World Design: Space Is a Decision

A world reads as flat and arbitrary when its arrangement was never decided — props scattered uniformly, geometry placed where it was convenient to code, a play space that looks the same in every direction. Spatial design is deciding what goes where and why, from the player's point of view. Make those decisions at design time (in the PRD, before code): layout shapes the systems that build it, and it cannot be retrofitted during polish.

This skill defines what to think about; how you achieve it is yours to choose.

Compose for the gameplay camera

The only view that matters is the one the player actually has. A layout that looks great from an editor god-view can be unreadable from a chase camera two meters off the ground.

• What fills the frame in the first three seconds? That shot is the game's first impression — design it like one.
• At the actual camera height/FOV, is there depth layering — foreground detail, midground play space, background silhouette and sky — or one undifferentiated plane?
• Where does the horizon sit, and is there something interesting on it in every direction the player will commonly face?

Orientation: the player always knows where they are

• Landmarks: a few large, distinct, widely visible features (each unique — a repeated landmark misleads). From any reachable point, can the player name their location and point toward the objective?
• Directional asymmetry: the world should look different facing north vs. south — asymmetric skylines, a light source with a definite direction, distinct edge treatments. Uniform surroundings make every wrong turn feel identical.
• Local wayfinding: lighting, color, and geometry can lead the eye along the intended path — players follow light, contrast, and converging lines without being told.

Scale, proportion, and density rhythm

• Give the world a scale reference the player's own avatar/vehicle establishes — then keep proportions consistent with it. One out-of-scale prop breaks the whole illusion.
• Vary density deliberately: busy zones (detail, landmarks, action) alternating with quiet ones (negative space, long sightlines). Uniform density — whether empty or cluttered — reads as noise. The quiet stretches are what make the busy ones land.
• Cluster props the way the world's fiction would (things accumulate near walls, paths, and structures), not uniformly at random.

Space serves the mechanics

The layout is a gameplay system, not scenery. Ask what the space must do for the game and let that drive placement:

• Can the player read the next challenge before it arrives — corners, threats, goals, or decision points telegraphed by geometry and landmarks?
• Does the space shape routes and choices (cover, chokepoints, paths of least resistance), or is every location mechanically identical?
• Is the problem space visible (or discoverable) from the camera the player actually controls? Does the framing itself hint at what to do next?

Approaches worth exploring

Suggestions, not a curriculum: techniques such as greybox blocking before the art pass, Poisson-disc or noise-driven scattering, spline-based paths and networks, space partitioning and room-graph generation, wave-function collapse, L-systems, hero landmarks over procedural fill, rule-of-thirds framing and leading lines. Pick what serves the concept, combine freely, and reach past this list when you know something stronger.

The walk test

Before calling the world done, playtest it as a place, not a program:

• Screenshot the actual gameplay camera at 4–5 key moments — does each frame look composed, with layered depth and a clear focal point?
• Spin the camera 360° from the center and from the edges: is any direction dead, repeated, or unreadable?
• Could a player who just spawned point toward the goal? Could they retrace their path?
• Does the layout visibly serve the core mechanic (challenges readable, routes shaped, goals framed) rather than merely hosting it?