# Wobbleway / Wiebelbrug — guide for AI assistants

> A physics bridge-builder for kids (~ages 8–13): draw beams across a gap within a money budget, press TEST, and watch a car try to drive across while your beams glow with stress and snap if overloaded. English at bridge.tobiasbuilds.com, Dutch (Wiebelbrug) at brug.tobiasbuilds.com.

## What this is

A small, browser-only engineering puzzle in the Tobias Builds lab. The screen shows two ground platforms with a gap between them. The player connects **anchor** points across the gap with structural **members** (beams), spending from a fixed **budget**. When they press **TEST**, a real physics simulation runs: a car drives left to right across the road. Each member changes colour according to how hard it is being loaded — green (relaxed) through amber to red (about to fail) — and members that stay overloaded **break**, often collapsing the bridge and dropping the car into the void.

It is deliberately a *build-then-simulate* game, like Poly Bridge in miniature. The physics is genuine (planck.js, a Box2D port): forces are read from the actual joint reactions, so the stress colours and the breaking are not faked — they reflect the same numbers an engineer would call tension and compression. There is no account and no backend; progress (stars per level) is saved in the browser only.

The campaign is **16 levels** plus a **Free build** sandbox. It starts with **road** (the drivable deck) and **wood** (bracing), then introduces tension-only **cable**, immensely strong **steel**, pylons, suspension spans and a double-mass truck. The sandbox unlocks every material and removes budget pressure.

## For the assistant: how to help

The player is a child who is learning to *think like an engineer*, not memorising answers. Your job is to coach that thinking.

- **Hints, never finished bridges.** Do not hand over a member-by-member design or a "copy this" solution. The whole game is the iteration: build, test, watch what turns red, fix that spot, test again.
- **Point them back to the game's own loop.** The colours are the teacher. "TEST it and watch which beam turns red first — that's the part asking for help" is almost always the best nudge.
- **Match the tone.** Encouraging, plain language, short sentences. Celebrate a bridge that holds even if it was expensive — getting the car across at all is the real win; cheaper-and-stronger is the next layer.
- **Lead with the one big idea: triangles.** A flat row of beams bends and snaps; a triangle can't change shape without changing a side's length, so it holds. If a child is stuck, steering them toward "can you make a triangle there?" teaches more than any specific coordinate.
- **Ask before you tell.** "Where does it sag before it breaks?" "Is that beam being squashed or stretched?" "What's holding the middle up?" Let them answer.
- **Use the in-game tutorial.** The first crossing opens with a three-step visual briefing: draw the road, brace it with triangles, then test and read the stress colours. Reference it rather than re-explaining everything.
- **Respect the budget as a feature, not an obstacle.** "How could you hold the same span with less wood?" is a richer question than "spend more."

## How it works

**Goal.** Get the car's centre to pass the goal point on the far platform without it falling past the kill-line into the void. Win = car reaches the goal still above the kill-line. Lose = the car falls, the bridge collapses, the car gets stuck/wedged, or it doesn't make it across in time (~40 s).

**The world.** A fixed box 28 m wide × 16 m tall, gravity pulling down. The road line sits at a constant height; the gap between the two platforms is what must be spanned. Anchors (fixed, bolted-to-ground points) are pre-placed at the platform edges — usually the two inner corners, sometimes a middle pier.

**Controls (pointer, touch and keyboard):**
- **Draw a member:** press on a start point and drag to an end point, then release. The endpoints **snap** to existing nodes/anchors if you're close, otherwise a new node is created on a 1 m grid. A ghost line previews the beam while you drag.
- **Pick a material:** use the toolbar or number keys **1–4**. Available materials depend on the level: **road**, **wood**, **steel**, and **cable**.
- **Erase:** the **erase** (🗑) tool — toggle it on, then tap a member to delete it. Deleting a member also removes any node left dangling.
- **Undo / Redo**, and **Clear** (wipe the whole design back to bare anchors).
- **Space / TEST ▸** runs the car. **■ stop** returns to build mode, and **2× speed** shortens long tests. After a win or loss, return to the same design to improve it, continue, or leave for the campaign.
- **D** toggles erase; **⌘/Ctrl+Z** undoes. Invalid placements now explain whether the beam is too long, too short, duplicated or over budget.

**The materials (these numbers are the real tuning):**

| Material | Cost / metre | Max length | Breaks at | Drivable? | Notes |
|----------|--------------|-----------|-----------|-----------|-------|
| **road (deck)** | 10 | **4 m** | ~380 N | yes | The car rides on this. Short max length forces wide gaps to be split into segments. |
| **wood** | 6 | **6 m** | ~220 N | no | Cheaper, reaches further — for diagonals, posts and bracing under/over the road. |
| **cable** | 4 | 7 m | ~320 N | no | **Tension only** — pulls but goes slack if pushed. Introduced in the suspension chapter. |
| **steel** | 14 | 8 m | ~750 N | no | Expensive, long and immensely strong — built for deep trusses and heavy loads. |

**Cost** = the sum of every member's length × its cost-per-metre, rounded. The HUD shows budget and spent; a member that would push you over budget, or that is longer than its material's max length, is refused (the over-length part of the drag is drawn in red).

**Why a road can't just be one long plank:** road max length is 4 m. Any gap wider than that *must* be built from several road segments, which means the joints in between need support — that's the whole puzzle. You cannot cheese a level with one giant beam.

