TOGOAT · build log Race · Files
Rubber band distance car · 2026

TOGOAT

Earnie's rubber band racer: a 3D-printed lattice chassis drawn after a 2015 design-school car, driven by one folded band through a 2:1 pulley.

Render of TOGOAT from the front three-quarter: an open lattice chassis in black and gray around a central tube, deep-dish black wheels with blue lips, a cone nose.
RenderThe car as printed: black and gray PLA chassis, black wheels with blue lips. The real tires are clear TPU; the render shows them black.
DriverEarnie
DesignEarnie, his dad, Claude
Roundsv1, v2, v3
PrintedSept 30 to Oct 1
RaceFri Oct 2, 2026
01

The brief

Build a car powered by rubber bands. It has to travel at least 5 meters in a straight line, and the longest run wins. Along the way, explain it: the simple machines inside it, Newton's three laws, the friction that slows it, and measured time, distance, velocity, acceleration, mass and force.

Earnie picked the reference: the Cirin, a rubber-band RC car that three Art Center College of Design students (Max Greenberg, Sameer Yeleswarapu and Ian Cullimore) built in 2015. Its frame is a one-piece printed nylon lattice wrapped round a carbon tube, closer to bone than to a kit car.

He directed the design through an AI model, Claude, and named it AIbert. The lattice shapes, the physics model and the print files came out of that conversation; the reference, the name, the colors, the build and the test runs were his. A few years ago a chassis like this was work for a design studio with industrial software. Here it took a weekend of conversation and a shop full of printers.

Two side views stacked: above, the Cirin's own gray CAD render; below, TOGOAT v3 at the same wheelbase.
ComparisonAbove, the Cirin's own CAD render (Greenberg, Yeleswarapu, Cullimore, 2015, from the project page). Below, TOGOAT v3 scaled to the same wheelbase.
02

v1: a flat lattice

The first version was a flat Voronoi-cell floor with a winding hook on the rear axle, the band hooked straight onto it. On Sept 29 it went through four independent reviews: collisions in the mesh, the drivetrain physics, the print process, and a check against the Cirin itself. Each finding was ranked by how much distance it would cost.

The worst one was simple geometry. The winding hook's tab hit the deck, so the axle could not turn a full circle. Of 57 hook positions checked, none cleared. That car goes nowhere.

The second worst was physics. The band wound onto a 13 mm collar would ask the tires for 2.7 N of push when they can hold about 0.45 N before spinningE. Most of the band's energy would have gone into wheelspin.

Four gray views of the v1 car: a flat lattice floor with round lattice wheels.
v1Flat lattice floor, lattice wheels, band on a hook.
FindingWhat it costsThe fix
Winding hook hits the deck; 0 of 57 positions clearThe car goes 0 mA 16 mm keep-out round the rear axle
Band on a 13 mm collar asks for 2.7 N against 0.45 N of gripEAbout 80% of the energy lost to wheelspinWind a thin string on the bare 3 mm axle instead
Nothing makes the band let go at the end of the runThe axle rewinds it and brakes the car: 30 m falls to 6.6 mEThe string ends in a loop over a short rounded peg, so it slips off
Rubber-sealed bearings drag more than printed bushings30 m falls to 4 mEShielded bearings, degreased
Set screws slip on round rodWheels spin on the axleFile a flat where each screw bites

E marks a number from the physics model, an estimate. Real floors give less.

03

v2: drawn from the inside out

v2 fixed every v1 finding. A braided bio-truss shroud wrapped a printed tube, the band stretched inside the tube, and a string wound onto the bare axle. Every hole line was probed so no strut crossed a pin.

Then, the night of Sept 29, it went beside the Cirin at the same wheelbase, and it lost. It looked like a carrot on wheels of swiss cheese. The numbers said why: the Cirin's wheelbase is about 3.5 tire diameters, v2's was 1.6; its tires are twice as tall as they are wide, v2's were seventeen times; its side view is a long low torpedo, v2's was a cone rising into the middle of the rear wheel.

The cause was the order of work. v2 started with the drivetrain, the tube first, and wrapped a braid round it. The Cirin started with the silhouette.

