AI collaboration case study · Active private beta

Independent product · 2026

The Chosen Quest

Rebuilding a classroom RPG into a cohesive desktop experience. A human-directed, AI-assisted redesign spanning gameplay systems, interface design, visual quality control, testing, and release preparation.

Role
Product direction, UI/UX design, workflow direction, playtesting & QA
Product
Desktop role-playing game
Platform
Java desktop
Status
Active private beta
The Chosen Quest title over a dark fantasy mountain valley with new quest and load quest controls
The Enhanced Edition establishes a painterly fantasy and restrained Art Deco direction.

01 / Overview

A visual refresh became a connected product effort.

The Chosen Quest began as a classroom console RPG. The Enhanced Edition asks what it takes to turn that foundation into a legible, atmospheric, and dependable desktop product without erasing the character of the original.

Character choices needed to persist into exploration and combat. New artwork needed responsive framing. Gameplay changes needed repeatable validation. The interface, underlying rules, and review process had to evolve together.

  1. 01

    Make the game easier to understand and operate.

  2. 02

    Give the experience a recognizable visual identity.

  3. 03

    Improve functionality without losing testable behavior.

02 / The collaboration model

AI increased execution capacity. Judgment stayed human.

Plausible output was never treated as finished output. Every change moved through a repeatable direction, implementation, and review loop.

AI assisted with codebase analysis, option generation, implementation, test coverage, documentation, and selected visual production. I retained responsibility for the product intent: defining constraints, choosing tradeoffs, reviewing results, and deciding what was ready to keep.

  1. 01Audit
  2. 02Define constraints
  3. 03Explore options
  4. 04Choose a direction
  5. 05Implement
  6. 06Render & test
  7. 07Review
  8. 08Refine

03 / Character system

One system, not sixteen unrelated screens.

A neutral charcoal, parchment, and antique-gold shell holds stable geometry while class color and race heraldry behave as modular identity layers.

The interface supports four races and four classes without sacrificing interaction states, contrast, or readability. Class identity leads; race ornament stays quieter; functional colors keep their established meanings.

OutcomeCharacter identity now follows the player into the hero rail, inventory, combat feedback, and outcome screens.
Character creation interface showing race and class choices, Combat Paths, portrait, meters, statistics, and equipment
Race, class, Combat Path, portrait, and equipment information share one responsive system.
4Playable races
4Playable classes
16Supported race/class combinations
1Connected identity system

04 / Gameplay systems

Turning class choice into a play style.

Combat Paths give each class two readable starting approaches while equipment can later unlock a different current style.

The Rogue exploration established the model: compare equipment, action sequence, strength, and tradeoff instead of relying on flavor text alone. The player's origin remains visible even as their equipment changes.

OutcomePath and style information persists through character creation, exploration, inventory, combat, saved games, and player-facing status.
Art Deco concept comparing Rogue Assassin and Skirmisher combat paths through equipment, action loops, strengths, and tradeoffs
The Rogue concept compares two starting identities through concrete play consequences.

05 / Quality process

Making polish testable.

A change was not complete because one screenshot looked right. It had to survive representative content, interaction states, target compositions, and gameplay behavior.

Render checks cover representative screens and control states; deterministic simulations exercise combat builds; structured playtesting turns observations into durable backlog work. This separates visual confidence from behavioral confidence while requiring both.

Rendered matrix of fantasy interface button states used for visual quality checks
Interaction states were treated as a system and reviewed through visual regression output.

Reflection

Intent, implementation, evidence, revision.

The most important artifact is not a single generated screen or feature. It is the process that connects those four things.

AI accelerated analysis and implementation, but it did not decide what made the product coherent. Strong constraints, durable documentation, automated checks, screenshots, and playtesting each caught a different class of problem.

Planned next

Playable browser comparison

Separate Original and Enhanced editions are being prepared for a guided comparison. They will remain unavailable until loading, input, save/reset, and responsive behavior are validated.

Browser editions in development