Community challenge #12: Puzzles (closed)

Thanks! The two I made are 5.3x6.3 inches but they can be scaled up or down relatively easily by editing the original sketch. The new manufacturing model and arrange tools feel like a nice improvement, Product Design Online released a good tutorial that explains them really well.

Thanks for noticing the robots by the way, Ill probably make a few posts about them on here soon!

I also added the fusion and cut rocket links to the top of the post!


Gopher hole puzzle

Walnut pieces with a sapele box, both finished with Danish oil

This is a classic puzzle and the pieces could be made in a number of ways, but having the precision-aligned CNC holes is visually appealing, and pocketing them out felt safer than using my drill press. A machine fit for the “buttons” could be designed but I did not take that route.

“Top”: Gopher Hole No Thru.c2d (254.4 KB)
“Bottom”: Gopher Hole Thru Cut.c2d (243.3 KB)

The puzzle design

The main square is 2.55" × 2.55" × 1.4". I used a thickness and width of wood I had available, hence the strange dimensions. I am new to CNCs so I put the design together in Illustrator and exported SVGs into Carbide Create. I used a 0.25" end mill and a 60° Vee bit to bevel the buttons and holes. I also modeled the pieces in Solidworks for extra practice and renders. One could easily add more pieces but I feel 6+2 is an appropriate level of complexity. There is also one duplicate piece.

The puzzle has taken both of my test subjects about 25 minutes to solve. There is one satisfying and simplifying "Eureka" moment:

As an example, a piece with 3 buttons on one side and 0 buttons on another side should obviously be placed with the buttons into the holes at the top or bottom of the puzzle. There are two such pieces in this puzzle (#2 and #3) (and one more “lucky guess” piece: #5). With this clue, it may only take 10 minutes to solve the rest.

Making the puzzle & lessons learned

  • Care must be taken to mirror the pieces correctly for the top and the bottom.
    It was hard for me to visualize the pieces and match the buttons and holes properly. I accidentally cut one set of holes right on top of a set of buttons…

  • To orient the piece when flipping:
    For the top of the pieces, I referenced off the bottom-left of the SVG. For the bottom of the pieces, I flipped the board across y and referenced off the top-left of the SVG. This was enough precision for me!

  • The pieces are probably best cut out on the CNC.
    I planned to draw the outlining rectangles using a CNC pocket and then use a table saw to cut them out. I would recommend using the CNC and tabs to dimension everything! This should improve safety and reduce sanding time.

  • Plan your containing box ahead of time!
    Consider oversizing the bottom and top hole grids so they fit in rabbets. My box has simple butt joints with homemade walnut dowels. I haven’t included any “plans” for the box as part of this project.

(Forgot to take a picture when she solved it!)


I love this, intact I love just about everything about it, well done. And I feel you are going to be someone to keep an eye out for, welcome to the Community!!



How are your projects coming along guys ?
I did the roughing pass for my puzzle cube tonight (I edited my post above to include a few pics), that was interesting (#BambooRocks)


Perhaps fortuitous that you didn’t have thick enough wood in another species, the bamboo looks great!
Looks like the runtime was going to be about 21 minutes, what did thee actual runtime turn out to be? Which end mills did you end up acquiring for the depth of cut required?


First attempt in aluminum going! Using classic belt Z and first part came out pretty nice and clean, with some small burrs to sand off:

Edit: Testing multiple variants of toolpaths in the same piece…learned quite a bit:

A much better result from lessons learned (and the pieces fit)! :

The Z-belt is having no problem chugging through the Aluminum which has been pleasant to see. The one annoyance has been with how thin of an onion skin to leave without accidentally punching through and having the workpiece start to get pulled up (!). I’ve found about 0.002" to be working for the moment.

I’ll update this with pics and more as I get through it. I’m holding my breath and seeing what tolerances I can get for this to work. 40 more pieces to go!


I did not pay attention, it was a bit longer than that, around 30min total I would say, those retracts slow things down and I did not bother optimizing them.

The endmills…well I just got the one with the longest LOC available (30mm / 1.2") from my usual local supplier, it turned out to be a 3-flute 1/4" with a steep helix, normally intended for cutting aluminium, but bamboo is so well behaved that anything sharp will work.