Render of v2: a white braided truss cone around a tube, thin disc wheels full of round holes.
v2The braided truss. Sound engineering, wrong silhouette.
04

v3: silhouette first

v3 starts with the outline: an F1 stance, a torpedo side profile, and a plan view shaped like a T, with broad arms at the front axle, a narrow waist and a rounded tail. A skin sits just under that envelope, with big lens-shaped openings wherever the loads leave room and tighter cells where they come in.

Measured off the Cirin's top view at the same wheelbase, its plan is not a straight taper. It pinches to a waist about 28 mm wide each side, then swells to 41 mm near the rear. That waist and swell is the gentle S-curve in the outline. Round 14 moved the rear seats inward by 11 mm so the rear wheels stand off the body on bare axle instead of packing against a wide tail.

Line chart of half-width along the car's length: dots for the Cirin, a blue line for round 13 and a red line for round 14, both following the Cirin's waist and rear swell.
DiagramPlan half-width from the rear axle to the nose. Dots: the Cirin, measured off its own top view at our wheelbase. Blue: round 13. Red: round 14, the shape that was printed.
Side render of v3 in white: a low torpedo lattice between six-spoke wheels, the top of the body below the rear tires.
SideThe torpedo. The body's top sits below the rear tires.
Top render of v3 in white: a T-shaped plan with arms to the front wheels, a waist, and a narrow tail between the rear wheels.
TopThe T, the waist, the tail. Rear wheels stand out on bare axle under printed sleeves.
Rear three-quarter render of v3 in white with deep-dish wheels.
RearDeep-dish six-spoke rims, rear only slightly larger than the front.
Close render of the four-part nose cone: base, two thin rings and a cap with a flat tip.
NoseFour parts, all cones and cylinders, so each one can be rebuilt in TinkerCAD from a table of sizes.
Technical drawing: cuts through the chassis along every pin and axle line, each marked clear.
DiagramEvery line the car needs open, the axle, the pins, the string path, cut through the chassis and checked. All 24 probes came back as designed.
05

The drive

Grip sets the limit. The floor can only push back on the rear tires so hard: the friction coefficient times the weight on them, about 0.91 N for a car near 300 gE. Any pull beyond that spins the wheels and turns stored energy into heat.

So the band hides inside the spine tube and pulls on a shuttle. A string ties off at the tube's rear mouth, runs forward round a bearing on the shuttle, and comes back out to the rear axle. That is a 2:1 pulley. The axle feels half the band's pull and takes twice the string, which keeps the launch pull under the grip limit and spreads the same energy over twice the distance.

The rear tire is 76 mm across and the axle is 3 mm, so each centimeter of string moves the car about 23 cm. Through the pulley, each centimeter the band shortens moves it about 46 cm.

The string's free end is a small loop dropped over a peg on the axle. When the band runs out, the loop slides off and the car coasts. Without that, the spinning axle would rewind the string the other way and brake the car.

Model distances, one folded #33 band: about 45 mE. Two folded #33s side by side: about 74 m, with launch pull still inside the grip limitE.

AnimationWind, release, coast. The readout tracks turns, shuttle travel and the 2:1 ratio.
Interactive · drive modelOpen full screen ↗
Drag to orbit. Labeled parts, explode slider, and the winding cycle step by step.
06

Printing it

The parts printed over two days on whichever machine was free, and most of the print decisions got made by what broke.

