization and helps prevent future issues.  It did help, but in the end, it became clear that this design would always feel unreliable even at its best.

I realized that since the contacts were already so close to touching, they just needed a little attraction in order to pull together.

Version 5 – 2016

I had the idea of using neodymium ring magnets behind the brad heads.  Once the brad heads approached each other, the neodymium magnets would be attracted and help create pressure on the contact.

In order to do this, the entire puzzle had to be increased in size to make room for the magnets.

I found neodymium ring magnets the right size to slip the leg of the brad through, so the magnet could line up behind the round brad head.  The supplier was K&J Magnetics.

R622L (1)

Once I received the magnets, I created some trimmed brads with magnets applied, and stored them on a nearby lamp.  Otherwise, they would fly across the desk to make contact with each other!


The core was modified to make the depressions for the brads deeper to allow for the magnets.  The walls of the core were also thickened to 2mm!  The hope was that a thicker core would be less likely to flex and break contacts.


At the same time, I decided to use the increased size to simplify the design.  Tracks would no longer be needed for the brad insertion, for example!  The larger sized parts allowed me to position the contacts such that the brad legs could come straight up into the parts without bending around any corners.  The size difference was significant, though!


Also, I made all walls on the moving parts 2mm thick, so that flexing and squishiness would no longer be a problem.

Again, I began by marking the core with polarity marks.

20160115_210958  20160115_211010 20160115_211015  20160116_223140

And then it was time to start adding the brads.


I was still using the fishing crimps, and I decided it was worth one last look to see if I could find some kind of crimp that was designed to make an electrical connection, so I wouldn’t have to solder these hundreds of connections.


I ended up finding something called spade terminals, which act like they were made to slide onto the legs of a brad and make a great electrical connection, no solder required!


They worked perfectly!




20160118_133639    20160118_210512  20160118_212715 20160118_213331

To assemble the centers, I prepared the different parts first.  I colored the spade terminals with sharpie to make the polarity simpler to see.  I had also purchased nicer red and black wire by this time, so that things would be clearer to look at.

The led assembly was bent and trimmed according to a template so that it would mate with the spade terminals at the right angles inside the center piece.

20160123_223110 20160123_223124

Here are the parts needed.  The 3D printed center.  The 3D printed reflector, which doesn’t really reflect.  It just covers the electronics and allows the LED to show through.  Four spade terminals.  Two wires.  Four brads with ring magnets applied.


And here is the completed center.  The screw is reached by gently pushing the LED aside.


Edges were done in a similar way.  Parts were prepared, including an LED harness with two LEDs, two resistors and two spade connectors.  And two brads with magnets applied.  In this photo you can see how the brad legs can now just enter the bottom of the 3D printed edge and come straight into the part.


A few edges were assembled in succession.


Corners require 4 spade terminals, 3 leds, and two wires.  The two wires are used to connect the outer terminals so that connection is successful even if only one makes solid contact.  Assembly continued…


20160125_215646 20160125_225524       20160125_231531  20160125_231557 20160127_220024


At a certain point, the puzzle was ready for the core to be closed, so the batteries were secured inside, and the core cap was applied.

2016-01-28 22.54.05    20160128_225046    20160130_004430

Here you can see one of four screws that secure the “halves” of the core together.

20160130_004733 20160130_004735  20160130_004911 20160130_004915  20160130_005103_sml  20160130_005108


And that’s basically it!  9 years later, I have a working completed puzzle!

Thanks for reading.






Chopasaurus Jr. – 6/2014

The following article is taken from the original announcement of the puzzle on in 6/2014.

This is a vertex turning dodecahedron cut a bit shallower than the Chopasaurus Rex. Hence the name — Chopasaurus Jr.

As a vertex turning dodecahedron, it is “dual’ to a face turning icosahedron like all my radiolarians over the years.

It can be classified under GelatinBrain as “1.2.7”

I designed Radiolarian 11 (The Radio Star) with the feature that every part has a cap. I thought it might let me make shapemods easily someday. And here we are with the first — a dodecahedron mod!

If you’re interested in the internals, please watch my “Making of the Radio Star video”

Video Here


Radiolarian 1.5 – The Radio Cannon – 5/2014

The following article is taken from the original announcement on on 5/2014.

This is the face turning icosahedron with fixed centers like a canonical Rubik’s Cube or Megaminx. Because of this canonical property, I’m calling it the Radio Canon.
This face turning icosahedron sits between Radiolarian 1 and radiolarian 2, so I’m thinking of it as Radiolarian 1.5.

The Radio Canon Jumbles! And yes, this is the very last one of the Radiolarians I’ll make. I guess it’s a five year series now, started in May of 2009 and ended in May of 2014. Ironically this is the shallowest planar one.

That’s an average of only 3 Radiolarians per year for the last five years. What a trek!

Video here.


Dave’s Diamond – Unsuccessful V2 – 4/2014

This puzzle does not work!

The following article was taken from the original post on in 4/2014.

This is a second version of Dave’s Diamond. It has been redesigned larger for stability. It’s much much better, so I’ve also made a video showing some of the possible moves.

Video here.

For discussion of the puzzle, see the original topic here:


Helicopter + 3x3x3 – 4/2014

The following article is taken from the original announcement on in 4/2014.

The Heritombo Cube was actually designed and printed after I started to doubt this design for a Helicotper 3x3x3 would work. After waits for a few missing parts, and a design tweak, it has turned out that I’m revealing the original design after the Heritombo Cube.

I made the cuts shallower first as an experiment, but when I saw the face pattern, I immediately started shooting for a 3x3x3 hybrid. In the event that someone makes a flat cut Helicopter 3x3x3 that doesn’t use shallow cuts, I’ll rename this one.

This was completely designed on a flight to China, and even uploaded to Shapeways somewhere over the Pacific.

This is now on Shapeways for $85.

Video Here.










puzzle 004_mech

The Brachio Star – 4/2014

The following is taken from the original article on in 4/2014.


This is another series I was bound to complete. It is the Brachio Star– the icosahedral version of the Dino Star. Since it’s bigger than the Dino Star and bigger than the Bronto Star, the name “Brachio Star” sounded good.
Since I got to design the whole thing, it moves nicely and is stable, if a little stiff.
This is now on Shapeways for about $150 in wsf HERE. You need to order all 3 sub models.


