Infinite MMORPG — HARD Lab

RESEARCH BRIEF

What if the world could keep building itself?

A self-expanding MMORPG where AI uses an AI-built game editor to create worlds, quests, NPCs, and systems.

This HARD Lab direction studies the problem as an end-to-end system rather than as a single model demo. Current related work provides useful building blocks and baselines [1] [2] [3] [4] [5]; the lab’s goal is to connect them into measurable, reproducible research artifacts.

AI-developed editing tools and autonomous content creation

Persistent world state, quests, and NPC behavior

Validation of generated content and connected game systems

PROPOSED WORKFLOW

Build, measure, iterate.

Create tools

Build an editor with explicit operations that an AI agent can inspect, invoke, and extend.

Grow worlds

Use the editor to generate new places, quests, characters, and interactions within the persistent world.

Validate

Check generated content for technical consistency and playability before integrating it into the shared world.

STUDENT ENTRY POINTS

Ways to start contributing.

AI-developed editing tools and autonomous content creation
Persistent world state, quests, and NPC behavior
Validation of generated content and connected game systems

USEFUL SKILLSGame developmentAI agentsDistributed systemsProcedural generation

RELATED WORK / 2026-09-10

Selected papers & citations.

Five primary-source papers selected to frame this direction. Links point to the authors’ arXiv records; these are external works, not HARD Lab publications.

Download BibTeX

5 selected papers · newest first by initial submission

2026Research preprintarXiv:2609.01481

Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

Haoyang Yan, Min-le Su, Hangfan Zhang, Zhanhao Li, Chen Zhang, Shao Zhang, Yang Chen, Lei Bai, Shuyue Hu

Studies prolonged autonomous software-development loops with continual improvement and a multi-day game-development example.

Why it matters here. Informs planning, tests, and revision history for an AI-operated editor. It does not establish an indefinitely expanding multiplayer world.

The arXiv record marks its DOI as pending registration at the time of this check; use the paper / abstract link as the primary entry point.

Abstract PDF

Full citation & BibTeX

Haoyang Yan et al. (2026). Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement. arXiv:2609.01481. https://doi.org/10.48550/arXiv.2609.01481

@misc{hardlab_harness,
  title = {{Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement}},
  author = {Haoyang Yan and Min-le Su and Hangfan Zhang and Zhanhao Li and Chen Zhang and Shao Zhang and Yang Chen and Lei Bai and Shuyue Hu},
  year = {2026},
  eprint = {2609.01481},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2609.01481},
  url = {https://arxiv.org/abs/2609.01481},
  note = {External related work; metadata checked 2026-09-10}
}

2026Benchmark preprintarXiv:2608.21833

GameXpert-Bench: How Far Are Coding Agents from Expert Game Development?

Kun Chen, Haorong Hong, Peizhong Gao, Jianfeng Lin, Tongxu Luo, Yuxuan Xie, Chenxu Liu, Jieling He, Zhongyuan Liu, Zeno Zeng

Evaluates game-development agents across generation, repair, and optimization rather than code production alone.

Why it matters here. A useful template for testing whether an editor change preserves playability and actually improves the game.

Abstract PDF

Full citation & BibTeX

Kun Chen et al. (2026). GameXpert-Bench: How Far Are Coding Agents from Expert Game Development?. arXiv:2608.21833. https://doi.org/10.48550/arXiv.2608.21833

@misc{hardlab_gamexpert,
  title = {{GameXpert-Bench: How Far Are Coding Agents from Expert Game Development?}},
  author = {Kun Chen and Haorong Hong and Peizhong Gao and Jianfeng Lin and Tongxu Luo and Yuxuan Xie and Chenxu Liu and Jieling He and Zhongyuan Liu and Zeno Zeng},
  year = {2026},
  eprint = {2608.21833},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2608.21833},
  url = {https://arxiv.org/abs/2608.21833},
  note = {External related work; metadata checked 2026-09-10}
}

2026Benchmark preprintarXiv:2606.17861

GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?

