ThinglingsBack to the playground
Build journal · Updated 9 November 2026

A small box.
A very big possibility.

Independent wireless parts. Everyday materials. One shared software language that makes a child’s creation move, react and play.

Keep the ambition.
Make the first proof concrete.

Start with two independent wheel modules and one sound-and-light module. Give each its own radio and AA battery pack, then let Android bring them together as one creation. A new attachment should become a new ability.

The long-term kit is a family of small self-contained modules, each with an identity, a capability and its own power. The phone discovers them and coordinates the creation. There are no wires between finished modules.

The first experiment, now ordered

  • Two movement modules: a geared wheel, driver, Waveshare ESP32-C3-Zero and three AA cells in each. Prove forward motion, turning and reliable stopping.
  • One sound-and-light module: a third ESP32-C3-Zero, an amplifier, a speaker and the board’s RGB light, with its own three AA cells.
  • One Android controller: discover and assign modules, coordinate movement, choose lights and trigger sounds. Software checks have passed; physical Bluetooth and motor tests are still to come.
  • Later experiments: a moving joint, sensors, paddles and guarded fans. These are future research, not part of the first order.

Prototype boards are tools for discovering the software and mechanics. They are not the intended production electronics or the eventual toy’s retail price.

A vocabulary of physical possibilities.

Many different creatures can share the same few mechanisms. A position-controlled hinge can become a jaw, a tail, a rudder, a waving hand or a pair of wings.

Roll

A motor and reduction gears turn a wheel. Four pods make a truck; two can make a wobbling creature.

First experiments

Wag & hinge

A small servo sets an angle. Use it for a tail, jaw, door, ear, rudder or sail trim.

Future experiments

Grip & lift

A servo with a linkage closes claws; a geared spool becomes a little crane or drawbridge.

Mechanical attachment

Push & pull

A crank or linear linkage extends a neck, opens a hatch or makes a cardboard character bow.

Mechanical attachment

Buzz & shiver

A vibration motor gives a creature a heartbeat, a shiver or skittering motion on a suitable body.

Future experiments

Glow

Small coloured lights become eyes, headlights, signals or a creature’s expression.

First experiments

Speak & sing

A piezo makes chirps cheaply. Recorded voices need a speaker, amplifier and audio-capable electronics.

First kit: amplified sound

Feel a touch

A button or touch sensor lets a child wake, pet or surprise their creation.

Future experiments

Sense a tilt

A motion sensor can detect a lift, a turn or a change of posture.

Future experiments

Notice nearby things

Distance sensing can trigger a stop, a surprised sound or a cautious movement.

Future experiments

Hear

A microphone can detect a clap or sound level. Understanding spoken instructions happens on the phone or through a voice service.

Next software step

See

Start with the phone camera for object geometry and tracking. A separate camera pod is possible but adds power, cost and data traffic.

Phone first

Paddle

Replace a wheel with a paddle adapter. Keep the motor above water and animate a bottle raft.

Reuse a wheel motor

Spout & jet

A miniature pump can make a whale spout or a fountain dance. Test whether its water jet can propel a tiny hull.

Thrust needs measuring

Blow & float

A guarded fan can push an airboat or steer a neutrally buoyant helium creature.

Mass and thrust check

Flap

A servo and crank can flap a paper butterfly or dragon wing. A balloon or stand can support the body.

Motion first

Hover & fly

A skirted hovercraft or a decorated flying core adds lift. Free flight needs a matched airframe and local stabilisation.

Dedicated expansion

Swim & dive

A sealed fin or thruster body needs buoyancy control. Use a surface radio buoy, tether or stored routine underwater.

Sealing and control research

The first experiment kit.

Three independent radio modules. Two wheels. Sound and light. Reusable parts to test the software before miniaturising the finished pods.

Parts ordered on 9 November 2026. This is an adult-built bench prototype awaiting assembly and physical tests, not a finished toy or an orderable Thinglings kit.

The ordered kit uses three Waveshare ESP32-C3-Zero boards, two geared wheels with drivers and one amplified speaker, powered by three separate three-AA battery packs.

Commodity scale

The WCH CH572D combines a processor and Bluetooth radio in one chip. LCSC’s indexed catalogue lists about US$0.21 at 1,000 units. That is a bare chip, before a circuit board, motor, power and enclosure. Price source · Manufacturer capabilities

Prototype scale

A ready-to-program board costs more because it includes a board, connectors and supporting parts. It lets us test software before designing tiny production electronics. A complete low-single-pound wheel pod remains a manufacturing target to validate, not a supplier quote.

Leave room for land, sky and water.

A protected fan pod could push a raft and later steer a helium creature. A servo could become a jaw, rudder or wing. The parts below are research leads checked on 9 November 2026, not purchases or tested expansions; prices and availability may change.

Future air mechanics research

Bench thrust, airboat, then airship
Part and supplierQuantityListed unit priceLine totalWhat to check
DFRobot 7 × 16 mm coreless motorsThe Pi Hut · Pack of four · Listed in stock1 pack£8.20 inc VAT£8.203.7 V nominal, 0.8 mm shaft. Measure loaded current and thrust.
Gemfan 1219 31 mm propellersFlying Tech · Four-pack · Select 0.8 mm shaft1 pack£2.50 inc VAT£2.50Matching shaft size; chosen variant stock and motor/prop performance need confirmation.

Listed motors and propellers only. Reuse compatible controller and driver nodes. This is not a complete flyable kit.