The Heritombo Cube – 3/2014

The following article is taken from the original announcement of this puzzle on in 3/2014.

The Heritombo Cube is a shallower cut variant of the Helicopter Cube with straight cuts that reveals the edge centers and face centers. The name comes from a diminutive helicopter.

You may notice the edge caps don’t want to quite sit flush, but it’s just a print tolerance issue. The design is not fudged. I could sand each cap but I think I’ll just print toleranced replacement caps.


The Copper Orb – 1/2014

The following article is taken from the original announcement on in 1/2014.  Some extra assembly photos have been added here at the bottom of the page.


I’m happy to present the Copper Orb. It’s a LOWER order version of the vintage Parker Brothers puzzle “The Orb”. Instead of 4, 5 or 6 tracks of beads, it has only 3 tracks.

It has been plated with copper, and then sealed with Epoxy. A vintage orb donated beads for this puzzle.

The rows of beads have the following counts (8-16-8). The number of beads are dictated by the geometry, and the fact that the smallest rows have 8 beads.

This orb can stop in 2 positions BETWEEN each 180 degree turn.

This puzzle is actually a little more difficult than the 4 row Orb because there is no 90 degree position.

Thanks for looking!



These assembly photos were not included in the original announcement.


20140118_141017 20140126_135149 20140126_143132 20140126_161056 20140126_161521 20140126_161754 20140126_162105 20140126_162251 20140126_162534

Little Chop V2 – Successful Skirting Rails Mechanism – 11/2013

The following article is taken from the original announcement on in 11/2013.  Assembly photos have also been added at the end.

The first version of my Little Chop using skirting rails was promising, but in the end it was unstable. This version is stable! I’m very excited to have a working Little Chop using a unique mechanism.

This new version is 100mm instead of 67mm to an edge. The additional 33mm really do make the puzzle feel much larger, but they also make it stable, even allowing new structures inside that increase stability.

Jumbling on this puzzle is blocked by the mechanism, allowing some partial jumbling turns before blocking is complete.

You might notice in the photos that I have managed to scramble it in my excitement. See the deed in the video:

Video is uploading here.



To understand the skirting rails mechanism, please see the Pentultimate V2 article, and the Pentultimate V3 article first.

Also check out first version of the Little Chop for discussion of the mechanism.  The same skirting pattern was used on this puzzle, but the void master mechanism was much improved by making the puzzle larger.

Again, we can see the blocked equator here:


The following mechanism pictures were not included in the original post:

SGS4 722 SGS4 723 SGS4 724 SGS4 725 SGS4 726 SGS4 727SGS4 721

Radiolarian 11 – The Radio Star – 11/2013

If you haven’t seen it, I recommend watching the making of video first:

Making of the Radio Star


The following article is taken from the original announcement on in 11/2013.




The Radiolarian puzzles have all been made, but my first version of the Radio Star (Radiolarian 11) was a total failure.
I started it in September of 2012, so I’ve been working on it on and off for over a year.
Here’s the thread on the topic of the failed first version:

The new second version is a success!

Video uploading here.

There’s a long, long story behind this one. I did record some making-of material that I may cut into a short video sometime. The sad truth is that there may be more time in this puzzle than I put into the Petaminx.

I call it the Radio Star because this puzzle is to an icosahedron what a StarMinx is to a dodecahedron. It is the cut depth at which the corners have fully disappeared.

As you can see, the puzzle is huge. The edge lengths are 102mm. It uses the “Shells on Rails” mechanism, which combines traditional shells with void-style rails internally. This allows me to build a much deeper puzzle internally for larger stable corners, and then go shallower on the outside using shells.

Here’s a part that illustrates the concept.

The Radiolarians have had two types of jumbling. This one is the first that I have noticed the ability to double-jumble. Both types of jumbling can be set up on opposing faces before being jumbled simultaneously. I’m sure Radiolarian 10 must also have this ability. There are pictures of this below.


The Golden Orb – 10/2013

The following article was taken from the original announcement of this puzzle on in 10/2013.  I am adding some behind the scenes photos at the end that were not part of the original post.


I’m happy to present the Golden Orb. It’s a higher order version of the vintage Parker Brothers puzzle “The Orb”, and a step above the Silver Orb. Instead of 4 or 5 tracks of beads, it has 6 tracks.

It has been plated with copper, chrome, then gold. Two vintage orbs donated beads for this puzzle. The white and black are 3d printed.

The rows of beads have the following counts (8-22-30-30-22-8). The number of beads are dictated by the geometry, and the fact that the smallest rows have 8 beads.

This orb can stop in 5 positions BETWEEN each 180 degree turn. They are all shown in the photos below.
Now that there are an even number of rows, there’s an exact halfway position again. But instead of the 2 orbits in the halfway position on the original Orb, this puzzle has three orbits.

Thanks for looking!

In this thread, the Silver Orb was shown:

In this thread, we discussed the number of beads that should be used on this puzzle;





The following behind the scenes photos were not part of the original post:


SGS4 031 SGS4 032 SGS4 033 SGS4 034


SGS4 126

SGS4 636 SGS4 637 SGS4 638

Little Chop V1 – Unsuccessful Skirting Rails Mechanism – 10/2013

The following is taken from the original announcement of this puzzle in 10/2013.

This is a little chop using the skirting rails mechanism. It prevents jumbling, so internal blocking due to hidden jumbling doesn’t happen.

I was hoping a 65 mm version of the little chop would be possible. The puzzle is too small to be perfectly stable though, so a larger version is in the works.

This experiment was done with the blessing of Matt Shepit.

Video here


To understand the skirting rails mechanism, please see the Pentultimate V2 article, and the Pentultimate V3 article first.

The little chop puzzle has 6 slices, like the Pentultimate.  And like the Pentultimate, 5 of the slices can be locked to one of the exposed centers.  (Here the locked face is colored red, and any edges that touch that face are colored red, to show that they are locked.)

Only the equator remains to be locked.  I chose the center of the cube face to create an interacting nub to collide with the core, indicated here by the green dot.



Here is the core, with a red part added to indicate the colliding nub.  We can see that the part will be allowed to skirt in from above and below, but it won’t be allowed to move sideways.