One interesting note is that I initially generated the toolpath with “two-way” adaptive to reduce cutting time further, and got a very clean cutting sound during the climb moves, but high-pitch (not chatter) intermittent whine when doing the conventional moves, which must have been some kind of resonance with such a long stickout (a good 1.75" I would say). Not sure why I would get it only when cutting conventional, anyway, I stopped the cut and regenerated it with one-way adaptive, increased the optimal load by 30% (just because I did not want to increase cutting time too much, and I felt like I had some margin left) and ran that. Nice brrr-brrrr-brrr cutting after that.


Looking forward to seeing that metal puzzle, I commend you for trying aluminium AND tight tolerances :slight_smile:


Turns out using a super long 1/8" square endmill for finishing instead of a ballnose, on a non-calibrated machine, with a flaky G-code sender setup did not yield great results (who would have thought…). I edited my post in the spirit of transparency and showing what not to do, and I’ll try again! That is, if I can find a long enough 1/8" ballnose.


This was fun. I actually bought my Shapeoko to make puzzles. ‘Trap’ was last years Christmas puzzle and ‘Grip’ is a similar puzzle from the same designer, Tim Alkema, who graciously gave me permission to use his designs in this challenge. Grip which was done from scratch for this challenge (well, at least from the significant learning of previously building Trap) and Trap was slightly modified adding 6mm in height to make a matched set with Grip.

I like puzzles that appear to be simple with only a few pieces, but are not so simple in practice. Trap, with only 3 pieces, is a level 12 puzzle, meaning it takes 12 moves to remove the first piece, but it doesn’t seem quite as hard as that might imply. Grip is a level 15 puzzle and seems much harder with its two unseen marbles that must be manipulated to remove the pin.

Design information for these puzzles (and many others designed by Tim) can be found at: Here you can visualize the paths and obstructions internal to the boxes.

Trap: box.c2d (180.5 KB)

Grip: box.c2d (173.0 KB)


Trap, left, & Grip


The parts are milled on an MDF fixture of about 400mm x 200mm; the design size in Carbide Create is 396mm x 204mm because those are the nearest even multiples of the 6mm grid used for this layout. The fixture has pockets for the pins and an outline for aligning the box blank (held in place with painter’s tape and super glue). All milling was done with a Yonico 6.35mm (0.25”) 3 flute downcut endmill.

Start with a blank about 275mm x 95mm x 18mm (only the 18mm dimension is critical). After the box sides are cut, they are rolled up into boxes and assembled using rubber bands for clamps and CA glue. Trap also requires a 48mm x 48mm x 12mm bottom (not cut on the Shapeoko).

The pins are cut from blanks of either 60mm or 72mm lengths of 24mm square stock; cut in two passes with a rotation between. The fixture has space for 3 pins for Trap, but only 1 for Grip (this is not random, but necessitated by the shape of the Grip pin). The extra space included in the fixture is for extra blank stock placed to reduce tearout. I use a snug fit and a downcut bit to hold the pins in place.

If you look at my design file, you may well conclude that I’m nuts. You may be right, but the design choices were made to reduce tearout and hand work to square up inside corners.

For my first attempt, I simply drew the shapes of the unfolded box sides with stepped rabbets at the corners; drew an oversized outline around all four pieces and created tool paths for pockets at 3 different depths. This wasn’t bad and would probably be a good approach for plastic; however, in wood there was tearout at each outside corner where the cutter moved from end grain to long grain (LL & UR with the grain running along X). So, my next attempt was to draw overlapping negative shapes and mill the end grain areas before the long grain; this worked well enough, but further optimization by moving some shapes near the corners from one piece to the adjacent piece reduced the hand work required on inside corners and that is the version included here.

Speeds & feeds are 12000-15000RPM & 1500mm/min giving a chipload of 0.035mm (0.0015”). Depth & steps evolved as I went, ending up on 6mm & 1mm which results in significant chip thinning that was not compensated, but the latest cuts were in cherry with no burning and are probably adequate. It was still cutting chips.


Apply your finish of choice; add 22mm marbles; and have fun!


And, thanks again to Tim Alkema.


As time runs down, I should at least START the post for my entry.

