Cogwalk

How to play

How to Play Cogwalk

Read every robot's starting direction, build one shared sequence, run it, and revise only the instruction that caused the first wrong move.

1. 01

   Identify each robot, its facing direction, and its matching goal. Trace a possible route for every robot before adding commands.

2. 02

   Add a forward instruction with Space, W, Up, or the on-screen step button. Add left and right turns with A/Left and D/Right or their buttons.

3. 03

   Watch the instruction counter. The strip has a level-specific limit, so remove moves that merely undo earlier turns or duplicate an avoidable detour.

4. 04

   Press Enter or GO to run the complete strip. Follow all robots through every command rather than watching only the one closest to a goal.

5. 05

   If the attempt fails, note the first command where a robot leaves its useful route. Use Backspace or tap an instruction to remove it, or press R/Clear when the plan needs a new structure.

6. 06

   Use walls as controlled delays. A blocked forward command costs that robot a step while allowing another robot to move, which can synchronize different route lengths.

About the game

Write one short program for robots that face different directions

Cogwalk is a compact programming puzzle built around a shared command strip. You place step-forward, turn-left, and turn-right instructions, then press GO. Every robot follows the complete strip at the same time. Because the robots can start in different squares and face different directions, a command that helps one robot can send another toward the wrong side of the board.

The board turns walls into useful timing tools. A robot that tries to step into a wall remains in its current square for that instruction while the others continue. That deliberate stall lets the group fall out of sync without adding separate programs. Later levels introduce ice, turntables, and a third robot, but the central question stays the same: what single sequence makes every starting position reach the correct goal?

The first level demonstrates the full loop without hiding the result. Adding two forward instructions and pressing GO moves the robot to its target and advances the counter from level 1 to level 2. From there, each board adds a new reason to inspect orientation, obstacles, goal colors, and the length limit before running a program.

Controls

Controls for Cogwalk

Space / W / Up

Add a step-forward instruction.

A / Left

Add a turn-left instruction.

D / Right

Add a turn-right instruction.

Enter

Run the entire command strip.

Backspace

Undo the last instruction.

R

Clear the current program.

Touch

Tap command buttons, GO, or an instruction to remove it.

Useful strategy

Treat walls as timing devices and debug the first divergence

Read every move from each robot's point of view

A forward command does not mean up on the screen. It means one square in the direction that robot currently faces. Before building the strip, mark each robot's orientation mentally and trace its first two possible steps. This prevents a common mistake where the route looks correct for one robot while another immediately walks away from its goal.

Turns are relative too. A left turn changes each robot by ninety degrees from its own orientation, so their new screen directions may still differ. When you add a turn, say where every robot will face afterward. If one robot needs the turn and another does not, look for a wall or board feature that can absorb a later forward step without ruining the second route.

Goals should be read as destinations for the matching robot, not interchangeable endpoints. Compare color and position, then work backward one or two squares. The square before a goal often reveals the required final orientation and makes the last commands easier to determine.

Use blocked steps to create different timing from one program

Walls do not end a run. When a robot receives a forward command into a wall, it stays in place for that step while the rest of the crew continues. This is the puzzle's main synchronization tool. Instead of trying to give one robot fewer instructions, position it so one shared instruction has no movement effect for that robot.

Look for routes where a robot can safely face a boundary while another still needs to advance. Add the forward command, let the blocked robot wait, and then use the next turn to bring both plans back together. A useful stall has a clear exit; repeatedly pushing into a wall without changing the other route usually wastes strip space.

A delay can happen early or near the goal. Early stalls help equalize routes of different lengths, while a late stall can hold a robot safely on or beside its destination as the others finish. Test whether the goal remains valid when later commands run, because reaching it early may not be enough if subsequent instructions move the robot away.

Debug one command at a time instead of rebuilding blindly

Press GO with a specific prediction: after command one, robot A should be here and robot B should be there. Watch for the first point where reality differs. That command is the best place to revise. Changes made later in the strip cannot repair a robot that already entered the wrong corridor unless the board deliberately routes it back.

Use Backspace or tap an instruction when the structure is mostly sound. Clear the full strip only when the route assumption is wrong, such as choosing the wrong side of a wall or requiring more commands than the level allows. Preserving the verified prefix saves time and keeps successful reasoning visible.

After three attempts, the official game can reveal a hint that adds a correct next instruction from the current position. Treat that hint as a lesson about the verified prefix, then predict what follows. It is more useful to understand why the command fits every robot than to copy it without rechecking orientation.

Separate your command from ice and turntable movement

Later boards add mechanics that change what happens after a command. On ice, one forward action may carry a robot farther than a normal floor square. A turntable can change orientation without consuming another turn command. Observe the board response once, then update your route model before adding more instructions.

Write the sequence as alternating causes and results: programmed step, automatic slide, new square; programmed turn, turntable response, new facing. This prevents double-counting movement or adding an unnecessary correction. The same shared strip still governs every robot, but different terrain can produce different outcomes from one instruction.

When a third robot appears, avoid solving pairs independently and hoping the last one fits. Check all three after every planned command. A wall stall or automatic rotation that solves timing for two robots may create the exact delay the third needs, making board features more valuable as the crew grows.