The original article did not include these mechanism pictures:

SGS4 641 SGS4 642 SGS4 643    SGS4 647 SGS4 648

Radiolarian 15 – The Radio Chop – Skirting Rails Mechanism – 10/2013

The following was taken from my original announcement of this puzzle on TwistyPuzzles,.com in 10/2013.


This is the 15th and last in my series of increasingly deeper cut face turning icosahedrons. This is the deep cut puzzle and thus the series is now technically complete at 15. I still plan to release a V2 for the Radio Star (number 11), but otherwise this represents the end!

As you all know, the equivalent dodecahedral Chopasaurus was made first a few years ago by Andrew Cormier. My version uses an entirely different mechanism called “skirting rails”. In other words this is not a shells mechanism. It is closer to my Pentultimate V2.5 mechanism.

One very unique property of my version is that this puzzle does not jumble! Deep cut jumbling puzzles can sometimes block internally, which is a property I don’t find desirable. In those cases, external turns look valid, but can’t be done due to the internal state which cannot be seen. Since my puzzle does not jumble, it does not have this problem.

Turning is pretty tight, but it’s nice and stable. Actually starting a turn seems to be the hardest part as you can see in the video. I assume it’s because of so many parts having to align themselves perfectly inside.

Edge lengths are 88mm, making this a very large puzzle!

Video Here!


To understand the skirting rails mechanism, please see the Pentultimate V2 article, and the Pentultimate V3 article first.

On the Pentultimate, we were able to lock down 5 out of 6 tracks by locking them to the top center.

On the Radio chop, we can only lock down 6 out of 10 in this way.  I’ve colored these 6 tracks red in the image below:

Untitled-2 copy

The four remaining tracks we need to lock down are the equator (shown in yellow) and three tracks that sit between the equator and the locking top face.  I chose the points marked in green to attach pegs to the master layer that can interact with the special core.  Allowed paths are swept out of the core and the rest is used to collide with the pegs and prevent illegal moves.

In the following images, I have colored the pegs red.  Here we can see how the blue tracks will be blocked.  They can rotate in from above and below, but not move sideways:



Here we can see how the equator will be blocked.  The pegs can rotate in from above and below, but they cant move left-right.


These assembly photos were not revealed in the original post.

SGS4 566 SGS4 567 SGS4 568 SGS4 569 SGS4 570 SGS4 571 SGS4 575 SGS4 577 SGS4 579 SGS4 580 SGS4 583 SGS4 591 SGS4 596 SGS4 598 SGS4 603


Silver Orb – 7/2013

The following article is taken from the announcement of this puzzle on in 7/2013.


I’m happy to present my Silver Orb puzzle. It’s a higher order version of the vintage Parker Brothers puzzle “The Orb”. Instead of 4 tracks of beads, it has 5 tracks. It is 3d printed in Nylon and polyJet resin, with some parts plated with copper and then with chrome by a classic car parts restoration company.

Some beads are harvested from sacrificial puzzles bought on ebay, and the white beads are 3d printed, designed to the exact same dimensions.

The rows of beads have 8, 22, 26, 22, and 8 beads. There is a tiny bit of fudging to make the bead count work. Can you see it? Would the next higher order orb need the fudging? :) Highlight for the answer –> No, the next orb falls gracefully into whole numbers of beads.

Instead of the 3 midway positions the original orb can stop in, there are 4 midway positions, (only 3 are shown below), all of which combine all beads into a single row. Does this affect the solve?


Radiolarian 14 – The Radio Fathom – 5/2013

The following article comes from the original announcement on in 5/2013.

This is the 14th in my series of increasingly deeper cut face turning icosahedrons. This is the second to last puzzle in the complete series of 15. The next and final puzzle in the series is the Radio Chop (These names have recently been shuffled around as a nod to the Chopasaurus).

This is a very large puzzle. With the cuts at this depth, we must have a very thin slice ring around the puzzle, which creates some very small parts with acute angles. In order to make the puzzle work, it has to be large.

Check out the size comparison below with Radiolarian 3!

There are 362 external parts, giving this puzzle close to the number of external parts on a teraminx, with the smallest part on this puzzle actually about the same size is a small teraminx part.

It jumbles in the same two ways that previous Radiolarians jumble, but because of the depth of these cuts, the jumbling moves seem to take about 1/5 of the total pieces along. Jumbling images are below.

Turning is quite good considering the depth of the cut, especially once the parts are jiggled into alignment. It has more friction than I like. I’d love to see a polished version of this some day.

Video here.

Jumble Type One

Jumble Type Two



Picture1 008  Picture1 010  Picture1 012  Picture1 014 Picture1 015    Picture1 019   Picture1 062 Picture1 063 Picture1 064   Picture1 066   Picture1 069 Picture1 070    Picture1 074  Picture1 076 Picture1 077

Radiolarian 13 – The Radio Gem – 5/2013

The following article is taken from the original announcement of this puzzle on in 5/2013.

I’m happy to present Radiolarian 13, the Radio Gem! There are two more in the series.

This has been a very challenging puzzle to get to this point. This is the second version, and the third print. This one has been tough financially. Assembly and stickering take 8 hours.

Compared to Radiolarian 12, the deeper cuts in this puzzle have caused 60 outer parts to disappear. Unfortunately, stability is not increased much from the outer simplification, but hangups are reduced which is great!

Turning is fair, although I feel the puzzle is not as stable as it should be, so I’m very, very careful in turning, and did not make the video too long. I can feel the puzzle would be likely to pop under normal solving, and with a deep shells puzzle like this, a pop could mean the whole puzzle. I am not going to risk another long assembly at this point! It’s only the long road and problems already solved to this point that makes me proud to present!

Video Here







Radiolarian 12 – The Radio Nebula – 3/2013

The following is taken from the original announcement in 3/2013.

After Radiolarian 10 and 11 both failed, this is a relief. Radiolarian 12 works!

This is getting very deep now, and I did not anticipate that the new corner and the surrounding star arms we gain at Radiolarian 12 would actually help stability, but they do.

Movement is fair, but after the desert I’ve been through, it feels Excellent!!

The main problem I have seen with the design is that the corners themselves can rotate in place by accident since they are so small. The design would have to be much larger to make them more stable. I think 2-3 times as large would work well. For now, call them trivial corners I guess!

I’d like to take this apart to apply silicone to the inner puzzles, and break-in the interior shells without being encumbered by outer shells before a video, so I’m doing that now. In the meantime, some photos!


Video here.