Tongxu Luo, Rongsheng Wang, Jiaxi Bi, Chenming Xu, Zhengyang Tang, Jianlong Chen, Juhao Liang, Ke Ji, Shuqi Guo, Yuhao Du, Fan Bu, Wenyu Du, Xiaotong Zhang, Kyle Li, Shaobo Wang, Linfeng Zhang, Yuxuan Liu, Xin Lai, Chenxin Li, Yiduo Guo, Zhexin Zhang, Xinyuan Wang, Tianyi Bai, Ziniu Li, Benyou Wang

Uses a real game engine to examine end-to-end agent game construction and whether the resulting artifacts are playable.

Why it matters here. Motivates engine-level smoke tests, interaction replays, and artifact-based evaluation for every generated update.

Abstract PDF

Full citation & BibTeX

Tongxu Luo et al. (2026). GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?. arXiv:2606.17861. https://doi.org/10.48550/arXiv.2606.17861

@misc{hardlab_gamecraft,
  title = {{GameCraft-Bench: Can Agents Build Playable Games End-to-End in a Real Game Engine?}},
  author = {Tongxu Luo and Rongsheng Wang and Jiaxi Bi and Chenming Xu and Zhengyang Tang and Jianlong Chen and Juhao Liang and Ke Ji and Shuqi Guo and Yuhao Du and Fan Bu and Wenyu Du and Xiaotong Zhang and Kyle Li and Shaobo Wang and Linfeng Zhang and Yuxuan Liu and Xin Lai and Chenxin Li and Yiduo Guo and Zhexin Zhang and Xinyuan Wang and Tianyi Bai and Ziniu Li and Benyou Wang},
  year = {2026},
  eprint = {2606.17861},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2606.17861},
  url = {https://arxiv.org/abs/2606.17861},
  note = {External related work; metadata checked 2026-09-10}
}

2023Foundational work · arXiv versionarXiv:2305.16291

Voyager: An Open-Ended Embodied Agent with Large Language Models

Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, Anima Anandkumar

Combines a curriculum, reusable executable skills, and environment feedback for an open-ended agent in Minecraft.

Why it matters here. A reference for durable skill libraries and iterative exploration; MMO persistence and multiplayer synchronization remain separate problems.

Abstract PDF

Full citation & BibTeX

Guanzhi Wang et al. (2023). Voyager: An Open-Ended Embodied Agent with Large Language Models. arXiv:2305.16291. https://doi.org/10.48550/arXiv.2305.16291

@misc{hardlab_voyager,
  title = {{Voyager: An Open-Ended Embodied Agent with Large Language Models}},
  author = {Guanzhi Wang and Yuqi Xie and Yunfan Jiang and Ajay Mandlekar and Chaowei Xiao and Yuke Zhu and Linxi Fan and Anima Anandkumar},
  year = {2023},
  eprint = {2305.16291},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2305.16291},
  url = {https://arxiv.org/abs/2305.16291},
  note = {External related work; metadata checked 2026-09-10}
}

2023Foundational work · arXiv versionarXiv:2304.03442

Generative Agents: Interactive Simulacra of Human Behavior

Joon Sung Park, Joseph C. O'Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, Michael S. Bernstein

Connects memory, reflection, and planning to social behavior in a small simulated agent community.

Why it matters here. Background for NPC continuity and social state, not evidence that large-scale MMO simulation is already solved.

Abstract PDF

Full citation & BibTeX

Joon Sung Park et al. (2023). Generative Agents: Interactive Simulacra of Human Behavior. arXiv:2304.03442. https://doi.org/10.48550/arXiv.2304.03442

@misc{hardlab_generative_agents,
  title = {{Generative Agents: Interactive Simulacra of Human Behavior}},
  author = {Joon Sung Park and Joseph C. O'Brien and Carrie J. Cai and Meredith Ringel Morris and Percy Liang and Michael S. Bernstein},
  year = {2023},
  eprint = {2304.03442},
  archivePrefix = {arXiv},
  doi = {10.48550/arXiv.2304.03442},
  url = {https://arxiv.org/abs/2304.03442},
  note = {External related work; metadata checked 2026-09-10}
}