Gaming Tech in October 2025: AI Coaches, World Sims, and Spatial Intelligence

Microsoft Rolls Out “Gaming Copilot” for Windows 11

Last week, Microsoft's Windows 11 updates introduced Copilot Vision, Voice, and Actions, features that let users speak to their computer, share on-screen context dynamically, and delegate tasks to local AI agents. The move transforms Windows 11 into what Microsoft calls the AI PC.

Users can now capture content, send summaries, and even automate tasks like organizing files or responding to messages—all without external APIs.

While I'm not too keen on the above features as we already similar things on mobiles and even on PC before AI era, the thing that got me excited the most is the new Gaming Copilot. It's a built-in AI assistant that acts as both coach and companion.

Currently in beta, Gaming Copilot analyzes game data in real time and offers tactical suggestions, build optimizations, narrative hints depending on the game’s genre and it can dynamically recommend (or adjust if you give it the permission) configuration tweaks for better performance.

A local inference mode (running on-device AI) means these adjustments happen without constant cloud processing, ensuring low latency and privacy protection.

xAI Moves Into the Video Game Market

Elon Musk’s AI company, xAI, announced plans this month to enter the gaming industry, marking its first major expansion outside of text and visual model development.

Unlike other AI models that primarily generates text or images, xAI's models are designed to understand 3D environments and comprehend real-world physics, such as how objects move or how light behaves. The company now aims to leverage this technology to enter the gaming sector and to be honest, I'm quite surprised they thought about this so late.

According to Musk’s post on X, the company is building AI-driven “world simulation” games, powered by the same generative architectures used in xAI’s world models research.

These models are capable of simulating physics-consistent environments and emotional NPC (non-player character) interactions, allowing spontaneous storytelling going beyond what the game developers scripted beforehand. The company is even hiring gamers to tutor their AI model Grok and willing to pay upto $100/hour.

General Intuition Raises $134 Million to Teach AI “Spatial Intelligence” Through Gaming

General Intuition, a Swiss-US AI startup, has raised $134 million in seed funding to develop agents capable of deep spatial reasoning. And interestingly, they are using video games footage to train their models.

Founded by former engineers from the video platform Medal, General Intuition uses recorded game footage as training data to help models understand how objects and entities move, collide, and obey physics in three‑dimensional worlds. It’s the same skill that lets a gamer aim ahead of a moving target… or a robot hand successfully catch a falling object.

The company’s research centers in New York and Geneva are now expanding into embodied AI, using reinforcement learning and game engines to build agents that learn these skills not only for games, but also for real-world applications like robotics or drone navigation.

Personally, I'm not sure how their developers are going to pull things off considering the physics in the game engines and the real world is world's apart (pun intended). But the idea is interesting and I'd be keenly following their progress.

Sources: TechCrunch, GGBA