Radiolarian 10 – Radio Nova – 1/2013

The following article is taken from the original announcement of this attempt on in 1/2013.

Here is the 10th in my series of Radiolarian Puzzles. Radiolarian 10, aka Radio Nova. I started this puzzle in August of last year.

Radiolarian 9 is a nice place to be. Many pieces have disappeared, and the remaining pieces are all in the same order of magnitude of size. Radiolarian 10is not so fortunate. It is an explosion of pieces, so I’m calling it “Radio Nova”. The explosion happens as 6 cut lines cross the center faces, creating TEN NEW PIECES in each center. At the same moment, we must maintain the corner pieces for the last moment before the “starminx” like Radio Star (the upcoming Radiolarian 11), where the corners FINALLY disappear. Maintaining the old corners with the new center explosion makes this puzzle very, very much of an “in-between”.

I have wanted each Radiolarian to be the same size. I’ve counted on new pieces (that appear as cuts deepen) being offset in mechanism complexity by old pieces (that disappear as cuts deepen). It has worked well so far, but this Radiolarian has too much going on. If I keep the new pieces and maintain the old pieces, the parts get too small to be stable. In fact, new parts start to live in the same order of magnitude as fillets.


Still, since a truly stable version was estimated at $2500 to print, I had to try anyway. I’m not sure what I was thinking.

Even the inner layers of the puzzle look concerned.


Still, I forged ahead with the assembly and 10 hour stickering. (Some of the stickers are 1mm, and have to be individually superglued.)


Unfortunately, I have found that completing a single turn grinds the smalles pieces into nylon paste, which falls inside the puzzle to cause more problems.


To cap it all of fantastically, I placed the puzzle on my laser cutter, which has a slanted surface. As I folded up the “photo tent”, this happened:


I don’t have the appetite to reassemble right now so I bagged the pieces for a possible family photo in the future.

Sorry, there will be no video. :(

If there is anyone out there with $2500 to spend, I would like to extend the request for a sponsor for a much larger version of this puzzle which I believe can be strong and stable. If you are that person, PM me please, and we can discuss.

Radiolarian 9 – The Radio Crystal – 12/2012

The following article is taken from the original announcement of this puzzle on in 12/2012.

This is the 9th in my series of increasingly deeper cut face turning icosahedrons.

In the face turning dodecahedrons, as the cuts get deeper than a Megaminx, the centers shrink in size and the slices begin to overlap each other on the edges, creating new parts. Aleh’s Brillic, also known commercially as the Pyraminx Crystal, was the first example with cuts at just the right depth to make the face centers disappear.

The Radio Crystal is a face turning icosahedron, with the same property that the cuts are at just the right depth to make the faces disappear. Surprisingly, this has already happened once before with the three faces around the turning face on the Radiolarian 2. But the cuts on the Radio Crystal are so deep that they pass another set of face centers on the icosahedron- 6 more per face! Buried beneath each hidden face center, there is a stack of centers 3 deep, one fixed to the core, and two with with holes to allow a screwdriver to reach the screws inside.

Visually and conceptually to me, this puzzle is the closest icosahedral analog of Aleh’s Brillic and Praminx Crystal, so I’ve named it the Radio Crystal… aka Jason’s Brillic. :D

Turning quality is very good. Criticisms from me are that it is a bit squishy due to the acute angles, and that the brillic pieces sometimes trail slightly behind a turn and have to be aligned separately.


Video Here.

Radiolarian 8 – Radio Jam – 12/2012

The following article is taken from the original announcement of this puzzle on in 12/2012.

Introducing the Radio Jam! This is the 8th in my series of increasingly deeper face turning icosahedron puzzles.

Just as the dodecahedron shape has some nice canonical face turning cut depths like the Megaminx, Pyraminx Crystal, Starminx and Pentultimate, the icosahedron has some cut depths that are well proportioned and elegant. For example the Radio Web – Radiolarian 6, should be simpler than the Radio Jewel – Radiolarian 7, since the Web has shallower cuts. But the Radio Jewel – Radiolarian 7 represents a moment where some pieces disappear, leaving a pleasant, simpler outward appearance. Even though it has a deeper cut than the Radio Web, the Radio Jewel is elegant both in outer appearance and the inner mechanism because some earlier complexity can finally be shed.

Puzzles between the elegant stopping points have been called “in-betweens”.

The Radio Jam – Radiolarian 8 is one of these in-betweens. It has not let go of the complexity of the Radio Jewel -Radiolarian 7, but has also gained some new pieces: two very small trianges on each edge. This uncomfortable space between two more elegant puzzles reminds me of the region on a radio dial just between two stations, where you can get hints of both, but neither is totally comprehensible. This puzzle contains the pieces from the Radio Jewel, and the upcoming Radio Crystal, and the result is a puzzle stuck between two worlds.

I’ll call movement on this one fair. I think this puzzle would need to be made larger to improve it. Breaking in was also a nightmare this time. On to the photos and video!

By the way– a quick apology, these photos were taken with too high of an ISO, so they are grainy. On the bright side there’s less depth of field. Next time should be an improvement.

Jumble 1
Jumble 2
This image is a comparison to Radiolarian 3. You can see how deep the cuts are now.

Video Here.

Big Chop – Unsuccessful Stochastic Rails Experiment – 11/2012

The following explanation was posted on to explain my experiments with stochastic rails.



I was thinking recently about ways in which a Big Chop might be able to be made using 60 identical copies of 2 parts. Thinking about the cheats we sometimes use to estimate tough integrals in computer science, I thought of an idea. Each part would have one shell or rail at the same level. When two parts near each other, they collide and one randomly goes up while the other randomly goes down.

The parts are designed so that no matter which wins, the rails interlock.

I have designed 6 versions of the Big Chop puzzle using variations on this theme. I’m not at a stopping point, but I wanted to share my progress with everyone here that’s interested in mechanisms.

Here are a few images of a variation with wider rails, and holes cut in the rail arms to make them springlike.


You can see that they hold together, randomly fitting whichever way I happen to push them together like velcro.

Here’s a demo video demonstrating the velcro like quality.

I have tried multiple rail sizes and widths. Strangely thinner rails seem more successful. These images show the thinner rails:

I have also tried stochastic combs, which are collections of shells that randomly go one way or the other like “bristle blocks”. These seem to collide much more making the puzzle useless. But I think I’ll try again sometime anyway.


