Fit — the routing drill

Drag the base handle

You do not aim the tip. You drag the base plate; a limited-joint arm carries the motion out to the tip. Its joints are constrained — locked twist, clamped bend — and a keep-out obstacle sits between you and the target. There is no straight-line reach: you must route the chain around the obstacle to fit the tip through. The target is hidden — the only guide is the warmer/colder strip, which reads the tip's position. Match it, then hold to lock. Left-drag the plate (forgiving grab) · right-drag or hold Space to orbit · wheel zooms.

Tip proximity

— pick a target —

left-drag the pink plate (forgiving grab) · right-drag / Space+drag to orbit · wheel zoom · the target is hidden — read the strip

press New target to begin

Not aligned
hold inside the tolerance to lock

0LOCKS

—BEST ERROR

—LAST LOCK

Press New target. A hidden position appears beyond the obstacle. Drag the base plate until the strip reads "warm", then "locked".

Why did I miss?

Error is split so a miss is legible — is the tip simply too far from the target (reach), or is the chain blocked by the obstacle so it can't route through (clearance)?

Assists & options

Reveal target (training wheels)

Show tip marker Show linkage

Show tolerance sphere (on lock approach)

Keyboard controls

Click the 3D view first (it takes focus). ←→↑↓ orbit · Home reset view · wheel zoom.
Drag the pink base plate to drive the chain. The tip is the referent of all feedback — always.

What this drill teaches

One sub-skill only: fit — routing/reach through available degrees of freedom. The chain has four segments; joints are angle-clamped and one twist axis is locked, so a naive straight reach toward the target fails — it runs the chain straight into the grey obstacle. You have to find the joint configuration that threads around it. Obstacles are visible (they are the constraint, not the hidden part); only the target position is hidden. Indirect control: you drive the base plate, a constrained CCD solver carries it to the tip, so you learn the chain's routing behavior — and because feedback tracks the tip against a hidden target, you learn the target geometry, not an input-space macro. Target is randomized per attempt, static within it. Max two uncertainty sources: hidden target + indirect (constrained) linkage. No physics noise, no ragdoll — the solver is fully deterministic.