1
Building an LLM Agent to Play Dwarf Fortress
用LLM打造《矮人要塞》智能体：从ASCII界面到生存目标，破解复古策略游戏的AI挑战。
Article URL: https://blog.trine.dev/posts/2026-02-28-df-ai-exp/ Comments URL: https://news.ycombinator.com/item?id=49208128 Points: 1 # Comments: 0
2
Show HN: GPT-5.6 Sol vs. Claude Fable 5 in CNC Red Alert 2
两大AI模型在红色警戒2中兵戎相见，带字幕展示实时决策与战略思维，还有回放文件可细品。
link includes video with thinking captions, map state along with thinking of each player and replay file which can be viewed in game.chronodivide.com …
3
AlphaZero in Sparsely Rewarded Games: Limits and Auxiliary Supervision
DeepMind的AlphaZero在稀疏奖励游戏中表现有限，论文提出辅助监督方法提升性能。
arXiv:2607.08984v1 Announce Type: cross Abstract: AlphaZero has demonstrated that a neural-guided Monte Carlo Tree Search can achieve superhuman perfo…
4
调查发现 31% 的 Steam 用户对游戏中使用 AI 不满，43% 表示可以接受
Steam玩家对游戏AI态度分化：31%不满，43%接受，你站哪边？
IT之家 7 月 8 日消息，据外媒 Gamesradar 今天（8 日）报道，GameDiscoverCo newsletter 对 3800 名 Steam 玩家进行了调查，结果显示，相当多受访者 对 AI 进入电子游戏并不反感 。 在受访者中，高达 23.4% 的人称自己“完全不介意”，另有 …
5
Create your own AI, then watch it battle others in your browser
在浏览器里亲手训练AI智能体，还能观看它们同场竞技，零门槛体验强化学习与涌现行为的进化乐趣。
Article URL: https://kim-ai-gpu.github.io/2026/07/04/introducing-agenlus-browser-rl/ Comments URL: https://news.ycombinator.com/item?id=48784825 Point…
6
Introducing OmniCore: A Neural Brain for Your Game’s NPCs
用神经网络给游戏NPC装上“大脑”，一个尚在打磨的原型，却透着未来感。
Before this ever reaches a formal release, I want other game developers to put it through real conditions — not a polished trailer, not a curated show…
7
IRumAI: Reinforcement Learning for Indian Rummy
首个用强化学习攻克印度拉米棋牌博弈的智能体，让隐藏信息对局告别组合搜索慢推理。
arXiv:2606.21975v1 Announce Type: new Abstract: Despite its massive player base and complex hidden-information dynamics, Indian Rummy has received no …
8
Hierarchical Control in Multi-Agent Games: LLM-based Planning and RL Execution
多智能体游戏中用LLM做高层规划、RL做底层执行，分层控制新范式。
arXiv:2606.20014v1 Announce Type: cross Abstract: Reinforcement learning (RL) has achieved strong performance in sequential decision-making, yet scali…
9
Augmenting Game AI with Deep Reinforcement Learning
将深度强化学习融入游戏AI，探索下一代智能游戏交互与玩法革新。
arXiv:2606.20210v1 Announce Type: new Abstract: Immersion in video games depends not only on graphics, audio, and game mechanics, but also on the qual…
10
A robot is sprinting towards you. Do you want it running on Claude or Grok?
11个LLM在2D大逃杀中厮杀，Claude和Grok谁更能帮你活到最后？
Article URL: https://openrouter.ai/blog/insights/royale-last-agent-standing/ Comments URL: https://news.ycombinator.com/item?id=48576824 Points: 265 #…
11
EgoCS-400K: An Egocentric Gameplay Dataset for World Models
首个大规模自中心视角游戏数据集，为世界模型研究提供独特第一人称玩法数据
arXiv:2606.18180v1 Announce Type: new Abstract: The shift from video generation to interactive world modeling places new demands on data: beyond capti…
12
英伟达黄仁勋会见韩国 KRAFTON、NCSoft 游戏厂商代表，讨论游戏 AI 合作方案
英伟达CEO黄仁勋密会韩国两大游戏巨头，探讨AI+游戏新方向，网吧巧遇Faker还送出绝版RTX 5090签名卡。
IT之家 6 月 7 日消息，据韩联社今日报道，英伟达首席执行官黄仁勋今日在首尔江南区的网吧先后会见韩国游戏公司 KRAFTON 创始人兼董事长蒋柄圭和游戏开发商 NCSOFT 代表金泽辰。 IT之家注：KRAFTON 旗下拥有《绝地求生 PUBG》《云族裔 inZOI》《深海迷航》等系列游戏；NC…
13
Show HN: Rogue-Bench – LLMs play the game Rogue
用经典Rogue游戏评测LLM推理与规划能力，趣味与前沿兼备的新基准。
Article URL: https://iwhalen.github.io/rogue-bench/ Comments URL: https://news.ycombinator.com/item?id=48278735 Points: 1 # Comments: 0
14
《极限竞速：地平线 6》惊现“噩梦”AI 车手，背后玩家身份成谜
《极限竞速：地平线6》惊现超强AI车手bowieknife99，玩家直呼“跑不过三蹦子”，背后身份成谜引热议。
IT之家 5 月 26 日消息，《极限竞速：地平线 6》在高度还原的日本实景中带来了绝佳的游玩体验，但一个虚拟车手却在这款竞速游戏里搅得天翻地覆。如今玩家社群纷纷行动起来，想要终结其恐怖统治，同时深挖这个虚拟角色的来历。 上周，玩家们很快发现了一个反复出现的对手：其表现异常强悍，出招更是变幻莫测，让…
15
It's 2026 so where are all the AI NPCs?
2026年回头审视：AI NPC游戏体验为何仍停留在YouTube演示，而非真正的沉浸式革新？
Article URL: https://www.frisson-labs.com/ai-npcs-2026 Comments URL: https://news.ycombinator.com/item?id=48239307 Points: 4 # Comments: 2
16
My Old MacBook Air Couldn't Handle It — So I Used Google Colab to Train an AI＃1
旧电脑也能跑AI训练？用Google Colab免费云端GPU轻松搞定，新手必看！
Introduction I recently booted up an offline card game I used to love — and couldn't clear the hardest difficulty anymore. I used to be able to beat i…
17
Learning to play Minecraft with Video PreTraining
OpenAI 通过视频预训练让神经网络学会玩 Minecraft，仅需少量标注数据就能制作钻石工具，迈向通用计算机智能体。
We trained a neural network to play Minecraft by Video PreTraining (VPT) on a massive unlabeled video dataset of human Minecraft play, while using onl…
18
Learning Montezuma’s Revenge from a single demonstration
单个人类演示教会AI狂揽74,500分，刷新《蒙提祖马的复仇》最高分记录。
We’ve trained an agent to achieve a high score of 74,500 on Montezuma’s Revenge from a single human demonstration, better than any previously publishe…
19
The International 2018: Results
OpenAI Five对战Dota 2顶级玩家，前30分钟保持优势但最终落败，展现AI竞技潜力。
OpenAI Five lost two games against top Dota 2 players at The International in Vancouver this week, maintaining a good chance of winning for the first …