每日早报 2026-08-17

Chris Mundy, the showrunner on James Gunn's new DC series Lanterns, says the surprise ending of Episode 1 was ‘big and scary’ to commit to.

Lenore Zann, the voice of Rogue in animated Marvel projects since the '90s, published a special video to congratulate Inde Navarrette after she was confirmed to play the live-action version of her character in Marvel's upcoming X-Men reboot.

The new Avengers: Doomsday trailer is here, and some fans believe it has a few clues that suggest the film is hiding two Doctor Dooms – and maybe even a cameo from Cillian Murphy.

Marvel star Simu Liu has offered a positive update on his long-in-development Sleeping Dogs movie adaptation, saying the film has "a REAL path forward" as discussions with game publisher Square Enix continue.

Spider-Man: Brand New Day has officially crossed $2 billion at the global box office, marking the first time Marvel's wall-crawling hero has ever reached the theatrical landmark.

Play your favorite PC, Steam, and Xbox games on the go with the best handheld gaming PCs.

Here are all the big reveals from D23's Disney Experiences Showcase, including the name and major attractions of Villains Land, new details on the rides and land of the Cars-themed Piston Peak National Park, previews of the Encanto and Indiana Jones experiences at Animal Kingdom, the reimagining of Disneyland's Tomorrowland, and even the return of Figment and Dreamfinder!

外卖的胃、久坐的腰、熬夜的脑，职场人的身体被“班味”全面入侵，“脆皮体质”成为职场人的常态标签。当“全民健康数智化建设”被写入国家发展蓝图，是时候解锁职场健康的秘密了。

A股周一开盘：超2600只个股飘红，三大指数集体高开；贵金属、通信设备上涨，酒类、生物科技下跌

每当夜深人静，乐潮电商负责人王雷总会习惯性地复盘拼多多后台的数据。对他而言，根据数据随时调整航向已是常态。就拿店里的一款意式风格地毯来说，当初设计师们原本押宝的是奶油风，但王雷在后台敏锐地捕捉到意式风格的点击率更高，转化率直接翻倍。他果断拍板：调整产能侧重，向意式风格倾斜。

跨国车企续“长约”，锚定中国市场确定性(和音)

今年夏天，海钓这项曾经的运动突然“破圈”。从短视频平台上的海钓直播到朋友圈里的渔获晒照，海钓正在从圈层内的小众爱好，逐步发展为特色垂类消费赛道。而它拉动的远不止一根鱼竿的生意，从渔具装备制造到租船出海服务，再到滨海住宿餐饮，一条由兴趣驱动的多元消费链条正在形成。

中新网8月17日电 据中国外汇交易中心网站消息，中国人民银行授权中国外汇交易中心公布，2026年8月17日银行间外汇市场人民币汇率中间价为：1美元对人民币6.7873元，上调5个基点。

近年来，数字经济蓬勃发展，网络直播成为潮流。越来越多人通过直播休闲、社交、消费、学习、工作，直播行业早已超越单纯的娱乐属性，成为数字经济的重要增长极与就业增收的核心载体。然而，在发展过程中，诱导未成年人消费、恶俗PK惩罚刺激消费等问题时有发生，对整体直播行业的发展产生了不良影响。近日，抖音直播发布关于“违规诱导消费”的治理公告，再次亮明了平台坚决打击违法违规直播的鲜明态度。

编者按： 　　家政服务业是重要的民生行业，在改善民生、稳定和扩大就业、促进城乡融合发展、带动扩大消费等方面发挥重要作用。近期，商务部会同多部门印发《关于促进家政服务业高质量发展的若干政策措施》，围绕加大对家政企业支持力度等5个方面提出19项具体举措。增加优质家政服务供给有哪些实招？如何更好发挥家政服务信用信息平台作用？记者进行了深入采访。 　　“小、散、乱”如何破解

警惕过度依赖AI陪伴

3年时间，年销售额从400多万美元增长到6000多万美元，海外市场从北美拓展到欧洲和日本——这是鑫帕特娜丝（青岛）家居饰品有限公司通过单家跨境电商平台出海交出的成绩单。

中新社湖南湘西8月16日电 题：湖南湘西“挤满”外国人

中新网长沙8月16日电(宋倩)北斗时空安全试验基地共建启动暨产业平台建设发布会16日在长沙举行，相关专家学者、企业代表等300余人齐聚，聚焦时空安全产业，共探共建共享产业生态。

2026年，“China Travel”从社交平台的热词变成海关入境章上的真实数据。上半年2291.4万人次外国人入境，同比增长20.4%，免签入境占比高达77.7%。