It looks like it COULD be complete at this point, and is actually a rather tough puzzle for most people presented in this form, but there is more that I have yet to complete. I will add the design files later, but for the moment will leave a little mystery with where this is going, though I have no doubt that there are several of y’all that have a good idea. In the end, there is a twist, though.

The general form is interlocking rods. I used oak dowels here, though the final version will be larger and aluminium. The tolerances on dowels are not great for this. The structure looks like

in CAD. The form is the rods half-through at each intersection, kind of like Lincoln logs, but to make the pentagonal structure, the notches are at compound angles. This made setting up the coordinate system interesting to machine.


You can see that the rods have the notch, and that I left the axes of both.

And added a work plane ON the axis of the part being cut, PARALLEL to the axis of the intersecting rod. The origin in in this plane, on the axis of the part being cut, but this is not the center axially. This gives a really nice coordinate system to fixture the parts for machining, and allows the notch to have no overhangs.

This means the notches will be machined witht he end of the tool, and the finishing will be multiple passes with a ball-end. Not ideal for a great surface, but the other option was fixture the part tilted up, making trimming the ends square a challenge, or requiring extra setups.

The wells at each end are for trimming to length.

First, glue a riser piece onto the wasteboard (12mm isn’t enough thickness), locate the origin, deck it, and machine the fixture.

I set the tools to the minimum acceptable extension

leaving about 1mm between the surface and the collet nut

to get the rough-in

then finished uder the close supervision of the shop supervisor

After finishing with the ball end, threading the hole for the hold down (you can thread MDF for machine screws. I usually get about 100 uses before the threads give out, by which time the fixture itself is usually done, too. MDF is not real rugged in this application, but it sure is easy) The first part was mounted

The clamp was made from a special work holding alloy known as “scrap bin aluminium bar” using the HWF process (Human With File).

Cut to length

and notch

then repeat for all parts to get

which is surprisingly difficult for most people to assemble to get

This is a puzzle as is, but not much of one. There is more to come, if I can find time this week to finish the next part between all of the zoom meetings and online training (that would take two hours total in person, though it will be maybe 15 to 20 this way). The CAD is done, but I need to CAM up and machine it.

EDIT: Part 2
Had some time this morning. Meeting cancelled.

The actual puzzle is an assembly puzzle:

The five rods need to be assembled in the cage. The enclosure was machined in to identical parts, then assembled using dowel pins for alignment

The model for each half

with both halves machined from PVC bar stock (because I had it on hand). The stock prep was done with handsaw, and then one end faced in the lathe to make a good surface to tape down. The machining was done with moderately aggressive parameters using a 3.2mm bull nose tool and a 1.6mm square end to bore the dowel holes. The 3.2mm tool was 1mm radial engagement at 5mm axial, 2000mm/min, 10Krpm. Nice chips

dowel pins in one part

then assembled

The puzzle is, of course,assembling the rods in the cage. I have left out one thing, to this point. As shown to this point, assembly is impossible. This leads to the nature of slight-of-hand.

There are two more rods. One un-notched, the other with to notches, hence the fancy fixture:

The un-notched rod fits in the lower part of the fixture to position the other rod at the proper orientation for the second slot to be cut

With these two rods, the puzzle can be assembled in the cage. The slight-of-hand?

Show the assembled puzzle to the victim. Disassemble, and palm the un-notched and twice-notched rods, replacing them with extra one notch rods.

When the victim can’t do it, show them, again swapping the rods.

(This is an old puzzle, I think Adams solve a version back in the day, as well as many Hong-Kong sources. I bought two to make a single magic trick. Only good one I ever designed myself)

The Inventor models (from my education license): (489.5 KB)

(this is the minimum model set to make the puzzle. There are about a dozen models total to construct it and develop the CAM. The cage model will need to be cut in half with a plane to machine, and your choice of alignment method added-separate pins like I used, fingers and holes machined in, whatever- as the model is the full cage.)


I won’t have time for this challenge. :disappointed:

Anyone in need of a good idea, I was going to make this. :wink:


I don’t think I could make one of those! That inside corner looks to be 1mm or less and an inch thick. Also how to deal with taper? Please teach me :slight_smile:

1 Like

Wire EDM Shapeoko accessory?
That’s gonna be a funny looking Shapeoko when you hook this up to it. :stuck_out_tongue_winking_eye:


Teach all of us…PLEASE!

