Blog/2026-01-11/Modeling With Claude: Difference between revisions

From Rest of What I Know
Added page description via AutoDescriptor bot
No edit summary
 
Line 34: Line 34:
== Results ==
== Results ==


[[File:Video - Power Grid Power Plant Card Display.webm|center|And this is how it snaps together]]
[[File:Video - Power Grid Power Plant Card Display.webm|center|frameless|upright=2|And this is how it snaps together]]


Claude delivered on pretty much all of them, though the lips on the left and right are a bit shallow, and the scallop is instead a big gouge. The magnet holes worked out perfectly. You can get the 3 mm x 1.5 mm magnets on Amazon pretty easily, and Claude added a little bit of slop so I could slather on some CA glue on the rim of each and shove it in the hole to dry.
Claude delivered on pretty much all of them, though the lips on the left and right are a bit shallow, and the scallop is instead a big gouge. The magnet holes worked out perfectly. You can get the 3 mm x 1.5 mm magnets on Amazon pretty easily, and Claude added a little bit of slop so I could slather on some CA glue on the rim of each and shove it in the hole to dry.

Latest revision as of 18:57, 11 January 2026

The little magnets allow you to connect the different sections temporarily

We've been playing quite a bit of Power Grid and one of the things that annoys me is when you have to pick up a power plant or move them around. I always have trouble picking one up from the second row and putting it in the first or picking up a plant because I can't put my finger under the damned card!

Past Related Stuff[edit]

Fortunately, we have a 3D printer so it was just a matter of creating the most obvious thing, a little flat surface to host the card on with finger divots. When I first started building things with my 3D printer, I used Claude Code with Python in order to build the model.

I use Claude for all sorts of things: code (obviously), creating the markup for wiki template uses, research, and also to model! And recently I'd read somewhere about using OpenSCAD as a way to represent the model. The problem with Claude and Python is that Claude attempts to write code to create an STL file, but the primitives it operates with are not particularly useful. It still gets there eventually but you frequently end up with self-intersecting meshes and so on.

This Time[edit]

This time, I have to say that the OpenSCAD approach works. With a few verbal descriptions of what I want, Claude was able to produce a small ramp. With some screenshots and instructions it was able to produce a lip, split it into 3 sections, and add the scallop for a finger.

The one thing that Claude (even with Opus 4.5 Max and the ability to create PNG renders to look at the intermediate output) failed to do was to put the little holes for magnets in the right place.

Frequently, the holes for the magnets would intersect with one or two of the flat surfaces. That sort of ruins the whole thing.

But even that was pretty easily surmounted. If you get Claude to use Python to model the angle of the ramp and where to place the holes so they don't intersect, it will figure that out.

Constraints[edit]

Anyway, I wanted a few things in the model and Claude delivered them all:

  • I wanted a bit of a ramp so that the cards would slide in one predictable direction
  • I wanted a lip on the left and right to prevent movement off the ramps
  • I wanted a lip at the lower end of the ramp to prevent cards from falling off
  • I wanted a bit of a scallop so I could put my finger underneath and pick up a card
  • I wanted it in a left 2-wide, a single wide, and a right single-wide so that we could shift from the 4x2 plant auctions to the 3x2 plant auctions neatly

Results[edit]

And this is how it snaps together

Claude delivered on pretty much all of them, though the lips on the left and right are a bit shallow, and the scallop is instead a big gouge. The magnet holes worked out perfectly. You can get the 3 mm x 1.5 mm magnets on Amazon pretty easily, and Claude added a little bit of slop so I could slather on some CA glue on the rim of each and shove it in the hole to dry.

In the end, the whole thing worked out quite well!

Bloopers[edit]

Polarity[edit]

Not everything went super well. I made the classic mistake of everyone who does this: I didn't keep the polarity right for the magnets. So what happened was that two of the surfaces would push against each other rather than pulling together. Rookie mistake!

I had a trick to make sure I'd get the polarity right. I put the first one in, then I put a second one on top of it and slid it aside till it was in the second hole. Then I'd put the next one on top of the first one and then place the hole on top of it and move it away. In this way the polarity is always aligned between the meeting surfaces.

I lost 3 magnets and two prints to this because the left-side and the middle, and the left-side and the right-side were both misaligned. Whoops!

CA Glue[edit]

Cyanoacrylate glue is pretty good for stick the magnets in their holes, but it works better on the side surfaces than on the poles. My first few tries, I had it on the far pole, but it would detach every time.

While I was doing the polarity trick previously mentioned, the extra magnet I was just about to stick in a hole would be sticking to a magnet that is already in a hole and it's the only one sticking out of that surface, so I could easily paint the outside with the superglue tip and shove it in the opposing hole.

Worked like a charm!

CA Glue is apparently not the best choice for PETG, but there was no way I was breaking out epoxy for this.

Unfortunately, while doing this, I managed to stick 5 of the magnets together. They were both magnetically and adhesively connected and there was no prying them apart.

Clumsiness[edit]

The magnets are tiny! 3 mm x 1.5 mm is very small, and my tweezers were magnetic which makes the whole thing annoying because the magnet will flip over trying to stay with the tweezers rather than stick in the plastic. I used a little flat cardboard box to rub and push the magnets in place, which worked well.

What didn't work well was showing Julie the magnets and then dropping one. It took us 5 min to find it again but we absolutely had to find it since it was in the living room where Astra does play.