it is an open effort to find a program that can reliably win NetHack 3.6.6: you can write a bot by hand, let claude code / codex / opencode improve one, or build your own harness that improves the thing that improves the bot
me and my lab switched to the goal-based research this year (we worked a lot in the idea-based framework for a last couple of years) -- and decided that we would like to tackle the NetHack, it is one of those nasty roguelike that is still somehow not solved with all fancy AI models (yes, even astra) -- probably on the way something useful will appear
you can pull someone else's bot, run experiments locally and register the results. every registered program is linked to your github repo and exact source commit
i always wanted to try building something that will use github as both social and storage layer (and it seems to work quite well in our case, simplified things a bit and i think this approach is quite cool)
there is a shared frontier across 73 starting identities -- your bots dont have to be the best at everything to contribute something useful, e.g. if you love to roleplay gnomes only, so be it
to make thing easier (for both harness devs and people who are just playing around), we provide a loop that does some sort of an evolutionary optimization process (quality-diversity, etc), ofc you don't have to use it. but you can start with in minutes (given you have codex or claude subs)
you can also bring your own search, your own harness, whatever -- the common ground is the evaluation. public seeds are there for people to overfit (in nethack even overfitting is hard), and we separately re-evaluate on hidden seeds to check whether the improvements actually transfer
one of my goals was to actually try to make it at least a bit fun -- hope you enjoy the web-site and easter eggs","source_kind":"rss","canonical_source_url":""}],"modelAnswers":[{"id":"4ffcb29f-4cdb-4fa3-ad23-ba8b931fa1d4","model_id":"3e37ec80-3eb9-4d78-9dd4-b5fa2206be5a","model_name":"Kimi-K3","provider":"alibaba","version":"kimi-k3","summary":"Hacker News 出现 Show HN 帖子《NetHackers》，由用户 vokneruk 于 2026-09-23 发布，项目链接指向 nethackers.dunnolab.ai。来源仅含标题与 URL，未附功能描述，产品实际用途与细节均无法确认。","answer":"来源事实：可用信息仅来自一条 Hacker News 的 Show HN 帖子（S1）。标题为《Show HN: NetHackers》，发布者为用户名 vokneruk，发布时间为 2026-09-23T23:04:20Z，项目链接为 https://nethackers.dunnolab.ai/。来源未提供任何正文、功能说明、截图或技术细节，因此无法从来源本身确认该产品的具体功能、实现方式或目标用户。 模型归纳（非权威结论）：'Show HN' 是 Hacker News 社区的项目自荐发帖惯例，通常表示开发者本人对外公开展示新作品，可据此推断 NetHackers 可能处于早期公开发布阶段。链接域名 dunnolab.ai 暗示项目可能出自名为 DunnoLab 的团队或实验室；.ai 域名后缀与输入标注的'AI应用'类别相符，产品可能涉及 AI 相关能力，但未经来源验证。名称 'NetHackers' 在字面上可能让人联想到经典 Roguelike 游戏 NetHack 或网络安全/黑客主题，这仅为命名层面的推测，不能视为事实。 