Summation with Auren Hoffman Moritz Baier-Lentz on why the future of AI runs on video games
14 snips
Aug 4, 2026 
Moritz Baier-Lentz, investor and founder in gaming and AI who helped build Goldman Sachs’s gaming practice and co-founded General Intuition. He argues video games are the ideal training data for future AI. He discusses play as core to learning. He outlines how game design, player-driven worlds, and social platforms shape next-gen AI and hit games.
AI Snips
Chapters
Transcript
Episode notes
Record Everything Now For Future Agents
- Record as much of your digital life as possible now to build a context window for future agents.
- Moritz recommends Google Takeout, Zoom recordings, emails, calendars and storing them locally (e.g., Obsidian).
Learning Happens During Play Not Just Language
- Human learning largely unfolds in space and time through play rather than pure language abstractions.
- Moritz argues language is a lossy abstraction and world models trained on action data better mirror real learning.
Video Games Are Prime Data For World Models
- Video games are the best proxy for tightly-coupled human action and environment data useful for training world and action models.
- Games provide clear states, short feedback loops, measurable goals, and massive human play data across many environments.