Changelog
This is the day-by-day record of ForgeaX building itself. The engine and platform are written almost entirely by AI in a self-running development loop — and this is its long-term development memory: one page per day of the features it shipped for itself and how it built them.
Hellforge 1.2.1: visible cursor and a brighter public demo
ForgeaX Studio v0.3.27: smoother setup, richer previews, and independent Game Runtime packages
AI-native editor control, video workflows, and a refined Studio workspace
Asset Canvas workflow, stronger scene authoring, and desktop stability
Material Editor, local-space gizmo, multi-select, and Play mode stability
v0.3.22 — Play mode & blank screen fixes, editor UX, panel contributions, video assets
v0.3.21 — Play mode freeze fix, new game blank screen fix, CLI zod dependency
Auto AnimationPlayer on skinned models, generative visual priors, and workspace upload
UE5-style camera fly mode, game action projection, and viewport stability
Play input ownership, MeshRenderer materials UI, and Hellforge character select
Collider debug overlay, safer asset workflows, and AI asset import to game
Faster project startup, stronger Play mode, and smoother onboarding
Engine Tier 2 runtime + Single-realm editor + Gateway asset/component surface + Asset-resident plugins
Tier-2 runtime + editor cleanup + Gateway component APIs + embedded asset plugins
Plugin host architecture + Play mode fixes + Keyboard shortcut forwarding + Game-video QTE timeline
Viewport Gizmo overhaul + Composer image paste + Agent persona i18n + iOS export
Hierarchy multi-select + AI semantic operation layer + editor stability fixes
ECS query caching + viewport gizmo GPU-driven + fx-math standalone
3D generation fixes + workbench i18n + TUI agent terminal
Windows Play mode stabilized + LLM debug proxy for developers
Forge gets long-term memory: the more you talk, the more it knows
R4 frontend lands + viewport 2×2 redesign
Architecture unbundled: four frontend modules go standalone + OSS mirror hardened
Forge learns teamwork: many agents at once
One command — now on Windows too
A UE5-grade plan for the content browser & drag-to-scene, plus an editor QA gate
Full-chain observability: see where it's slow at a glance
Open a project and it just renders + agents get long-term memory
A new asset browser + smoother first-run onboarding
An asset inspector + switch-without-stalling + Windows support
The project goes open source + the desktop app stops crash-looping
Generated 3D gets rigged and animated + automated editor QA
Full bilingual UI + complete devtools panels
Desktop hardening: games no longer stuck on Loading
Animated agent avatars + one source of truth for project data
Slide-along-walls movement + a matured standalone editor
Characters finally cast shadows + a fully standalone editor
A dark-fantasy ARPG sample + upload-a-model-and-use-it
Effects stop washing white + walk-and-jump in imported scenes
Multiple levels per project + a shadow-system upgrade
Characters actually animate + a steadier chat
Text-to-3D lands in your game + agents verify before continuing
Real sky out of the box + text-to-3D models
A build-gate so the AI can't fake "done"
A big editor integration + textured models render live
The core editor panels arrive + Windows support
Drag a model in and it becomes a scene
Editor mode, complete
The editor mode is actually usable
An editor mode gets its roadmap + the engine preps for visual editing
One command to launch the desktop app + a node editor joins
The desktop app runs end to end + crisp in-world text
Export a game as a standalone site + skybox and bloom
A desktop app arrives + a shared game library
The engine gains physics & picking + a rebuilt UI foundation
A dedicated animation workbench + live tool streaming
The engine stands fully on its own + richer materials
One-tap context compaction + a WebGPU software fallback
Skeletal animation + scene & observatory workbenches
The game runtime moves fully to the in-house engine
Confirm before risky actions + engine ambient light & shadows
The engine learns shadows + a new narrative workbench
A refreshed interface + an embeddable workbench
Chat reconnects to the new backend + manage models & keys in settings
A crash-recoverable backend + the engine's first real lighting
Open any folder as a workspace + in-app version & shortcuts
Inline 3D & image preview + @-mention pills
The agent bus comes into the UI + dashboard filters
Chat UI grows up: resizable layout, nested tool calls, sub-agent cards
Close the browser — agents keep working
Live hot-reload preview + session replay
Engine asset system v1 + deep chat polish
From zero to a chat-made spinning cube
Build & runtime architecture takes shape
Project kickoff