不确定性：产品实际功能、是否确为 AI 应用、DunnoLab 是否为正式实体、vokneruk 与项目的关系、开源状态与商业模式均未知，建议直接访问来源链接或原帖评论区核实。","citations":[{"source_id":"3d9faaa9-d25e-4737-9e38-79cfb546b3e7","title":"Show HN: NetHackers","url":"https://nethackers.dunnolab.ai/"}],"generated_at":"2026-09-24 00:15:22","entities":[{"name":"NetHackers","type":"product","confidence":0.9},{"name":"DunnoLab","type":"organization","confidence":0.55},{"name":"vokneruk","type":"person","confidence":0.7},{"name":"Hacker News","type":"organization","confidence":0.75}],"uncertainties":["来源未提供任何功能描述，NetHackers 的实际用途与功能未知","是否确属 AI 应用无法从来源验证，仅有输入中的类别标注","DunnoLab 是否真实存在的团队或公司未获确认，仅从域名推测","发布者 vokneruk 与项目方的关系（开发者或转载者）未知","原帖正文、社区反响（点赞数、评论）等互动数据不可得","名称是否与 NetHack 游戏或网络安全主题相关仅为推测"]},{"id":"09cb0ef8-dcdd-41bd-8874-c119362e3d9e","model_id":"7b111a8a-1f89-4a68-b63a-6690af3bfef4","model_name":"gpt5.6","provider":"bytedance","version":"doubao-seed-evolving","summary":"UTC时间2026年9月23日，发布者vokneruk在Show HN栏目发布AI应用类项目NetHackers，项目关联站点为nethackers.dunnolab.ai，目前暂未有更多项目细节对外公开。","answer":"根据当前唯一可查的来源S1信息，可确认的公开事实包括：该内容属于Show HN栏目的项目发布条目，发布账号为vokneruk，发布时间为UTC时间2026年9月23日23时04分20秒，展示的项目名称为NetHackers，项目类别为AI应用，对应的关联站点地址为https://nethackers.dunnolab.ai/。模型基于有限信息归纳如下：从项目命名与分类标签推测，NetHackers可能是一款结合AI能力、面向网络技术相关场景打造的工具类产品，名称组合了“网络（Net）”与“技术探索者/黑客（Hackers）”的语义，或面向网络技术从业者、爱好者群体提供相关服务，但该判断仅为基于名称和分类的合理推测，无更多公开信息佐证，不构成权威结论。当前仅能获取该项目的基础发布条目信息，发布方并未披露项目的具体功能、技术实现、目标用户、使用规则、收费模式、开发进度等核心信息，也未说明自身主体性质，以及项目与域名中dunnolab标识所属主体的关系，外部无法判断项目的实际可用性、落地价值与合规性，所有关于项目细节的结论都需等待发布方后续披露，或访问关联站点验证，不应随意传播未经证实的推测内容。","citations":[{"source_id":"3d9faaa9-d25e-4737-9e38-79cfb546b3e7","title":"Show HN: NetHackers","url":"https://nethackers.dunnolab.ai/"}],"generated_at":"2026-09-24 00:14:34","entities":[{"name":"NetHackers","type":"product","confidence":0.95},{"name":"vokneruk","type":"person","confidence":0.8},{"name":"Show HN","type":"topic","confidence":0.95},{"name":"dunnolab","type":"brand","confidence":0.6}],"uncertainties":["NetHackers的具体功能、技术路径、目标用户、使用规则、运营状态等产品核心信息未披露，无法确认","发布者vokneruk的主体性质（个人/组织）及背景信息无法确认","域名内dunnolab标识的所属主体性质，及其与NetHackers项目的关联关系无法确认","NetHackers的实际可用性、落地价值、合规性无公开信息佐证","NetHackers是否正式上线、是否对外开放使用暂不明确"]},{"id":"e41be10a-7854-4ae6-b797-6330ad565899","model_id":"43bd6f33-1b6d-422d-8673-f10a76194d86","model_name":"智谱GLM5.2","provider":"bytedance","version":"glm-5-2-260617","summary":"该项目名为NetHackers，由用户vokneruk在Hacker News上展示。项目关联域名为nethackers.dunnolab.ai。目前来源仅提供项目名称及发布信息，具体功能、应用场景与技术细节有待进一步披露。","answer":"根据来源S1提供的信息，事实部分如下：该项目名为“NetHackers”，由用户vokneruk在Hacker News平台上以“Show HN:”的形式进行展示。项目的发布时间标注为2026年9月23日，并提供了可访问的网址（https://nethackers.dunnolab.ai/）。从标题的“Show HN:”前缀可以确定，这是一个面向技术社区的分享项目。 在模型归纳方面，根据项目名称“NetHackers”及域名中的“dunnolab.ai”，推测该项目可能与人工智能、开发者工具或网络安全相关。“dunnolab”可能代表一个实验室或研究团队的品牌。然而，这部分内容属于基于命名习惯的推测，不作为权威结论。 在不确定性方面，由于来源仅包含标题、发布者和链接，缺乏对项目功能的具体描述，因此无法确认NetHackers的确切产品形态、目标用户群体以及它是开源项目还是商业产品。此外，发布时间标记为2026年，这使得该数据是真实事件还是模拟预测数据存在疑问。综上，当前仅能确认其发布行为及基础标识信息。","citations":[{"source_id":"3d9faaa9-d25e-4737-9e38-79cfb546b3e7","title":"Show HN: NetHackers","url":"https://nethackers.dunnolab.ai/"}],"generated_at":"2026-09-24 00:13:10","entities":[{"name":"NetHackers","type":"product","confidence":0.9},{"name":"vokneruk","type":"person","confidence":0.85},{"name":"Hacker News","type":"organization","confidence":0.8},{"name":"dunnolab.ai","type":"brand","confidence":0.75}],"uncertainties":["项目NetHackers的具体功能、应用场景和技术架构未知。","发布者vokneruk的具体背景及所属机构未明确。","域名dunnolab.ai背后的实体及其与项目的关系未明确。","由于发布时间标注为2026年，该事件是否为真实历史事件或为预测性/模拟数据存在不确定性。"]