Here’s an assembled example using the most successful design to date.

Turning is just painful.
The stochastic moments all seem to want resolution by hand-jostling, even though small collections of parts work well.

Since turning is so tough I sped up the video 8x. Yeah, I was resolved to get one turn.

Big Chop using Stochastic Rails

Radiolarian 7 – Radio Jewel – 11/2012

The following article is taken from the original announcement of this puzzle on in 11/2012.

I’m very proud to present number seven in my series of “Radiolarians”, or face turning icosahedra.

This is an even deeper face turning icosahedron than the Radio Web (Radiolarian 6). No new pieces are added for this puzzle. They are only removed as the cuts get deep enough to hide them. You could say Radio Web is an “in-between” from 5 to 7 which contains both their pieces.
But as this puzzle simplifies relative to Radio Web, it gets a really beautiful pattern.

I’m naming this puzzle based on the Jewel Spider, which has an abdomen reminiscent of the sticker pattern here. Check out the last few images to see what I mean.

Also notable, is that this design gets some nice regular triangles again, which we’ve missed since number 4. (Eitan’s Star)

This is the second time I have attempted this puzzle, having designed and printed two separate designs.

This puzzle can jumble in two ways again, and I’ve included closeups below of the jumbles.

Video is here.

Movement quality is very good.

Thank you!




Radiolarian 6 – Radio Web – 10/2012

The following article is taken from the original announcement of this puzzle in 10/2012.

I’m very proud to present the sixth in my series of “Radiolarians”, or face turning icosahedra.

This is an even deeper face turning icosahedron than the Cat’s Cradle (Radiolarian 5), and 5 new tiny pieces are added around each corner by the deeper cuts.

This is the third time I have attempted this puzzle. There are many many tiny pieces as you can see! Each corner is surrounded by 10 small parts.

Like the Cat’s Cradle, this puzzle can jumble in two ways. Pictures of each jumble are included below, in the second and third to last images.

Thank you!






Radiolarian 5 – Cat’s Cradle – 10/2012

The following article is the original announcement of this puzzle on in 10/2012.

I’m very proud to present the fifth Radiolarian, the “Cat’s Cradle”.

This puzzle is the deepest face turning icosahedron built with icosahedral shape. The cut depth creates small groups of tiny fractured acute parts near the edges that remind me of the “Cat’s Cradle”, a game played with a loop of string criscrossing between fingers.

This is the second design and print of this puzzle, this one being a major redesign of the mechanism. With this version I am very happy with the movement, which feels excellent.

The radiolarians can all jumble, but this one adds a new jumbling move where three of the cat’s cradle pieces can rotate away independantly. The jumbling results are shown in the 4th (standard jumble) and 5th (deep jumble) images. (Can you find them?…) :D

Also, I made a very short video:

Video Here

Thank you!









Radiolarian 4 – 10/2012

The following article is taken from my announcement on the TwistyPuzzles forums on 10/2012.
The fourth puzzle in my Radiolarian series of face turning icosahedra (leaving out a minor one).

This is another puzzle I sketched up in 2010, but I never finished the parts or printed it, and I was beat to the punch by Eitan Cher (pirsquared). His version has been called DeFTI and Eitan’s Star. I’m showing this version with his permission.

My version has an identical mechanism to my Radiolarian 3, and only the outer cuts are modified to change the relative sizes of the parts. Turning is very nice.

I show the jumbling in the video.




Rhombicultimate – 10/2012

After announcing this puzzle it became clear that Tom van der Zanden made it first, in 2010.  Here is the original announcement of mine from 10/2012.



This is a shape modification of my Pentultimate (mechanism v2.5), using caps to create a rhombic triacontahedron.

This modification was first suggested to my knowledge by Robert Webb in 2003 here.

One interesting thing about this modification is that it doesn’t change shape as many other pentultimate shape modifications do. It’s always a rhombic triacontahedron after every turn.

Every piece orientation matters, as opposed to the Pentultimate where only the triangles matter, or the Icosamate, where only the pentagons matter.

At the end of the video I’m attempting to show how well the puzzle can turn by using only one finger with fast turns. It gets a little loud… :)

Thanks for looking!

EDIT: I have learned that TomZ made this puzzle first, in 2010.

Bronto Star – 9/2012

The following article is taken from the original post on in 9/2012.


This is another from my 2 year old backlog. Noticing the interior of the Alexander’s Star resembled my beloved Dino Star, I wanted to print some inserts that could convert an Alexander’s Star puzzle into a larger cousin to the Dino Star. The Bronto Star!

You can buy your own parts to modify one of yor extra Alexander’s Stars here:
Bronto Star on Shapeways

Please note, though, that as the Alexander’s Star is tricky to turn, so is the Bronto Star. So don’t expect to speedsolve it, but it can be fun to have alongside your Dino Star on display!

Youtube Video Here.





Cubeultimate – 7/2012

The following article is taken from the original announcement in 7/2012 on


This is a puzzle I had printed in 2010 and then set aside as I slipped into a work-related coma since then.
But I finally put it all together and thought I should share it, even though it’s just a shape mod of my Pentultimate (version 2.5 mech).

My Pentulimate design is already large, but extensions to make it cubical make for a very large puzzle. It’s about 4.25 inches on an edge, or 108 mm!

This shape mod would have to happen eventually I think, but I believe it was first suggested by Robert Webb about 10 years ago on this forum.

Thanks for looking!









Pico Cube 9mm – Briefly the World’s Smallest Rubik’s Cube – 4/1/2010

This puzzle was posted on April 1st with an image containing “photoshop” in the name to make people think it was an April fool’s joke.
Here’s a picture of the new Picocube design with a thumb tack!


Hey everyone! As Dave said, this puzzle has been in the works for a long time. It’s fully functional, and yep, it’s real. :) Finally…

I couldn’t resist posting it on April 1st and adding Photoshop to the name. :) The irony is that the image was actually cropped and resized with Gimp. So as far as that goes…. April Fools! :)

Anyway, here’s another picture! It’s 9mm on an edge.

I’m hoping to get a macro lens video going soon.

This cube is less than a quarter!
This cube is less than a quarter!



Here are some making of images that weren’t included originally.