1 Like

Ha! Sorry to create the stir.


Sorry, I got all excited thinking you would be making this on a Shapeoko, per the rules :wink:
But it would be interesting to watch that process in action.


The Attribute Mazes have names based on guiding Dungeons and Dragons attributes. (Of course, any puzzle can be solved with enough Strength or Charisma). The objective is the same for all three mazes: guide the ball bearing from start to exit without turning the puzzle upside down. The bearing is 0.25" and the puzzle dimensions are 4" x 2.75" x 1.125".

  • WIS (Wisdom): a perception check; find the magnet to guide the bearing past the holes in a simple maze
  • DEX (Dexterity): a tricky spiral ramp must be navigated so the bearing can fall through the center and exit
  • LOG (Logic): a lever-activated double-trammel mechanism allows the bearing to avoid a T-shaped trap


I would like to continue making these types of puzzles with the attribute as a hint and puzzle names like WIS+1. I am especially proud of the trammel mechanism because I’ve never seen that in a puzzle before. I was inspired by the Trammel of Archimedes and wanted to figure out how to put it in a puzzle.

For each puzzle, the process was:

  1. Prepare stock at correct thickness
  2. CNC the puzzle base and lid (using only 0.25" end mill) using Shapeoko hardwood settings
  3. Create auxiliary pieces (ramp, trammel blocks, lever)
  4. (Drill dowel holes and) add dowels; attach lid
  5. Hand sand until satisfied (320 grit)
  6. Engrave puzzle name on lid with 60deg V-bit
  7. Finish with Danish oil and a coat or two of shellac


I had some fatal mistakes here in the final steps that have put this project into the “remake later” category. This should have been the easiest puzzle to make. The maze is simple but the ball will fall out the bottom if the magnet is not used. The maze must still be “mapped out” because moving the magnet in the wrong way will drop the ball. The magnet is hidden in one of the four dowels (with the other three glued in). Major mistakes:

  • I thought I could hand-drill a straight 3/8" hole through the bottom piece, guided by the top piece. I should have used a drill press or trusted the CNC.
  • My dowels are too thick for the holes; I should have taken the time to sand them down! I ended up cracking three out of four corners while hammering them in.
  • The first hole can be bypassed by quickly tilting the puzzle, giving the ball enough momentum. But I haven’t been able to build enough momentum to bypass the second hole, so I consider the holes well-placed.
  • I lost my magnets! They must be stuck to something metal because I can’t find them anywhere right now. Also, check your magnet strength before you put the lid on to make sure they are strong enough.


The trick here is to gently guide the ball up a spiral ramp and into the center hole, where it drops down and easily exits. I made the raised ramp out of MDF with the idea that it could be sanded and sculpted to make the task easier if necessary. My version is tricky even if you can see the ramp, so I will be adding a slight groove to the ramp using a chisel or carving tool.

  • In my first design I didn’t pocket quite enough space in the top piece to allow the ball all the way up. This was straightforward to fix with a chisel, but it definitely doesn’t look great.
  • A simpler design would probably be a better starter puzzle (for example, a round container with an internal, square wall and a single opening).



The double-trammel design here means a single lever moves the two wooden blocks in a slightly confusing manner, but the mechanism is visible through the holes in the boxes. Each block has two holes to allow the ball through. A shortcut can even be used to bypass half of the maze.

  • Solidworks was useful for the initial dimensions because the constraints are easy to set and test. Fusion360 may work just as well, but I’m not very skilled yet.
  • I used cut-off nails with wide heads to hold the trammel blocks to the lever.
  • If you know of any other puzzle that uses this mechanism (especially with more than 1.5 tracks) let me know – I’m very curious what else can be done!



It’s not much to watch. The workpiece is submerged in water with a splash guard over the top. all you really see is the wire feeding.

1 Like

Does anyone else have fun with naming their gcode files?

Finally making some headway on my entry after settling on some adjusted dimensions to get the parts to “smoothly” fit. I had a heck of a time getting the small piece tweaked to fit in the bigger one…Fusion made it look simple with plenty of clearance :neutral_face:.