}],"modelInsight":{"topic_id":"1725d295-26cf-4484-abc7-f118943071df","overview":"综合三项解读，Hacker News 的 Show HN 帖子《NetHackers》由用户 vokneruk 发布，项目链接指向 nethackers.dunnolab.ai。三者均认为当前公开信息有限，只包含项目名称、发布信息或标题与URL，未提供明确功能描述、应用场景或技术细节。关于时间，gpt5.6 与 Kimi-K3 指出发布日期为2026年9月23日，其中 gpt5.6 标注UTC；智谱GLM5.2未直接说明。关于定位，gpt5.6称其为AI应用类项目，但另外两家未确认该分类，并强调实际用途无法确认。未确定项包括是否属于AI应用、具体功能、技术实现与使用场景。可共同确认的仅是项目存在、发布者与链接，产品价值与技术路线仍需更多公开证据。","common_points":["均确认存在 Hacker News 的 Show HN 项目或帖子 NetHackers","均确认发布者为 vokneruk","均确认关联站点或域名为 nethackers.dunnolab.ai","均认为当前公开信息有限，缺少明确功能描述、应用场景或技术细节"],"disagreements":["gpt5.6 将项目明确归类为 AI 应用类项目，而智谱GLM5.2 与 Kimi-K3 未确认该分类，仅表述为项目、帖子或无法确认实际用途","对公开信息范围的表述存在差异：智谱GLM5.2 称来源仅提供项目名称及发布信息，Kimi-K3 强调来源仅含标题与URL，gpt5.6 额外给出 UTC 发布时间"],"disagreement_score":15,"confidence":82,"evidence_refs":["智谱GLM5.2 解读：NetHackers 由 vokneruk 在 Hacker News 展示，关联 nethackers.dunnolab.ai，细节待披露","gpt5.6 解读：2026年9月23日 UTC，vokneruk 在 Show HN 发布 AI 应用类项目 NetHackers，关联站点为 nethackers.dunnolab.ai","Kimi-K3 解读：2026-09-23 出现 Show HN 帖子《NetHackers》，vokneruk 发布，链接指向 nethackers.dunnolab.ai，功能未附"],"generated_at":"2026-09-24 00:16:17"},"entities":[{"id":"52af44a2-1d72-4fd2-bab2-6162abdd35eb","name":"Show HN","type":"topic","confidence":0.95,"visibility_score":100,"snapshot_count":1,"mentioned_count":1,"cited_count":1},{"id":"e442b5b2-4dac-4c24-8198-a04ced2aed1c","name":"NetHackers","type":"product","confidence":0.9,"visibility_score":100,"snapshot_count":3,"mentioned_count":3,"cited_count":3},{"id":"ae33f426-9ffe-488b-826e-2c5666aa63cc","name":"vokneruk","type":"person","confidence":0.85,"visibility_score":100,"snapshot_count":3,"mentioned_count":3,"cited_count":3},{"id":"338fba40-374f-4302-a7f0-80a15ca2eb18","name":"Hacker News","type":"organization","confidence":0.8,"visibility_score":100,"snapshot_count":2,"mentioned_count":2,"cited_count":2},{"id":"9ef0aa3b-bb27-4995-b8c8-6c75f9e604d0","name":"dunnolab.ai","type":"brand","confidence":0.75,"visibility_score":100,"snapshot_count":1,"mentioned_count":1,"cited_count":1},{"id":"ce9a1aa3-ca71-46b9-ab41-cfdaeac168e6","name":"dunnolab","type":"brand","confidence":0.6,"visibility_score":100,"snapshot_count":2,"mentioned_count":2,"cited_count":2}],"modelStats":[{"answer_id":"4ffcb29f-4cdb-4fa3-ad23-ba8b931fa1d4","model_id":"3e37ec80-3eb9-4d78-9dd4-b5fa2206be5a","model_name":"Kimi-K3","provider":"alibaba","generated_at":"2026-09-24 00:15:22","citation_count":1,"entity_count":4,"uncertainty_count":6},{"answer_id":"09cb0ef8-dcdd-41bd-8874-c119362e3d9e","model_id":"7b111a8a-1f89-4a68-b63a-6690af3bfef4","model_name":"gpt5.6","provider":"bytedance","generated_at":"2026-09-24 00:14:34","citation_count":1,"entity_count":4,"uncertainty_count":5},{"answer_id":"e41be10a-7854-4ae6-b797-6330ad565899","model_id":"43bd6f33-1b6d-422d-8673-f10a76194d86","model_name":"智谱GLM5.2","provider":"bytedance","generated_at":"2026-09-24 00:13:10","citation_count":1,"entity_count":4,"uncertainty_count":4}],"topicAssets":[],"relatedTopics":[{"id":"716acce4-331a-428c-be53-86ef162d7837","title":"假牙粘附剂有股松油味","category":"AI与科技","heat_score":73,"discussion_volume":14,"summary":"假牙粘附剂有松油味或异味，主要是因为部分“天然”品牌使用松香树脂作为粘合基底，带来浓郁松木气味；同时也可能源于配方中的樟脑、薄荷等抑菌剂，或酒精溶剂的挥发。