Every Dungeon
Assembled by AI.
Every Run, Unique.
Majestic Mind Games builds action roguelites where levels and enemies are generated on the fly. No two playthroughs share the same corridors or the same fight.
Roguelites That Think, Not Scripts That Repeat
Our proprietary AI generates every dungeon layout, enemy set, and room encounter fresh. Design intuition encoded into constraint grammars, not hand-placed triggers.
Ruinveil
A fast-paced action roguelite where every dungeon is built by our AI in real time. The layout you run today will never exist again. The enemies you face tonight were assembled fresh from an ecology that tracks your playstyle.
Three AI Systems, One Dungeon
Our generation stack links a level grammar, an enemy ecology, and a seed replay system. Each layer feeds the next so the whole dungeon feels authored, not assembled.
Level Grammar Engine
Spatial rules encode dungeon tension, pacing, and discovery into constraint sets. The generator assembles rooms from valid chains, guaranteeing every layout is completable and interesting.
Enemy Ecology System
Enemies are placed by role in an ecological context: predators, harassers, sentinels. Each room's inhabitants interact with each other and with the layout so encounters feel designed, not randomly dumped.
Seed Replay System
Every run encodes to a seed string. Replicate a run exactly, share it with friends, or submit it for speedrunning. The AI is transparent: you can always trace back to where a dungeon came from.
"The dungeon feels like someone designed it for me. I checked the seed. It was assembled in 74ms from scratch when I hit the button. That is the future of the genre."verdant_ruin, early-access participant, closed beta July 2026
From the Devlog
All Posts
Closed Beta Feedback: 40 Players, 200 Hours, What We Fixed
We ran a 40-player closed beta over two weeks. This post summarizes the feedback we received, the bugs we found, and the three design changes we made in response.
Read Post
Dynamic Difficulty: Matching the Challenge to the Player
Hard-coded difficulty tiers do not work with a procedural dungeon. Our approach dynamically adjusts room density, enemy aggression, and loot scarcity based on run-level performance signals.
Read Post
Measuring Run Variety: How We Know the AI Is Working
We built internal metrics to track layout uniqueness, enemy set overlap, and encounter repetition across runs. The data tells us when the generator gets lazy before players do.
Read PostFollow the Build. Play the Beta.
Ruinveil is in active development. Read our devlog to see how it works, or wishlist the game to be notified when beta access opens.