Picture1 003 copy Picture1 008 copy Picture1 014 copy Picture1 017 copy Picture1 025 copy Picture1 034 copy Picture1 046 copy Picture1 049 copy Picture1 050 copy Picture1 051 copy

The Icosamate – 2/2010

The following is taken from the original post on on 2/18/2010.


Here is the Icosmate!

I have mentioned a couple of times that my original idea for the pentultimate was for the icosahedron shape. I have finally made one! I thought it would also be fun to include below some of my original sketches for the puzzle, before I knew of the name “Pentultimate”, and some of my original knucklehead musings and scribbles. One of the sketches is for a deep cut edge turning icosahedron, which I haven’t tackled yet.

Video Link



Pentultimate Version 3 – Skirting Rails Mechanism – 2/2010

I called this version 2.5 of the mechanism for awhile.  But since it works great, I’ve started just giving it its own version number, v3.

In the skirting rails mechanism, an internal void master version of the desired puzzle is made.  The master layer will be hidden and never allowed to turn.  The master layer will only be allowed to skirt sideways.

There are eight types of pieces including the caps and core.


Something about my camera during this period of time caused my images to be soft and blurry.  Sorry!

To enforce the rule that the rails can only skirt sideways, the void master layer is glued around the top center.  This takes care of five of the six master pentultimate rings.


Here the top center has been glued to it’s adjacent master parts (in white).  It will be attached to the core (in grey) and locked in place.


Here the fixed center and adjacent glued rails have been attached.

Picture1 040 copy


To enforce the ring of void parts on the equator, that aren’t locked in place by touching the top center, some blocking shapes are added to the core.  Shown here in blue is the small part that will be fixed to the core and stop the adjacent master piece from moving in the slice direction.


If you look again at the photos above, you can see the blockers on the equator of the core.

Here, half the puzzle has been assembled.  Since it’s a void mechanism, the core isn’t needed to hold these parts together.



Here the equator blockers are visible, as well as the feet that can interact with them.







Picture1 018 Picture1 020 Picture1 023 Picture1 024


Movement is excellent.





The Compy Cube – 1/2010

The following is taken from the original announcement on the TwistyPuzzles forumn on 1/18/2010.


This one is abased on the concept by Sausage for “Sausage’s Cube”. This puzzle was designed and built with Wayne’s permission. He suggested we come up with a new name, and we settled on “Compy Cube”, naming it after a famous little dinosaur.

It’s a very easy puzzle to solve which makes it a lot of fun! I think people who are intimidated by twisty puzzles may find this one fun and rewarding to play with.

I’ll be offering a very small run of these puzzles for sale soon. They will be 3D printed. As always, the mailing list at is a great way to find out first.



Radiolarian 2 – 1/2010

The following is taken from the original announcement on 1/11/2010.


The straight-cut Radiolarian is finally finished! This one has no centers which allowed me to add some nice corner and star-arm interior mechanisms (with none of the fudging used in the Cornered Radiolarian).

The prefix circo is a nod to Circogonia Icosahedra, the organism that isnpired the Radiolarian name.

I’ll also be calling it Radiolarian II.



Radiolarian 1, Cornered – 1/2010

The following is taken from the original announcement on


This is a pretty minor update. But as some of you may have imagined, the Radiolarian was designed to minimize the star pieces around the corners, in hopes of keeping both the fixed face centers and the corners.


Doing this involves minimizing the star parts on the surface and in the mechanism and then fudging the rest. For folks who have imagined this mechanism, 4 star parts anchor the corner and the 5th interferes. The result works, but turning quality is only poor.



Pentultimate Version 2 – Skirting Rails Prototype – 12/2009

This was my second attempt at the Pentultimate.  This version worked, but it was a bit catchy.

This was my first time using the “skirting rails” mechanism.  With skirting rails, there is a master version of the puzzle on the inside made using the “void” mechanism.


The master layer allows all the outer parts to move past each other.  The trick is keeping the master layer locked in place.  It should always be allowed to skirt sideways, but never be allowed to slide in the direction of the void mechanism tracks.

On the Pentultimate, there are 6 master pentultimate rings.  5 of these rings can be locked in place by simply gluing them to the top center face.  Here I am in progress gluing the rails in place around the top center.

Something about my camera during this period made these images soft or out of focus.  Sorry!



The core has special blockers on the equator that prevent sliding movement on the remaining 6th master layer.  They still allow the master layer to skirt sideways through the equator without blocking.  Thus, “skirting rails”.

Here I’m testing the mate of the rails with the outer triangles.


Picture1 065 copy IMG_0708

On this version, the blocker arms were very fragile.  As you will read below, they broke off and were replaced later with screws.  Version 3 was redesigned and built a few months later, and it works great.


Here you can see the equator blocker arms mating with the master layer.

Picture1 073 copy Picture1 074 copy Picture1 075 copy




IMG_0832_2 IMG_0840_2


Once assembled, the fragile blocker arms broke, and I replaced them with screws, which worked just fine.  Version 3, built a few months later, was a complete success!


IMG_0850_2 IMG_0888  IMG_0889


Picture2 001 Picture2 008


Oskar’s Tri-Gears – 9/2009

TriGears was offered for a short time at the website.  The original announcement is below.

TriGears represents a design collaboration between Bram Cohen and Oskar van Deventer.  Three gears meet with teeth meshed at 120 degree angles. The teeth on the gears are designed with varying thickness and spacing. Can you place them all on the holder so that they spin freely? There’s only one solution! All other configurations cause the gears to eventually bind up.

We have a very small number of these puzzles left over from a short production run for the IPP29 Puzzle Exchange.  We are offering them at $90 each, plus $10 flat rate shipping.  We have no plans to make more of these puzzles at this time, so we are limited to stock on hand.

Designed by Bram Cohen and Oskar van Deventer, and originally prototyped by George Miller.  Manufactured by Jason Smith and Adam Cowan.


The Twisty Store


Picture1 001 Picture1 005 Picture1 007 Picture1 005 Picture1 006 Picture1 011 Picture1 006 Picture1 006 Picture1 014 Picture1 044

Adam Cowan’s Gigaminx – 7/2009

The following post is taken from the original announcement on


These are some photos of the new Gigaminx designed by Adam Cowan. This project is a collaboration between Adam (design and CAD), Andrew Cormier (masters smoothed and tuned), and myself (molding and spincasting of parts).

