AI People adds local LLM support

This is the rare AI-game update that actually changes the economics, not just the vibe: moving inference local turns NPC chat from a recurring cloud bill into a device capability. It also nudges AI People closer to a real testbed for on-device agents, even if the hardware requirements still keep it out of reach for a lot of players.

• –GoodAI is positioning AI People as an offline-first AI sandbox, which is a much stronger story than “cloud chatbot game.”
• –Local inference can eliminate token burn for scenarios, especially if players also disable TTS and speech recognition.
• –The tradeoff is real: 8GB for the LLM alone plus audio tooling means the feature is still aimed at fairly beefy gaming PCs.
• –The update is interesting for developers because it shows how agentic NPCs, voice, and local inference can be packaged into a consumer game.
• –If GoodAI can cut the VRAM floor over time, this could become a template for more self-hosted AI gameplay loops.