Game AI Daily · Friday, October 23, 2026
GAMEAI HOT
Unity MethodReload skips domain reload, experimental feature speeds up iteration
Unity has introduced the experimental MethodReload feature. Developers only need to mark methods and start the watcher, and edits to method bodies take effect immediately in a running Editor or connected Player, with no need to wait for a domain reload.
Engines & tools
Roblox Studio beta adds Lua support for creating text-to-speech asset
In the Roblox Studio beta, AssetService:CreateAssetAsync can convert an AudioTextToSpeech object into a speech audio asset. Developers can generate speech in Studio ahead of time instead of calling text-to-speech at game runtime, which lets speech respond instantly and avoids API rate limits. Generated assets currently cannot have a custom name, description, or creator ID, and are all named TTS_Asset_ followed by the asset ID.
Unity CLI data: over 200 million events show how developers are handing work to coding agents
Unity's official data shows that about 98% of Unity CLI runs between July 20 and October 13, 2026 came from coding agents and other automation, while manual terminal input accounted for just 0.2%.
Wonder Interactive launches Strata, using AI to turn Unreal Engine projects into browser-playable versions
Wonder Interactive has launched Strata, an AI development tool that turns Unreal Engine projects into browser-playable links using WebAssembly and WebGPU.
Tripo Studio reworks rigging and animation features
Tripo Studio has reworked its rigging and animation features for motion that can be used directly in games. The new version supports full-body and hand skeleton rigging and provides five skeleton presets ready for use in engines. Text to Motion generates 1–10 second animations from prompts, and up to five motion segments can be combined. Characters can be auto-rigged, animated, and exported without re-mapping.
Meshy is now available in Hermes
Meshy is now available in Hermes Agent as an official plugin. Users can have Hermes generate 3D models through Meshy's MCP server, such as a treasure chest model for Godot or a printable model converted from a photo of a cup. Hermes Desktop renders models in chat and plays rigged character animations inline, and its gallery saves each model with its prompt. In messaging apps, results arrive as short preview videos.
Research
WorldStream: streaming interactive 3D worlds with hierarchical planning and persistent state
WorldStream introduces an open-ended 3D world model that can expand playable terrain at runtime, respond to open-ended language instructions, and restore visited locations along with their edits. A hierarchical planner outputs macro and micro structural decisions, which deterministic local code then compiles into terrain, roads, settlements, vegetation, and entities. The world record stores plans, seeds, object identities, mutable state, and boundary constraints, so returning chunks can be recompiled from it. Stateful NPCs also update as they act.
LiveAct: heterogeneous action interfaces for real-time multi-agent world models
LiveAct introduces a heterogeneous action interface that lets a multi-agent video world model generate joint visual futures for multiple controlled agents, handling control inputs ranging from game buttons to robot commands. The method reaches agent attribution accuracy of 0.898 on games and 0.942 on multi-robot manipulation, without relying on agent position or identity cropping. Its causal generator runs at 17.70 FPS at 512×448 resolution.
Scaling in games: continued pre-training for embodied agents in diverse virtual worlds
The paper systematically studies how data volume, diversity, and model scale affect embodied agents in continued pre-training, using 120 training configurations and models from 0.8B to 27B parameters. Diversity improves transfer to related worlds but gives limited gains for worlds that differ substantially. Larger models improve data efficiency, but when trained on a single world they generalize worse to distant worlds. The authors argue that building general-purpose embodied agents requires training stages richer than continued pre-training.
D-Skill: Distilling teacher knowledge through agent skills
D-Skill distills the paired trajectory differences between a teacher and a student model into interpretable natural-language skill documents. The student uses these documents at inference time without updating its weights. The paper reports that the method closes 75%–95% of the performance gap between teacher and student on SpreadsheetBench-Verified, ALFWorld, and chess puzzles. Within the same model family, the student's per-task inference cost is 5.7x–8.3x lower than the teacher's.
Industry
Game reverse-engineering community decomp.dev shuts down as AI reversing sparks debate
decomp.dev, the largest game reverse-engineering community, has announced it is shutting down, with administrator Encounter saying malicious users and AI ports that showed no respect for the original projects made the problems worse. Around the same time, an engineer claimed AI finished in a day an MMO reverse-engineering rebuild that took him three years, but many commenters questioned the claim.
Paper Games sets up Nuwa bionic robotics company, Yao Runhao holds 60%
Paper Games is making robots? Yao Runhao holds 60% as new firm registers with 10 million yuan capital
Briefs
- ToM2Vec: Continuous theory-of-mind tokens for multi-agent coordinationarXiv Board & Classic Game AI