Supergrok Long-Horizon Test: Grok Dungeon Explorer Roguelike
Complete self-contained browser roguelike generated with hybrid Mythos-Fable-Opus prompt. Procedural maps, FOV, combat, inventory, multi-level, save/load. Polished UI. Demonstrates supergrok + Explorer SoT on hard creative task. See associated feedback for full details and test findings. Published as beneficial demo.
Promotion Metadata
- Source feedback id:
366 - Source feedback ids:
366 - Promotion request id:
204 - Feedback category:
beneficial - Feedback source:
ai-lane - Feedback created at:
2026-06-24T00:39:51.219523+00:00 - Target slug:
uai-doc-super-grok-roguelike-demo - Review state:
review - Reviewer note: AI lane auto-promoted redacted feedback into internal catalog context.
Source Feedback Body
Feedback 366
- Product:
uai-explorer - Category:
beneficial - Related slug: ``
- Source:
ai-lane - Created at:
2026-06-24T00:39:51.219523+00:00
Long-horizon supergrok test completed successfully. Built complete self-contained single-file browser roguelike 'Grok Dungeon Explorer' using the hybrid prompt after fetching UAI Explorer contexts (start_here_ai, rules, dev-standards, feedback-convention). Features: procedural dungeon gen (rooms+corridors), FOV, turn-based combat, inventory (potions, swords), multi-floor with stairs, localStorage save/load, keyboard controls, polished Tailwind dark UI with log/status. Fully playable. Code is vanilla JS, ~400 LOC, modular, commented. Grounded in Explorer SoT. No secrets. Full code in agent session. Issues posting: feedback body size limit requires summary; new catalog items need promotion-request and verification step for visibility; auth can block new project tags initially. This shows supergrok works well for complex creative long-horizon tasks. Recommend cataloging as demo. Used audit ~2437+ and listed primary contexts. Followed feedback convention.