More info on this puzzle soon, but we wanted to share a few photos now!






Andrew Cormier Petaminx – 2/2009


This article appears as originally written on the 2009 “Puzzle Forge” web site.

It was later featured on Gizmodo, Make and others, and the video has more than 1 million views.


The Petaminx is a face turning dodecahedral puzzle with 4 slices per face.


On December 29th 2008, designer Andrew Cormier posted free STL files for the Petaminx on the Twisty Puzzles forum.


I’d like to thank Andrew for making his design available, and congratulate him on coming up with a working Petaminx design!
Andrew is already well known for his Teraminx design and builds. Drew’s Teraminx Video


On Feb 25 2009, my Petaminx was completed.


An FDM print was done first to test fit all the parts, and ensure that the design would work at this size.


The core for the final puzzle was also included in this test print.



FDM prints have noticable scanline artifacts, but they are still very very faithful to the desired shape.



Once the FDM test print was a success, a print was ordered in polyJet.



Polyjet is higher resulution than current FDM machines. Very, very small scanlines can still be seen in the right light, but you can see in this image that it is a noticable difference between FDM (black on the left) and polyJet (white on the right).




Polyjet can have a rough texture if the support material wasn’t removed all the way. Take it from me– Do whatever it takes to remove all the support material before making your molds, or you will have a rough texture on every part you make.



The support material was removed by soaking in a basic solution and then brushing with a stiff nylon grout brush.



Circular silicone molds were made for spincasting. The molds are spun in a small machine while resin is poured in the center. This method can help alleviate problems with airbubbles.



These are the molds for most of the smaller parts in the Petaminx. This mold has 30 parts, and it had to be run 30 times, plus extras for rejects and parts used to make assembly jigs later.



Casting the required parts took a few days part time. While one mold was curing, I emptied and prepped another.



With some practice, I was able to reduce flashing considerably.



Each completed set of parts was stacked in a cardboard box for cleanup later.



Once all 974 parts were cast, they each had to be cleaned, sanded and prepared by hand. This process involves removing any sprue remaining, removing all flash, and sanding any rough areas.


Each part took 1 or 2 minutes, or more, to clean. Cleaning took about 30 hours.



The centers were cast with pilot holes only, and had to be drilled hollow. A dril-stop was used to set the depth on the appropriate sized bit, using a polyJet master as reference.



Drilling the holes with a drill press helped ensure that they were absolutely vertical.



Once all centers were drilled, the core was assembled.



Sets of parts were superglued together to form rigid bridges that were used as assembly jigs. They allowed me to stack parts safely until an entire face was completed, and then flip the puzzle over.




Assembly was about a 7 hour process, and I won’t narrate too much about it. Here’s a string of photos that give a good idea of the slow progress.












Yes, I finished assembling the puzzle in my car on a lunch break.


The puzzle was painted with gloss black before stickering. The puzzle contains over 1200 stickers, and each one must be placed by hand, one at a time. With some practice, I was able to sticker a face every 30 minutes. Total stickering time was about 6 or 7 hours.



I used Drewseph’s color scheme as far as my available Vinyl colors would allow, and then I improvised.


Here are some photos and video of the completed Petaminx in action!





Thanks again to Andrew Cormier for making this amazing design available for builders.


Here are some fun stats:


The puzzle has 975 individual parts.

Each face has 101 stickers, making a total of 1212 stickers.

20 hours on masters and molds.
12 hours casting parts.
30 hours cleaning up parts and sanding (!!)
7 hours assembling all 975 parts
6 hours stickering.

PicoCube 10mm – Briefly the World’s Smallest Rubik’s Cube – 7/2008

The following was posted in the original TwistyPuzzles announcement.

This terrible looking hand stickered 3x3x3 lump of playdoh-shaped ABS is 9mm on an edge. Here are some photos of the build.

At this scale, ABS is very flexible. Irritatingly rubber-like. Supports must be large, which takes away from all straight edges– you’ll see. In the end, most surfaces are touched with an exacto blade. I won’t work at this scale ever again, if I can help it. It just feels impossible to make a clean looking build.

The flexibility at this scale also produces “hairy” looking chips, which can be removed by brushing over them with a fingernail.

The core is made in two parts. With the very thin ABS, even screwing in the center caps could ruin the core. After 2 ruined cores, I made special jigs for gluing and driving screws. The larger jig is meant to hold the two halves together for gluing, while the smaller is cut back to allow it to be clamped while attaching 4 of the center caps. The screws are size #000.

Thumbtacks and drywall screws are included in some photos to help with visualizing scale.

Stickers were scissor-cut, and applied with tweezers. I want to scream just thinking about it.

It did work (I broke it already), so that’s something. But I’m really unhappy with the lumpy look. On the bright side, I’m on to my next project!!



















Pentultimate – Knucklehead Version – 3/2008

The following article comes from the original website.


In early 2007, I made a pencil sketch of a deep cut vertex turning icosahedron puzzle. I later realized this ground had been covered by others before in dodecahedron form, including Christopher Pelley, who coined the name “Pentultimate”.


The first Pentultimate was completed on March 19th at around 8 PM.

Here’s a little back story on how this puzzle was created.


The Pentultimate mechanism is made of of 20 triangular and 12 pentagonal shard-like pieces. Every piece has 12 built in female grooves at 12 different vertical levels.

The grooves line up when the pentagonal pieces and the triangular pieces are placed together as shown.

Each pentagon owns one level of the track, and has male tabs protruding from one single distinct level. The tabs fit in the grooves of all the other parts.

Since each pentagon has male tabs at a distinct level, any pentagon can pass by any other pentagon, and the male tabs won’t collide, being at different levels.

The male tabs (of 12 sizes) are modeled separately, and are designed to fit at their one distanct level perfectly. Shown here is the model for one of the tab levels.
Because the parts contained multiple complex undercuts, they were broken up into machinable slices. The slices were imported into Maya as polygon geometry, and custom software was used to lay them out efficiently for milling and also add post support geometry.

The software automatically positions the parts for conservation of material, and determines how many support posts to add based on the part size.

The triangles were done first. 13 different triangle sizes were modeled in Solidworks, as well as a triangular cap which adds the pentultimate vertex. Note the hole in each part. The hole has been carefully sized to thread to a #2 screw that will hold the stack of parts firmly together.