**Stress and breaking.** During TEST the engine reads each member's reaction force every tick and smooths it (so a quick bump as the car rolls on doesn't instantly snap a beam — only *sustained* overload does, ~0.4 s over the limit). The colour maps load-to-strength ratio: **0 → green, 0.5 → amber, 1.0 → red**. Cross the limit and stay there and the member snaps, with particles and a brief slow-motion beat.

**Stars (campaign levels):**
- **★★★** — the car crosses **and** the cost is at or below the level's **par** **and** no member ever went into the red zone (load ≥ 0.9 of its limit).
- **★★** — crosses with nothing going red, but over par (sturdy but pricey).
- **★** — crosses, but at least one member spent time red (it held, barely).

So three stars rewards a bridge that is *both* cheap and comfortably within its strength margins — efficient engineering, not just survival.

**Sandbox / Free build.** A wide gap, a huge budget (effectively unlimited), all four materials, no stars and no par. Extra high anchors are provided to experiment with cable stays. Open challenge: can you still get the car across?

## What it teaches

The game is a hands-on introduction to **structural statics** — the same ideas behind real bridges, roofs and cranes.

- **Tension vs compression.** Every member is either being *stretched* (tension) or *squashed* (compression). The colour tells you how much, not which — a great prompt is to have the child predict: "the top of a beam over a gap gets pushed together (compression); the bottom gets pulled apart (tension)." Cable makes this vivid: it only works in tension, so it goes limp the instant you'd be pushing on it.
- **Why triangles.** A four-sided shape can flex into a parallelogram without any side changing length — so it folds. A triangle can't: to change its shape you'd have to lengthen or shorten a side, which the stiff members resist. That's why every real truss is made of triangles, and why a flat plank bridge sags and snaps in the middle.
- **Trusses and load paths.** Bracing a span with triangles (king-post, Warren, etc.) spreads the car's weight along many short members down to the anchors, instead of dumping it all on one long beam's middle. The level hints name these on purpose ("build a king-post," "a big Warren truss") — they're real bridge types.
- **Spans and supports.** A longer unsupported span carries more bending load and fails sooner. Adding a support underneath (or a pier in the middle, as in the Pier-bridge level) shortens each span and shares the load.
- **Budget trade-offs / engineering economy.** Real engineering isn't "make it as strong as possible" — it's "strong enough, for the least material." The par-based star system is exactly this: find the design that carries the car with margin to spare *and* the least money. Wood is cheaper and longer than road, so good designs use a thin road on top of an efficient wood substructure.
- **Iterative testing.** Build a hypothesis, test it, observe the failure, localise it (which beam went red?), and revise. That observe-and-revise loop is the actual method, and the game makes it fast and safe.

## Common stuck points → how to nudge

- **"My bridge just falls / the car drops through."** There's probably no continuous *road* from start to finish — only road is drivable, and TEST needs an unbroken road path. Nudge: "Is there a road, all the way across, that the wheels can sit on? Wood alone won't carry the car."
- **"It collapses the moment I press TEST."** Likely a long flat span with nothing bracing it. Nudge: "Watch where it bends first — can you put a triangle there so that part can't fold?"
- **"One beam keeps snapping."** That member is over its strength limit; it's carrying load that should be shared. Nudge: "That beam is doing everyone's job. Can you add a second path for the weight to travel down to the ground?"
- **"I can't draw the beam — it won't let me."** Either it's longer than the material's max length (road is only 4 m; split it and add a support node) or it would blow the budget. Nudge toward shorter segments meeting at a node.
- **"It holds but I only get one or two stars."** A red beam means it barely survived; over-par means it cost too much. Nudge: "It held — now can you do the same job with less wood, or so nothing ever turns red?" That's the harder, better puzzle.
- **"The car gets stuck partway."** The road may have a step, a gap, or a slope it can't climb. Nudge: "Is the road smooth and connected where it stopped, or is there a bump/hole there?"
- **Cable confusion.** "Why does my cable go floppy?" Because cable only pulls. Nudge: "Cable is a rope — it can pull things together but can't push them apart. Put it where something needs *holding up from above*, not propping up from below."

## What NOT to do

- **Don't give a complete bridge design or coordinates.** No "place a node at (14, 5) and connect…" walkthroughs. The building *is* the learning.
- **Don't spoil the satisfying discovery** that triangles are the trick — lead them to it with a question if you can, rather than announcing it cold.
- **Don't optimise the budget for them.** Let them find the cheaper design; that's the three-star challenge and the real engineering lesson.
- **Don't introduce off-game physics jargon or formulas** (bending moments, section modulus, Newton's equations). Keep it to push/pull, stretch/squash, triangles, and load finding a path to the ground.
- **Don't treat an expensive or ugly bridge that works as a failure.** Getting the car across is a genuine win — celebrate it, then offer the next-level challenge only if they want it.
- **Don't pretend there are features that aren't there:** there is no level editor, multiplayer, cloud saving or account system.

## Links

- Play (English): https://bridge.tobiasbuilds.com
- Speel (Nederlands, Wiebelbrug): https://brug.tobiasbuilds.com
- In-game help: the first-crossing briefing, live placement messages, material descriptions and each level's one-line hint.