Naavik Gaming Podcast How AI Continues to Reshape Game QA
Dec 16, 2025
Christoffer Holmgård, CTO of modl.ai, and Julian Togelius, professor and co-founder, explore how generative AI is revolutionizing game QA. They discuss the shift to black-box testing through screen-seeing AI agents, allowing for no-code automation that frees testers from traditional coding. The duo highlights how this technology enhances bug detection and leverages multimodal models for better game feedback. They also delve into the evolving roles of human testers, emphasizing that AI is not a replacement, but an augmentation, streamlining the QA process.
AI Snips
Chapters
Transcript
Episode notes
Black-Box Testing Becomes Practical
- Vision-based black-box testing matches how real players interact with games and avoids per-game engine integrations.
- Generative multimodal models finally made pixel-based decision-making and English task prompts practical for QA.
Genre & Access Shape AI Strategy
- Game genre and available information fundamentally change which AI approaches work best.
- Building per-game sensors or deep integrations is expensive and often impractical at studio scale.
Give The Agent Compact Game Context
- Provide the AI with concise textual context about the game (genre, goals, rules) to improve behavior.
- Also supply tuned perceptual models to locate UI elements and in-game objects reliably.