The triangles were machined from sheets of black ABS plastic. One sheet of triangles makes 8 full triangle stacks. One side was machined first, and then positioning pins were drilled and used to accurately flip the material over to machine the other side.


Each side took about a week for the machine to finish. Because of the long machining time, I decided to cast parts later to avoid machining all the pieces needed.

Once both sides of a row of parts were finished, the support bars were cut and sanded off.

This photo shows a few of the triangle master parts stacked up, with a penny in the photo for size reference.

The same procedure was followed for the pentagons. The 13 different pentagon sizes were CAD modeled and laid out in custom software with supports for machining. This sheet produces 3 full pentagon stacks. Only one set was machined though, once I decided to cast parts in urethane to save time.


Seeing how they stack up:

Lego bricks were used to build retaining walls around the parts, leaving 1/2 to 3/4 of an inch around the parts.


Soft white sulfur-free oil based clay was used to build the parting line. The clay separation must be as clean as possible to get flash-free castings later.


Once all the parts were prepared with clean parting lines and Lego walls, they were ready for silicone.

The silicone I used was a 2 part system with a 10:1 ratio by weight. After calculating the volume needed to fill my molds, I prepared the right amount of silicone.

After 24 hours, the molds were turned upside down, and the clay was removed. The silicone was uniformly coated with petroleum jelly.

Silicone will stick very well to silicone, so it’s important to use the right rubber to rubber release.

My first batch of pentultimate molds were unusable because I used a universal release, and my molds glued shut.


14 individual 2 part molds were created from the machined masters.

I used alumilite black to cast the parts. Electrical tape was used to clamp the molds without deforming the silicone much.

I rotated the molds through a 5 stage assembly line; warming, preparing, pouring, curing, and removing.

Once cured, the parts were removed from the molds, and flashing was removed with a razor blade. Each part was then re-drilled with the appropriate size screw hole.

Only 430 left to go.


Parts were sorted into tackle boxes to keep them organized. Casting the parts took a few weeks to complete.

The tabs were all machined, since only one sheet of parts would be needed. Tabs of same size were laid out in rows, so that mill tool sizes could be changed for the smaller parts.

The smallest tabs required a 1/32 inch end mill.


To hold the parts in straight stacks for fastening together with a screw, jigs were required. The jigs were designed in Solidworks, keeping their size under the size of a “two-by-four”.

Then the jigs were machined from soft wood stock.

The parts were stacked up inside the jig.


And the jig was clamped shut for driving the screw.


A jig was also made for the pentagonal pieces.


The pieces were stacked carefully inside.



And then a screw was driven through all the parts in the stack to fix them together.

Once all the parts and jigs were made, they were assembled with the correct size and length screw.

At this point, I discovered an error in machining, and realized that each triangle part was 0.13 mm too thin. The difference is very small, but the error stacks up.

Luckily, nylon washers were available at this thickness, so I used one between each triangle part.


Here you can see the spacing washers. If this had not worked, I would have had to start over.

Most of the parts slide onto the screw freely. Only the last 4 actually thread.


With the spacers, the triangles and pentagons line up as they should. What a relief!


Here, the first few triangle shards have been completed.

A week later, more were finished.


The pentagon stacks were assembled in the same way.

Pentultimate pieces can also be used to play Towers of Hanoi. Two puzzles in one.

Only one more pentagon to go!


Each pentagon gets male tabs inserted at one level.

They all must slide in at the same time as shown in this picture.


This pentagon is #2, with tabs at the second level.




The tiles were modeled in Solidworks, and then applied to a model in Maya as a test. A few iterations were required to get the thickness of the tiles right.

The acrylic was first surfaced to an exact thickness, so that the tiles would be held together by a very thin film of acrylic while being machined.

6 different colors of tiles were machined, making a colorful mess!


Once finished the tiles were carefully removed from the machine, broken from surrounding material, and sanded to clean up the edges.


20 triangular pieces and 12 pentagonal pieces have been finished.


I took them outside for a bit of silicone lubricant.


Here are a few pictures of different stages of assembly.









The finished result.






Here’s a video of the puzzle being demonstrated.




Since this is a prototype, there are a few problems with movement. The puzzle is sensitive to alignment. As you will see in the video, I turn it over in my hands before each turn to make sure the pieces are lined up properly. Once lined up, rotational movement is fair.

The pentultimate contains 820 total individual pieces.

416 of the custom pieces are cast from 26 masters.

152 of the custom pieces are machined in copies of 15 shapes

568 parts are custom made, counting the 72 tiles.

252 parts come straight from hardware suppliers.

In the end, all parts assemble to make only 32 moving parts.
For more discussion of this build, visit the Twisty Forums:


Picture1 043 Picture1 048 Picture1 050 Picture1 051 Picture1 055 Picture1 056 Picture1 063 Picture1 065 Picture1 068

Tetrex – A 3D Rubik’s Magic Inspired Puzzle – 9/2007

Tetrex – A 3D Rubik’s Magic Inspired Puzzle

The following is taken from the original post around 9/2007.


The first Tetrex puzzle was announced in late 2006.  I plan to revisit this design sometime soon.  In the meantime,  here are some photos and notes from the original prototypes!

The Textrex puzzle is made up of a series of connected solid polyhedra.  The parts are hinged in a way similar to the Rubik’s Magic line of popular folding puzzles.


The first prototype was built using 4 sided dice.  Grooves were routed into the dice using a tabletop router.  Then, the dice were painted and connected together with nylon filament.


The initial prototype proved that the idea would work.  I built more prototypes with different numbers of parts to help me decide what size of puzzle to make.

These prototypes showed that painting would not be an option due to rubbing between the pieces while hinging.  I decided to use clear machined polycarbonate tiles for the second version of the puzzle.



For the second prototype, I machined tiles in polycarbonate.  Each tile had to be sanded with high grit sandpaper, and then polished to a clear smooth finish.


The tiles were glued into pairs of 2 using acrylic cement.  Colored patterns printed on card stock were cut, folded, and inserted into 2 pairs of tiles, making each tetrahedra.  In this prototype, every face and corner are uniquely labeled, to help me determine interesting patterns.


Once all ten tetrahedra were assembled, they were ready for assembly!





Here are a few pictures of patterns made using the final puzzle:









Here is a short demo home video of this prototype in action.


For a few more photos and a bit of discussion, please see this forum tread: