Sheet 01prototype, not a product
A glove that tells your fingers what to do.A camera watches. A solver works out the next move. The glove taps the exact finger, a beat before you need it.
RoboFingers is an open research project into guided hands. The first thing I'm teaching it to guide is a Rubik's cube solve, and the plans are free for anyone who wants to build one.
index, right layer up
- R
- U
- R′
- U′
tap sent, nothing has moved yet
- thumb
- index
- middle
- ring
- pinky
- project
- RoboFingers
- drawn by
- Your name here, age 13
- stage
- Bench prototype
- plans
- Free, on request
Machines can see and think faster than us. Our hands are what's slow.
A phone can read a Rubik's cube in a quarter of a second and know every move needed to solve it. A beginner still needs a month, because knowing the move and making your hand do it are two different problems. Most skills worth having are stuck behind that gap: chords, knots, soldering, sign language, a decent tennis serve.
Screens are a bad way across it. You look down, you look up, you've lost your place. So RoboFingers doesn't put the instruction on a screen. It puts it on the finger that has to move, as a short tap you feel through a small motor in a fingertip cap.
The part I care most about is the ending. If the glove taps forever, I've built a crutch. It should get quieter every time you get it right, and then stop. That's the actual research question, and I don't know the answer yet.
A tap that lands exactly on the move is already late. A hand needs roughly a fifth of a second of warning, so every cue goes out early on purpose.
- step 1
See
A phone camera reads the cube's colours, or your hand position, forty times a second.
- step 2
Solve
A solver picks the next move. On the cube that's a known algorithm, not magic.
- step 3
Cue
The move becomes one tap on one finger, sent early enough to be useful.
- step 4
Check
The camera confirms the turn happened. Get it right often enough and the taps fade out.
Build one: the cube rig.
Five vibration motors, a microcontroller, a phone on a stand, and about forty dollars of parts. You put it on, hold a scrambled cube, and it taps your fingers through the solve. That's the whole first build.
| no. | part | cost |
|---|---|---|
| 1 | Coin vibration motors, 10 mm, five of them — one per fingertip cap | $9 |
| 2 | 3D printed knuckle bridge and five fingertip caps (STL files included) | $3 |
| 3 | ESP32 dev board plus a small motor driver, strapped to the wrist | $11 |
| 4 | 500 mAh battery, charger board, and an elastic wrist strap | $8 |
| 5 | Any phone on a $6 stand, running the camera app | $6 |
| 6 | Ribbon wire, heat shrink, velcro, and a standard 3x3 cube | $9 |
| Roughly, if you already own a soldering iron | $46 |
Where it actually is right now
- worksReading a cube's colours from one camera in decent light
- worksFive motors firing independently from the wrist board
- roughCue timing. It's early, but not consistently early
- roughThe fingertip caps. Version three still slips off my thumb
- not yetConfirming a turn happened without being told
- not yetFading the cues out as you learn. This is the whole point, and it's last
Build it with me.
I'm putting together a plan pack for the cube rig and sending it to the first hundred people who want one. It's free. There's no kit for sale and nothing to pledge. What I need back is people actually building it and telling me where it falls apart.
- Plan sheet
- One PDF, printable, wiring and dimensions on it
- STL files
- Fingertip caps in three sizes, knuckle bridge, wrist tray
- Parts list
- With links, and the cheaper substitutes that still work
- Firmware
- ESP32 sketch and the phone-side cube reader
- Cue map
- Which finger gets told what, and why I chose it
batch 01 — plans go out when the pack is finished
Everything else fingers could be told to do.
The cube is the easy one, which is why it's first: the moves are exact, a camera can check the answer, and nothing goes wrong if it's late. These are the directions I want to take it, roughly in the order I think I can manage them.
Learning an instrument
Chord shapes tapped onto the fingers that have to move, and the beat pulsed on the wrist so you feel the tempo instead of watching it. Most people who pick up a guitar quit inside a year, and it isn't for lack of wanting to.
sketchingFlying and driving RC
I fly planes and run boats. Trim and throttle corrections could arrive as a nudge on the thumb instead of a number on a screen you shouldn't be looking at.
want to trySpotting what you missed
A camera on a car or a bike notices the thing in the blind spot and points at it, on the hand already holding the wheel. A direction is a much faster signal than a beep.
ideaTyping and signing
Same problem as chords, different alphabet. Learning to sign or to touch type is finger placement plus rhythm, and both are things a glove can teach.
ideaGetting a hand working again
If cues that fade out can build a skill, the same idea might help someone rebuild one after an injury. That's far past anything I can test, but it's the reason the fading matters.
long way off
Who's making this.
I'm 13, I live in Bay Area, California, and I've been taking things apart since I could hold a screwdriver. Mostly planes, boats and cars — anything with a motor and a linkage. RoboFingers started because I got fast at solving a cube and realised the hard part was never knowing the moves.
I work on this in the evenings and at weekends. It's a real research project in the sense that I'm trying to find something out, and not one in the sense that anybody is paying for it. If it turns into something bigger, good. If it stays a glove that helps you solve a cube, that's still worth having made.
What would help most
- 01Someone who has driven small haptic motors before and can tell me what I'm doing wrong with the timing
- 02People with hands that aren't mine, to test the fingertip caps on
- 03A hand at cube-solving algorithms, so the move list comes out short as well as correct
- 04Anyone who has taught a beginner something physical and knows when to stop helping