£10.70

Also needed before the air experiment

  • Full propeller guards, mounts, a suitable lightweight balloon/envelope and helium. Allow £15–£35 as an estimate; obtain a local fill quote.
  • Suitable power, drivers and a weighed design. Do not assume a small party balloon can carry the electronics.
  • For unsupported quadcopter flight, use an onboard flight controller and matched frame. A BETAFPV Matrix brushed controller is listed from US$34.99, but is a later option, requires additional parts and is not directly phone-Bluetooth controlled.

Future water, fin and sail research

Raft, rudder, fountain, wagging fin
Part and supplierQuantityListed unit priceLine totalWhat to check
Kitronik 3 V submersible pumpsPart 25110 · Listed in stock2£2.76 inc VAT£5.52Pump stays submerged. Flow is not reversible; jet-propulsion thrust is unmeasured.
6 mm ID pump tubingThe Pi Hut · Accessory on pump page1 metre£1.50 inc VAT£1.50Confirm outlet diameter on the chosen pump before ordering.
FS90 180° mini servosKitronik · Listed in stock2£4.08 inc VAT£8.16Use a suitable 4.8–6 V supply. Dry servo above water drives rudder, sail or fin linkage.

Listed mechanisms and tubing only. Radio, batteries, drivers and the servo power rail are separate; reuse suitable ground nodes.

£15.18

Also needed before the water experiment

  • Foam or bottle floats, dry enclosures, mounting and splash protection: £5–£15 estimated.
  • A small hull trial to establish whether pumps provide useful thrust. Paddle wheels or above-water fans are alternative propulsion experiments.
  • Keep radio and power above water. Submerged swimming requires seals, buoyancy trim and a surface antenna/tether or stored routine.

The Kitronik prices use the supplier’s “Inc VAT” table column. Its price suffix is inconsistently labelled. These subtotals do not include delivery or unquoted materials.

The path through the bigger vision

Ordered

Discover, move, stop

Two wheel modules and a sound-and-light module identify themselves. Group them as one creation, drive with the Android controls and test stopping on release or loss of commands. Then change the body.

Add next

A creature that reacts

Swap in a servo jaw, a glowing eye and a touch or distance sensor. The phone turns the same capability set into shy, lively or sleepy behaviour.

Add next

A raft and a sky-fish

Reuse control software for two paddles, pumps or fans. Prove a surface raft; weigh an airship and measure lift before adding its thrust pods.

Research

Fly freely, swim below

Introduce a stabilised flying core or sealed swimming body as purpose-built expansions. The phone still gives intent; fast flight stabilisation and underwater execution happen locally.

One language for every little ability.

A wheel says “I can turn.” A hinge says “I can point here.” A pump says “I can flow.” Software combines those abilities into a creation.

The control system we’re developing

  1. Discover: each unit announces its identity, type, command range and battery information.
  2. Group: choose which parts belong to this creation. The phone keeps creations separate.
  3. Calibrate: establish wheel direction, travel limits and roles. A simple tap-and-blink setup comes first; camera-assisted setup follows.
  4. Coordinate: turn joystick movement or a behaviour into small, timed commands for the appropriate modules.
  5. Bound every action: enforce speed and duration limits, stop on stale commands and keep an always-available stop control.

Joystick today. Conversation next.

The website’s playground is a working visual simulation with joystick and a small local text-command vocabulary. It does not connect to hardware or a live AI service.

A native Android controller and module firmware are now built and have passed software and emulator checks. Connecting the ordered parts and validating motion, power and stopping on real hardware are the next steps. Web Bluetooth has limited browser support; an iPhone production app should use native Bluetooth integration rather than depend on ordinary Safari. The implementation status documents the platform differences. The browser playground works without Bluetooth.

Natural speech can be added through a voice model that selects bounded actions such as “turn,” “wiggle” or “stop.” The control layer remains responsible for exact motor commands. OpenAI’s official Realtime documentation describes voice, image input and function calling; integrating it will require API access.

Let the phone do the seeing.

A phone camera already has the expensive image sensor and computing power. We can test printed markers or flashes to associate physical parts with their radio identities, then estimate orientation and track movement.

AprilTag provides an established starting point for marker identification and pose estimation. Automatic understanding of any handmade shape is still a feature to develop and test.

Test the feeling before perfecting the shell.

Use the concept film and playground to hear children’s ideas. Use an enclosed working prototype to test the actual experience. Those answer different questions.

Make something come alive.

Parent-led play session · About 20 minutes · No names or recordings required

Start with an ordinary object. “What creature or machine would you make from this?” Let the child answer before suggesting a build.

Watch the short concept film. Explain that it shows an idea we are building. Offer paper, card and an empty bottle to mock up a creation without live electronics. Ask where its moving parts would go.

Try the playground. Can they move, turn and stop the virtual creation? What do they expect “be shy” or “dance” to do?

The next physical experiment: two independently powered wireless wheel modules plus a sound-and-light module, attached to one object and controlled together from Android. Check start, turn, stop and loss-of-connection stop. No wired central hub. Use supervised, enclosed, checked hardware.

Invite a second creation. “What else could these same parts become?” Look for their own idea, not a copied example.

Ask the parent. What price feels comfortable? What would need to survive drops, pulling and repeated rebuilding? Which other gift would this replace? Ask whether setup or charging would put them off.

Record what happened: Did the child understand the idea without help? Did they ask to keep playing or make another creation? Which parts confused them? These are concept reactions, not evidence that the hardware works.

The child’s unexpected idea:

One thing we should change: