Test Quest: Easing D&D onboarding through a generative content pipeline, achieving 19x cost reduction and 21% confidence improvements.

CONTEXT

GRADUATION PROJECT (2024)

ROLE

PRODUCT DESIGNER

FUNCTION

USER RESEARCH, RAPID PROTOTYPING, USABILITY TESTING, THEMATIC ANALYSIS

STACK

FIGMA, PYTHON, OPENAI API

MEMORY ALLOCATION : OPTIMAL
ARCHIVE NODE : 0x4A7B

> ABOUT

Test Quest is a software tool that produces custom Dungeons & Dragons adventures to ease the onboarding process for new player groups. Over six months as a solo graduation project, I combined interaction design and a passion for tabletop games to help novices jump into the game faster and more confidently.

CHECKSUM : VERIFIED
KEY EXCHANGE : SUCCESS
LOADING FUNC #273 ##==
FRAGMENTATION : 2.1%
NPU1_OUTPUT : PROCEED
NPU2_OUTPUT : PROCEED
NPU3_OUTPUT : PROCEED
3S  UV 4D  XX 1C
3S  K  4D  Y  1B
3S  UV 4D  XX 1C
3S  K  4D  Y  1B

> BACKGROUND

D&D serves as more than entertainment — it's a collaborative sandbox where groups build social bonds through shared experience. Yet newcomers face significant barriers: complex rulebooks, expensive starter kits, and time-intensive preparation. Therefore, my objective was to reduce these onboarding frictions.

SYNTAX PARSING : COMPLETE
OUTPUT TARGET : x86_64

LITERATURE REVIEW

13 PAPERS ANALYZED

PLAYER INTERVIEWS

5 PARTICIPANTS

COMMUNITY SURVEY

250 RESPONSES

> PROBLEM SPACE

I deployed a mixed-methods research approach to identify the most relevant pain points within the D&D community:

Rule complexity impeding game flow

Low role-playing confidence preventing character immersion and detracting from the social experience

Character creation processes exceeding user patience thresholds

LOAD BALANCING : OPTIMAL

CCR-CODER COMPILING    HANDSHAKE ESTABLISHED: X01    ID: 7428-2339-0841-5069    

CCR-CODER COMPILING    HANDSHAKE ESTABLISHED: X03    ID: 2325-9994-0831-5069    

> DESIGN APPROACH

I applied the Double Diamond framework from brainstorming through testing. User insights and iterative prototyping helped me establish three core design principles:

Players must never feel overwhelmed or distracted during the game creation process.

Game materials must be fit to match user experience levels.

Content must maintain consistent story logic across all game elements, to avoid confusion.

3S  UV 4D  XX 1C
3S  K  4D  Y  1B
3S  UV 4D  XX 1C
3S  K  4D  Y  1B
MODEL LOADED : GPT-4
TEMPERATURE : 0.7
TOP_P : 0.9
MAX TOKENS : 2048
GENERATING ...

> CONTENT GENERATION PIPELINE

I iteratively refined the content generation architecture: users input minimal character and game preferences; the system then defines custom characters, with advice tailored to their chosen focus. This approach presented several benefits:

Automated generation reduced preparation time

Character-specific rule references were accessible on-demand, preventing mid-session interruptions

Customization options retained player investment and creative expression while maintaining a consistent game setting

PORT 8847K : SECURE
AUTH PASSKEY>>> [REDACTED]
SHARD TRANSFER INITIALIZING
PACKET INTEGRITY : 100%
G62
W58

> IMPLEMENTATION

Test Quest executes an LLM-based pipeline via a command-line interface. The system generates characters as JSON files in parallel, then integrates backstories into a unified knowledge base containing all campaign information.

The world lore guides sequential generation of three narrative acts, each containing unique characters and encounters calibrated to campaign focus. After generating each act, the system compacts it into a contextual overview for subsequent acts, maintaining narrative coherence and preventing information conflicts.

The generated materials are formatted to print-ready PDFs via the Homebrewery web app, outputting complete D&D adventures in 20 minutes at an average cost of €1.5 per set.

SHELL DIAGNOSTIC INITIALIZING [::
SHARD_00 [PRIMARY] ... OK
SHARD_01 [REPLICA] ... OK
SHARD_02 [REPLICA] ... OK
::]
3 892 561
7 443 205
20/25
REV 2

> RESULTS

12 hours of structured tests were organised with 15 novice participants to validate system effectiveness. I used pre/post-test surveys, focus group discussions, quantitative observation and thematic analysis to assess overall performance:

PLAYER CONFIDENCE

+21%

PREPARATION TIME

-80%

COST DECREASE

19x

NARRATIVE COHERENCE

73%

LEVEL : WARN [23 ENTRIES]
MESH TOPOLOGY : UNSTABLE
“It cut down the preparation time a lot, which was my main concern with trying out a new game I'd never played before”

KEITH

TEST PARTICIPANT #07

EDGE SERVER : AMS-02
CONTENT ID : 2164-37XX
PING : 19MS
EDGE SERVER : AMS-02
CONTENT ID : 2164-36XX
PING : 15MS
“I wouldn't have made such a character, because there is something restrictive about sharing your own ideas [...] I think for a lot of players it would help with role-playing.”

EVELYN

TEST PARTICIPANT #12

3S  UV 4D  XX 1C
3S  K  4D  Y  1B
3S  UV 4D  XX 1C
3S  K  4D  Y  1B
EDGE SERVER : AMS-02
CONTENT ID : 2164-38XX
PING : 24MS
MEMORY LEAK : DETECTED
FRAGMENTATION : 80.7%
CLUSTER STATUS : DETERIORATING
FLUSHING BUFFERS : COMPLETE
EXIT CODE : 0

> REFLECTION

I found context architecture to be the primary quality determinant: comprehensive information yielded superior character integration and backstory coherence

While Test Quest demonstrates how LLM-based content pipelines can reduce onboarding frictions, I found generated materials ultimately lack the imaginative depth of human-made content. I recommend gradually transitioning to own materials as experience grows to encourage creative exercise.

WARNING>>> BUFFER OVERFLOW
DD
5C  E9  DD
AF  BD  5C  E9  DD
DD
5C  E9  DD
AF  BD  5C  E9  DD
DD
5C  E9  DD
AF  BD  5C  E9  DD

> CONTACT

UPLINK ESTABLISHED

100%

IDENTIFIER VERIFIED

100%

CONNECTION SECURED

100%

Open communication channel?

CONNECT

ZZT
BBX