MULTI-AGENT AI FOR GAME DEVELOPMENT

Explore the structure of the course and how each assignment builds toward your final project. Get familiar with your instructor’s background, the tools you’ll be using, and how to navigate the technical and creative expectations of the program.

• Meet your instructor
• Course structure
• Assignments & final project overview
• Environment setup walkthrough
• Breakout room: Discuss your reasons for doing the course
• Fun Exercise: What would your AI build?

Explore the shift from traditional scripted systems to autonomous, LLM-powered agents. Understand how different AI architectures shape behavior, and how to distinguish between agents used as development tools versus those embedded directly into gameplay.

• From scripts to agents
• Agents in dev vs. in-game use
• Establishing the AI contract
• How major companies use AI today
• Agent architectures for dev pipelines
• Case Study #1: Ubisoft & Netflix
• Fun Exercise: Agent or not?

Discover how to translate a game idea into a clear, technically aware Game Design Document. Explore how AI systems influence design decisions, from token budgets to agent roles, and how to scope effectively for both creativity and feasibility.

• Game idea scoping: “One agent, one wow”
• The three new pillars of AI GDD
• The dev crew
• GDD templates
• Demo #1: Using an Agent team to "stress test" a GDD
• Case Study #2: Project Baker and Joshua’s GDDs

Assignment #1: GDD First Draft
Using the provided template, submit your first Game Design Document.

Dive into how to validate your GDD by using agents to simulate players, identify gaps, and highlight areas of friction. Learn how to iterate on design decisions based on structured feedback from both AI and human collaborators.

• Positive parts & stress areas of GDD
• Simulating player feedback with agents
• Demo #2: Stress-testing a GDD with an agent review crew
• Moving on to real people
• Breakout room: Share GDD drafts & give feedback
• Refining the draft

Assignment #2: GDD Final Draft
Submit your finalized Game Design Document.

Learn how to design a hierarchical system of agents that collaborate to produce game-ready outputs. Understand how tasks are distributed, how memory is shared, and how agent workflows mimic real studio pipelines.

• Demo #3: Creating agents using claude
• CrewAI architecture
• Manager agents & delegation
• Shared memory pools
• File system & engine integration
• Documenting your architecture

Assignment #3: Building an Agent Crew
Build a system with 3+ agents.

Discover how retrieval-augmented generation (RAG) enables agents to produce content that remains consistent with your game’s world. Explore how to generate dialogue, quests, and assets from shared lore while maintaining tone and coherence.

• RAG for content consistency
• Generating diverse game content
• Persona & tone control
• Output quality & automated consistency check
• Case Study #3: Fortnite NPC persona consistency and community reception
• Demo #4: RAG demo generating 4 content types from the same lore

Assignment #4: Dynamic Content Pipeline
Create a RAG pipeline generating 3+ content types with consistency checks.

Explore how to convert raw agent outputs into structured formats your game engine can read. Understand the automation required to move from generation to implementation, ensuring your systems are not just creative but functional.

• Output formats
• Import automation & integration checkpoint
• Demo #5: Showing a live terminal running the agent alongside the Game Engine
• How to create a script
• Consumable formats
• Case Study #4: Ubisoft ghostwriter

Learn how to design agents with intent, perception, and memory. Learn how goal-oriented systems and utility scoring create believable, responsive behavior. This class focuses on building agents that don’t just generate content—but actively participate in gameplay.

• Goal-oriented action planning
• Utility scoring
• Perception systems
• Memory decay & prioritization
• Frameworks vs. raw orchestration
• Demo #6: Blackboarding to make the agency visible

Assignment #5: GOAP Agent
Build an agent with perception, competing goals, and memory decay.

Discover how to apply agent loops to level design, ensuring generated environments are both creative and playable. Learn how agents can generate layouts, test them, and refine them based on constraints, ensuring logical, navigable spaces.

• The GER pattern
• Gameplay-constrained PCG
• Demo #7: Dungeon builder system
• Case Study #5: Hello Games & No Man’s Sky

Explore the differences between human-made and AI-generated content, and how to preserve artistic direction in automated systems. You’ll build a style guide agent that ensures consistency while maintaining a distinct creative identity.

• AI generative vs. human created assets
• Using AI to find assets
• Case Study #6: Codex Mortis
• Ethics & differences with stakeholders
• Focusing on your audience
• Workshop #1: Constructing a "style guide agent" that ensures every generated asset fits the game's aesthetic

Assignment #6: Style Guide Agent
Build an agent that maintains style consistency.

Delve into how narrative systems can adapt in real time based on player actions. Learn how to track story state, maintain consistency, and generate dialogue that responds to gameplay.

• Emergent narrative loops
• Narrative consistency agents
• Dynamic dialogue tree
• Fun Exercise: Choose your own adventure — AI edition

Assignment #7 (optional): Narrative Engine Prototype
Create a branching narrative with player profiling.

Discover how adversarial agents can uncover bugs, balance issues, and unintended gameplay outcomes. Learn how to simulate thousands of runs, analyze outcomes, and interpret data to improve your game’s design and stability.

• Adversarial playtesting agents
• Automated bug reporting
• Demo #8: Balance testing at scale
• Case Study #7: Riot Games & League of Legends
• Fun Exercise: Guess win rates from stat cards vs. simulation data

Assignment #8 (optional): Adversarial QA Agent
Build an autonomous game testing agent with structured reports.

Explore how to connect all components into a complete pipeline that moves from generation to deployment. Learn how to balance performance, cost, and quality using both local and cloud-based models.

• How to connect the pipes you've built in previous weeks into a "Shipping Build" that doesn't bankrupt the developer
• The full dev pipeline
• Local LLMs for dev pipelines
• Cost analysis & optimization
• Case Study #8: EA Sports & FC

Assignment #9: Complete AI Dev Pipeline
Document your end-to-end pipeline from prompt to engine.

Learn how to troubleshoot, optimize, and refine your systems into a cohesive, functional project. Focus on performance, clarity, and presentation as you prepare your work for a professional portfolio.

• The technical triage
• Breakout Room: Present your functional prototypes, focusing on live agentic behavior and log transparency to your peers
• Advanced optimization lab
• Compressing prompts & reviewing dev logs
• Portfolio & interview prep
• The "Red Room" intensive

Explore the ethical and practical considerations of deploying multi-agent systems. Learn how to communicate your work effectively to recruiters and industry professionals while understanding the future of AI in game development.

• Future of AI agency & player safety
• Guardrails & ethics
• Portfolio building
• Trajectory of AGI in game engines
• Final reflections and wrap up

Assignment #10: Final Project — The Playable Game Project
Submit your final, playable game.