  1. The chassis planOne piece on a Stratasys F170 in ABS, with support that dissolves in a heated sodium hydroxide bath. The machine had red ABS loaded, so Earnie chose his colors around a red body: blue wheels and nose rings, a white badge.
  2. DroppedThe support bath turned out empty, with no solution on hand. The F170 route went. The chassis moved to a Bambu X1C in PLA, with tree supports.
  3. The support fixHand-picking PLA tree supports out of a lattice is slow and tears the part. The fix was a PETG layer where each support touches the chassis. PLA does not bond to PETG, so the supports lift off clean. The cost: 244 filament swaps and a print that grew from 7 h 27 m to 13 h 42 m.
  4. Wheels, attempt oneLightweight foaming PLA with a blue lip. The rear wheels came off the bed mid-print and balled filament round the nozzle. The cause was a small footprint holding a 38 mm tall part, no brim, heavy cooling and a cool bed. Reprinted in plain black PLA with an 8 mm brim and a hotter bed; both printed clean.
  5. Tires, a spareA spare tire sliced standing on edge tipped over at home. Lesson kept in the files: tires print flat.
  6. BearingsThe online bearing order was canceled and no local store stocked them. Seven bearings came out of idler pulleys in a spare-parts bin. They were 693s, 8 mm across, where the pockets were drawn for 10 mm 623s, so printed adapter rings make up the gap.
  7. BadgeRoyal blue letters on matte black rendered fine and read badly in person: two dark colors at about 2:1 contrast. The badge went to white letters with a thin blue keyline.
Photo inside a 3D printer: one half-printed clear tire and a large tangle of loose filament on the bed.
PhotoThe spare tire that printed standing up, tipped, and kept extruding into air.
Close render of the TOGOAT badge: white letters with a thin blue outline inlaid in a black plate on a red tube.
RenderThe badge after the fix: white letters, blue keyline, black plate, shown on the red body that was planned. The keyline also hides any color bleed at the letter edges, since the badge prints face down.

Colorways Earnie weighed

Red chassis with solid blue wheels.
PickBlue on red: his choice, made after seeing the red body. When the chassis moved printers, the body became black and gray.
Red chassis with black and blue two-tone wheels.
As printedBlack wheels with a blue lip, the version that got printed.
Red chassis with white wheels.
OptionWhite wheels.
Red chassis with black wheels.
OptionBlack wheels, from stock already loaded.
Top render of the as-printed car: gray lattice with a TOGOAT badge on the tube, black and blue wheels.
As printedTop view, badge on the tube between the struts.
Small printed parts laid out: drive peg, shuttle, collars, spacers, sleeves.
HardwareDrive peg, shuttle, collars, sleeves and spacers. Rendered blue, printed in black PETG.
07

The build

AnimationExploded and reassembled.
AnimationThree build steps: rear seat bearings, front wheels, rear axle.

Twelve steps, in this order: clean the supports off the chassis; glue the nose stack; seat the badge; press the rear bearings into their adapter rings; stretch the tires on and seat the front bearings; mount the front wheels on pressed-in stubs; thread the rear axle through wheel, sleeve, bearing, drive peg, bearing, sleeve, wheel; spin-test every wheel; load the shuttle with its band and string; pin the band, tie the string's dead end and drop its loop over the peg; release-test with the wheels in the air; then floor runs along a tape line, one change per run.

Interactive · build bookOpen full screen ↗
Tap a step: its parts fly into place, earlier parts fade back, later ones hide.
08

The night before

The car was built by Thursday evening. It did not go.

  1. First testThe fishing line tangled on the axle and the car barely moved. Nothing held the winds in place on a bare 3 mm rod, so the line walked and knotted.
  2. A spoolA snap-on spool for the axle was modeled and sent to a red team before printing. It found two weak points, so it became a sealed, press-fit spool instead. Measured against the CAD, it only fit on one side of the peg.
  3. Wound by handBefore the spool printed, a hand-wound test: the car crawled about two feet and stopped, with the tension gone. Fine with the wheels in the air, dead on the floor.
  4. By eliminationEvery rotating joint is a fit on smooth steel rod, and a photo of the axle showed monofilament on bare polished steel with nothing to grip, plus a strand that had sprung out through the lattice. Those were real, and none explained a car that ran in the air.
  5. The likely answerThe design calls for one #33 band, doubled: folded in half, four strands, a 4.5 cm loop across a 4.5 cm gap between the nose pin and the shuttle. The car had two bands, unfolded. A 9 cm loop across that gap is slack for the last third of the shuttle's travel, and pulls about half the designed force for the rest.

One word, doubled, read two ways, cost the whole night.

The fold costs nothing and the design assumes it, so it went first on race morning. The other live suspect was a wheel slipping on its axle under the car's weight. A sharpie line across each rear hub and onto the axle tells the two apart: if the marks split after a run, the hub slipped.

The lesson for the next build doc: say exactly what doubled means.

09

Race day

Result pending

? m

Friday Oct 2, 2026. Measured distance goes here after the run.

RunChange madeDistance (m)Time (s)Avg velocity (m/s)
1
2
3

Model prediction for comparison: 45 m on one folded bandE. The minimum is 5 m.

10

Files