Dreamlands AI Worlds 🪐, Unreal MCP ⚡, Capcom RE:Dox OSS 🔓
AI-Powered Game Creation 🤖
Dreamlands 1.0 launches AI asset and world-building tools
Dreamlands 1.0 is publicly available with AI tools for texturing, material editing, mesh cleanup, automatic UVs, and procedural building and city generation. The company lists plugins for Unity, Unreal, UEFN, Roblox, and several 3D applications, plus a node graph and API for custom pipelines. Plans cost $20–$200 per month; Dreamlands says individual texturing edits take about 10 seconds and cost $0.10. Whenere reports using the platform to turn estate reference photos into editable models for its Jane Austen-inspired Fortnite world.
Unreal MCP connects AI agents to the Unreal editor
Unreal MCP embeds a Model Context Protocol server in Unreal Engine 5.8 and later, allowing AI agents to work directly with projects through editor tools. This Windows walkthrough connects Claude Code by enabling plugins, configuring server startup, and generating a client configuration file. It demonstrates creating named objects, assigning materials, moving a cube, and spacing selected cubes along an axis. Developers can also add custom tools to automate their workflows.
Engines, Tools & Open Source 🛠️
Capcom open sources RE:Dox structured data engine
Capcom has open sourced RE:Dox, a C# structured data engine for .NET, under Apache-2.0. It parses data into compact tokens that support reading, editing, format conversion, and automatic parallel deserialization. A shared converter model supports formats including JSON, JSON5, CBOR, and MessagePack, while JSON5 editing can preserve comments. Compatibility layers cover System.Text.Json, Newtonsoft.Json, and DataContractJsonSerializer; reported parallel deserialization gains reach about 2.8× over System.Text.Json on tested datasets.
Preparing Unity projects for CoreCLR
Unity’s planned CoreCLR transition for Unity 7 brings .NET 10 libraries, tiered just-in-time compilation, and a generational, compacting garbage collector, according to the video. Without domain reloads, static fields and event subscriptions persist between play sessions and need cleanup. Native code retaining pointers into managed memory requires pinning, and plugins need compatibility testing. IL2CPP remains on platforms without JIT support and gains library improvements, but retains its existing garbage collector. Reported performance gains come from community benchmarks of an experimental player.
Graphics & Game Development 🎮
TinyBVH 1.9.1 dev branch adds RDNA3/4 hardware ray tracing
TinyBVH 1.9.1's development branch adds initial RDNA3/4 hardware ray tracing support, contributed by Alessandro Salo of Utrecht University. Its OpenCL kernel uses intrinsics for ray–triangle and ray–bounding-box intersections, plus stack management. The announcement reports performance matching native DXR for primary rays, without providing benchmark conditions.
Sharing Pong game logic between SDL and SFML
OGLDEV builds Pong with SDL and SFML by separating game logic from rendering. A shared game component accepts elapsed time and paddle input, updates movement and collisions, and returns three rectangles for each backend to draw. SDL drives updates through callbacks, while SFML uses an explicit event and rendering loop. The tutorial also explains paddle acceleration, bounding-box collision checks, and position corrections after impacts. A planned Vulkan implementation is deferred to a later video.