Early concept
Workbench canvas + copilot
A rough pass on the graph canvas, module tray, and AI copilot before the interface became a polished editor.
Case Study
Levelyst.AI
Levelyst.AI is a concept for turning a game prompt into a playable prototype direction, using reviewable blueprints, reusable smart modules, and a simulation-ready workbench before a team spends time building from scratch.
RoleSolo Product DesignerYear2025TypeProduct Design / UI-UX Design
UX ResearchInteraction DesignAI Product ThinkingPrototyping
Overview
What Levelyst.AI needed to prove
Problem
Game ideas are easy to describe, but teams need something playable before feedback becomes specific and useful.
Audience
Solo builders, small game teams, and design collaborators who need a shared prototype direction before production work begins.
Role
Product strategy, workflow design, interface design, prototyping, and framing the narrative around how AI generation becomes a usable prototype flow.
Scope
Define the prompt-to-blueprint system, shape the smart-module workbench, and prove that generated output can stay editable through build, simulate, and debug.
Outcome
A concept that turns a rough brief into a reviewable plan, a modular graph, and a playable prototype path the team can critique immediately.
01Challenge
The real gap was not inspiration. It was getting an idea into a playable state fast enough to discuss.
Friction
Early game concepts usually break across docs, whiteboards, and half-built prototypes before anyone can react to the same thing.
Design Goal
Move from prompt to something testable quickly, while keeping the output loose enough for iteration instead of pretending it is final production content.
02Target Users
The target users all share one need: something concrete to react to before the project gets expensive.
Target User 1
Solo builders
Need a playable direction quickly so the first prototype is guided by a real structure instead of a blank scene.
Fast first pass
Target User 2
Small game teams
Need alignment before production so mechanics, tone, and scope are discussed against the same prototype direction.
Shared reference
Target User 3
Design collaborators
Need something concrete to critique so feedback can target flow, feel, and system choices instead of vague intent.
Better critique
03Early Sketches
Before the polished workbench existed, the concept was being tested through rough screens, paper flows, and module-first layouts.
These sketches helped lock the relationship between prompt input, reusable modules, and a workspace that could move into simulation without feeling like a black box.
Early concept
Dashboard + community flow
Explored how projects, templates, and community inspiration could live alongside quick project starts.
Early concept
Construct-first graph view
Tested a more graph-first construct mode where systems and flow beats could be shaped visually.
Early concept
Module-led builder
Focused on how modules, AI suggestions, and simulation controls could stay readable in one builder screen.
04Prompt to Prototype
One pass shows the full workflow from a blank project to a playable graybox.
The flow starts with a new project, turns the prompt into an AI Blueprint, generates the module graph, and drops straight into simulate so the concept can be tested instead of just admired.
Swipe to explore
05Game Families
The system can steer toward several families, but two of them are the clearest and most testable today.
Levelyst already supports a broader prompt surface, but the strongest runtime proof is still the 2D platformer and 3D FPS survival paths. The other families show where the product can steer next without pretending they are equally polished now.
Most refined nowPlayable path
2D Platformer
The most proven 2D path, already grounded in movement, camera framing, pickups, and checkpoint-ready prototype structure.
Most refined nowPlayable path
3D FPS Survival
The clearest 3D path today, built around first-person movement, hitscan combat, enemy pressure, and wave-based runtime testing.
Broader directionPrompt direction
Tactical Shooter
Aiming and encounter prompts can already steer in this direction, but the deeper team-tactics layer still needs a more dedicated runtime pass.
Broader directionPrompt direction
Top-Down RPG Adventure
Quest-like exploration and system-driven adventure prompts fit the planning model, even though the runtime path is not as refined yet.
Broader directionPrompt direction
Sandbox Builder
Open-ended building prompts can be translated toward a supported slice, but the construction loop is still more directional than production-ready.
Broader directionPrompt direction
Open-World Action
Broad GTA-like action prompts can be framed into a playable direction, though the family is still more conceptual than fully runtime-proven.
06Playable Output
The generated result is meant to be entered, tested, and compared.
2D platformer simulation
The 2D path lands in a rooftop platformer with movement, checkpoints, and pickups already wired into something you can immediately test.
3D FPS simulation
The same system can shift into a survival-FPS direction and still stay grounded in runtime behavior instead of a fake cinematic.
07Smart Module Library
Smart modules keep the system reusable instead of regenerating the whole prototype from scratch.
The module layer gives Levelyst.AI a stronger promise than image generation alone. Playable direction is assembled from reusable behaviors the team can keep shaping after the first pass.
PlayerDepends on Gravity Physics
Platformer Controller
Defines jump, run, and landing behavior for a deterministic side-scrolling character.
Physics
Gravity Physics
Provides the movement rules that make jumping, falling, and grounded timing feel believable.
CameraDepends on Platformer Controller
Side-Scroll Camera
Keeps the player readable in motion and frames level progression without extra setup.
PlayerDepends on Character Body
FPS Controller
Sets up first-person movement and look controls for shooter-style prototype families.
CombatDepends on FPS Controller
Hitscan Weapon
Adds immediate weapon feedback so shooter concepts can be tested as interaction, not just described.
AIDepends on Character Body
Basic Zombie
Drops in a simple enemy loop so encounter pacing can be validated inside the generated prototype.
Physics
Character Body
Acts as the runtime body foundation for movement, collision, and enemy behavior in 3D spaces.
SystemsDepends on Platformer Controller
Checkpoint System
Gives platformer runs a progression loop that can be evaluated without a larger save framework.
SystemsDepends on Platformer Controller
Coin Collectible
Adds pickup logic and reward beats so the prototype has something to optimize around.
SystemsDepends on Basic Zombie
Wave Manager
Coordinates encounter pacing for survival-style runs where enemies arrive in timed pressure loops.
08Workbench View
The module library stays grounded in an actual editor, not just an abstract system diagram.
Planning
Prompt, plan chips, and recommendations stay visible while the graph is being shaped into a testable prototype.
Dependency Trace
Inputs, outputs, and linked systems make it easier to see what has to connect before simulate is worth trusting.
The live workbench keeps the graph, module details, and planning context in one place so generation still feels editable after the first draft.
Swipe to explore
09Update Review
The generated prototype can still be reviewed and updated without losing the thread.
Review Gate
Updates can be inspected before the next generation step lands, which keeps the prototype from feeling like a black box.
Team Use
Explicit change review makes the output easier to share with collaborators who need to understand what the next pass is actually doing.
Instead of hiding the next pass behind a single regenerate action, Levelyst.AI can surface what changed before the team commits to it.
Swipe to explore
10Outcome
The strongest proof point is no longer the concept art. It is the playable direction.
The updated concept now reads as a product that helps a team move from prompt to blueprint to simulation, while keeping the generated output modular enough to refine instead of replace.
Runtime Paths
2D + 3D prototype directions
System Basis
Reusable smart modules
Core Promise
Prompt to playable prototype
Takeaways
- The review step is what makes the AI output feel believable instead of magical.
- Smart modules give the concept a stronger long-term product story than one-off generation alone.
View Prototype
Try it yourself
Swipe to explore