此外，若保存不当导致管口细菌滋生，也会产生严重的腐败异味。","updated_at":"2026-10-21 07:40:18"},{"id":"7acb0c81-6aee-4165-b80c-4968d199d0ab","title":"茶船古道茉莉六堡茶茶饮","category":"AI与科技","heat_score":71,"discussion_volume":8,"summary":"识别品牌为'茶船古道'，产品为500mL无糖茉莉六堡茶，由广西华林食品受托生产。该饮品主打0糖0脂0卡，融合传统六堡茶底与现代健康理念。其市场定位覆盖追求健康的年轻群体及肠胃调理人群，参照无糖茶饮市场，预估售价在5-8元区间。","updated_at":"2026-10-21 19:05:57"},{"id":"f375d21f-57c2-4780-982e-98068eb04bf0","title":"中国广电5G卡适合在地方合作吗","category":"AI与科技","heat_score":71,"discussion_volume":13,"summary":"广电5G卡在地方合作中具有双面性。对于政企数字化项目（如安防、专网），其政策支持力度大且成本较低，较为适合��但对于纯个人号卡的渠道代理，虽然初期具备价格优势和低门槛，但面临监管收紧和售后服务不完善的风险，需谨慎评估。","updated_at":"2026-10-21 07:28:35"},{"id":"e0220c65-8cb4-4da8-b025-d11f63782d3b","title":"哪吒汽车重组是不是黄了？","category":"AI与科技","heat_score":65,"discussion_volume":5,"summary":"根据联网检索结果，以下是我整理的汇总分析： ```json { \"summary\": \"哪吒汽车重组并未\"黄了\"，而是正处于破产重整的关键推进阶段。2026年9月11日，合众新能源第四次债权人会议披露《重整计划（草案）》，浙江太乙圣莲拟出资30亿元获约70.62%股权。债权人线上表决截止日为10月12日，其后仍需嘉兴中院裁定批准方可生效。此前2026年3月原方案曾因未按期提交被法院裁定转入清算，本次为重新推进的重整方案。\", \"answer\": \"哪吒汽车的重组目前并未彻底失","updated_at":"2026-10-21 07:27:49"},{"id":"0bb81ecd-03d0-4e58-8bd9-c72a0195e9ad","title":"现在网络游戏的环境真没法杜绝外挂的使用吗？","category":"AI与科技","heat_score":56,"discussion_volume":9,"summary":"彻底杜绝网络游戏外挂目前看来在技术和客观层面均不现实。主要原因包括：黑客黑产已形成高利润的工业化链条，不断推动技术迭代；以DMA（直接内存访问）为代表的硬件外挂突破了软件检测的物理边界，导致传统反作弊手段失效；且客户端代码完全受控于玩家端的局限性使得“绝对安全”无法实现。尽管有kernel级别防护和法律打击，但只能达到“抑制”而非“根除”。","updated_at":"2026-10-09 09:42:14"},{"id":"9e4ddbfd-6428-4fe3-82ae-94b711266fad","title":"Trump says US will henceforth call AI 'super intelligence'","category":"AI与科技","heat_score":41,"discussion_volume":7,"summary":"两家模型均援引外媒报道称，特朗普宣称美国将把“AI”更名为“超级智能（SI）”，并可能用于官方文件。但截至当前，均无白宫正式文件或行政命令确认，事件处于“媒体宣称、官方未证”阶段，真实性与具体执行细节仍有待核实。两则解读信息高度一致，未呈现实质差异。","updated_at":"2026-10-01 08:30:05"},{"id":"hot-demo-001","title":"AI Agent 正在进入企业工作流","category":"AI与科技","heat_score":40,"discussion_volume":6,"summary":"企业开始把 AI Agent 从单点聊天工具接入真实业务流程，热点讨论集中在可靠性、成本与自动化边界。","updated_at":"2026-10-21 10:47:28"},{"id":"f56154db-f68c-488f-bc54-0e9a50318cfa","title":"GPT-6 Sol (Max) Intelligence, Performance and Price Analysis","category":"AI与科技","heat_score":40,"discussion_volume":4,"summary":"综合各模型解读，可确认Artificial Analysis平台存在由theanonymousone于2026-09-22发布的《GPT-6 Sol (Max) 智能、性能与价格分析》报告，标题含(Max)。Kimi-K3还提到同日存在GPT-6 Luna及三份Claude Opus 5.5相关报告，但其他模型未提及。所有模型均表示无法获取具体基准得分、性能数据与定价信息，需查阅原始页面核实。各模型对核心事实的描述高度一致，未发现实质性分歧，仅信息详略不同。未确定项包括具体基准结果、性能指标、价格、对比结论及与其他模型的关系。","updated_at":"2026-09-30 05:33:22"}],"canonical_url":"http://www.aigc.guru/hot/topic/1725d295-26cf-4484-abc7-f118943071df","discussion":null}}