The Eye Screw

A dispatch from the bench, September 2026. This one is a correction to The Sweet'N Low Packet, which is five weeks old and already partly wrong.

---

What that post promised

Post 08 solved wand blinds two ways at once. The good half was that wand mode stopped being a mechanism and became an install orientation: clock the body about thirty degrees around the tilt rod and the drive axle points down toward where the wand hangs, the way a native wand tilter's input shaft is already inclined.

The other half was a part. A two-piece snap knuckle, each half carrying a pair of balls, snapping together at ninety degrees into a joint that bent almost thirty degrees in any direction with zero rotational slack, popped apart under an overload yank, and pressed back together by hand. Two identical printed parts, no fasteners. It was the nicest thing I have drawn on this project, and it turned out to be the kinematic heart of a joint Carl Weiss patented in 1923.

It is not in the product.

What replaced it

The manual stub on the end of the main axle now carries a pilot hole. An eye screw threads into it, your existing wand hooks into the eye, and that is the entire wand attachment. The pilot is there so the screw cuts its own thread without splitting the printed stub, and because it runs in and out on that thread, you can wind it to wherever your wand's hook wants to sit.

Printed parts in the wand path: two, then one, now zero.

The part, for anyone copying this: a plated steel screw eye, 25 mm overall, 12 mm across the eye and 7 mm inside it, 10 mm of self-tapping thread at about 2.5 to 3 mm. A thousand of them cost $16.99, which puts the wand attachment on this product under two cents.

The whole wand connection. It was on the drawing the whole time as a slot I had stopped looking at.

Why the simpler one is also the better one

It has no fit window. The knuckle's ball-and-socket pairs did. In August the dowel holes went from exact diameters to a taper, because fit testing showed the gap between too tight — filament dowels buckling — and too loose was narrower than my printer's own repeatability. A screw in a pilot hole has no such window, and its diameter is not critical.

Nothing to validate. The snap forces were computed and the prints were on the bench waiting to be measured against the math. That test is cancelled rather than passed, which is the cheapest outcome.

Nothing to stock, nothing to print, nothing to break. No SKU, no STL, no delamination mode, no layer line in the torque path. Lose one and it is a hardware-store part.

What the knuckle was buying, and what it wasn't

It was buying angular slack with no rotational slack. A dangling connection cannot carry twist around a corner, so if the axle and the wand point in meaningfully different directions, something has to bend without winding up.

Here is the part that stings. The knuckle bent almost thirty degrees. The eye screw and a wand hook tolerate something like seventy. The part I designed, computed and published was the less forgiving of the two, and the more forgiving one was already in a bin at the hardware store. The install clocking gets the axle within a few degrees of the wand's hang anyway, so seventy degrees is margin I will never spend.

Third time

This journal's recurring lesson, for the third time: every time this machine got better, it got better by losing parts. Post 05 deleted a finished twenty-five-part machine. Post 06 deleted a finished gearbox over one option box on a cut sheet. This one deleted my favorite part, five weeks after publishing it.

The install chamfer stays. The clocked mount stays. The spool stays for cord blinds, unchanged. Everything that was geometry survived; the thing that was a part did not.

The TiltRod, by TiltForge, is a retrofit smart blind motor — repairable, open, Matter-native, and the cord always works. Founders pricing on the first 100 units; first units ship November 2026. Check your headrail at tiltforge.com/fit or follow the build at tiltforge.com/#waitlist.