中新网上海8月16日电 (记者 陈静)15日正在举行的2026上海书展上，《微弱烛光：PA时代与人类沉浮》新书发布会举行。

近期，外资在中国高端制造领域投资动作频频，一批重大项目接连落地、开工或投产，主要集中于汽车核心零部件再制造、动力电池、化工新材料及半导体等高新技术产业领域。受访专家认为，外资加码中国高端制造的背后，是中国产业升级红利与全球供应链重构的双重驱动，跨国公司正以实际行动为“投资中国”投下信任票。

海关总署数据显示，7月份，全国货物贸易进出口总值4.66万亿元，其中，高技术产品出口增长超五成。面对全球数字化、绿色化发展趋势，中国人工智能、绿色低碳等产业出口动力强劲，出口含“新”量含“绿”量不断攀升。

中新网乌鲁木齐8月16日电 (记者 戚亚平)16日，“鲁兵携手·产业共兴”山东—新疆产业合作对接活动(新疆生产建设兵团第十二师)在乌鲁木齐市举行。活动邀请两地80余家企业参与，涉及农业机械及装备、食品、消费品、生物科技等领域；同时安排了产业合作对接推介、实地对接等环节，力争达成实质性合作。

中新网廊坊8月16日电 8月14日至15日，由中共廊坊市委网信办组织开展的“我是廊坊推荐官”之行在河北廊坊举行。活动聚焦新材料、空港装备、智慧康养、数字档案、航空实训、柔性显示、高端医疗、文化科教八大领域，系统记录并展示廊坊现代化产业体系建设的阶段性成果。

中新社福建武夷山8月16日电 题：“吃茶去”品出武夷山星村镇的茶旅共富图景

中新网乌兰察布8月16日电 (记者 张林虎)16日，“卓资熏鸡香·四季好风光”品牌推介会暨全国美食主播溯源直播带货活动在内蒙古自治区乌兰察布市卓资县举行。

中新网杭州8月16日电(郭其钰)8月16日，浙江杭州具身智能人才培养与考试认证中心及宇树科技机器人(具身智能)职业技能培训中心在杭州滨江成立。该中心由国家人工智能应用中试基地(具身智能)联动宇树科技共建，是目前全国唯一的具身智能人才培养与考试认证中心。

中新网嘉兴8月16日电(傅飞扬)8月15日至8月17日，第35届海宁中国家纺城家用纺织品博览会(简称“海宁家博会”)在浙江嘉兴海宁市举行。连日来，海宁中国家纺城周边人潮涌动，来自全国各地的采购商拖着行李箱、拿着采购清单涌入展馆。

中新网鹤岗8月16日电 题：鹤岗双赛联动：林海奔跑与炭火烟火解锁“流量变留量”的文旅密码

中新网三亚8月16日电 (张月和)“心中早就期盼着开渔了，我们已经准备了半个多月，网具、冰块、燃油、淡水全部补足到位。”8月16日正午，三亚崖州中心渔港内，渔民谭农权站在船头，一边检查缆绳一边说，最希望的是人员船舶平安顺利，满载而归。他的身后，成排渔船正等待离港指令。

眼下正值旅游旺季，湖南多地古城古镇夜间客流涌动，热闹非凡。而几年前，这些地方还曾白天游人匆匆、入夜便归于沉寂。近年来，湖南各地立足湖湘文化、历史和非遗资源，摒弃短期流量思维，推动旅游从过路式打卡转向沉浸式文化体验与过夜深度消费，为全省文旅高质量发展注入新动能。

中新网湛江8月16日电 (梁盛 谢剑锋)16日12时，南海伏季休渔期正式结束，广东湛江湾汽笛齐鸣，万千渔船列队驶向深海。当天，2026年湛江开渔季启动仪式暨“湛品”展示活动同步举行，产业合作成果丰硕，线下达成签约总额50.5亿元，覆盖保险保障、银行授信、产销对接、冷链物流等海洋渔业全产业链。

中新网云南昭通8月16日电 (时文枝)“家人们，昭通早熟糖心苹果熟了，又甜又脆又多汁！产地直发，快来选购！”16日，在云南省昭通市昭阳区洒渔镇的苹果种植基地，主播范强正对着手机镜头热情推介早熟苹果“瑞雪”。随着新一年苹果丰收季的到来，这位助农主播的直播间再度升温，订单纷至沓来。

中新网河池8月16日电(记者 黄艳梅)碧水幽幽，鱼影灵动，桨板泛波……今年暑期，广西西北小城天峨县凭借“蓝刀鱼风暴”火爆出圈，数万尾蓝刀鱼在清澈见底的红水河中翻涌如银鳞流动，吸引大批游客慕名前来体验这场“奔县游”的清凉之约。

中新网北京8月16日电 题：AI与算法成为“流量入口” 专家建言电商领域科学应对

中新网长春8月16日电 (李彦国 高龙安)中国食用菌产业在保障国内供给的同时，一方面向外输出种植技术助力非洲粮食安全，另一方面搭建国际平台开展菌物前沿科研合作。

当地时间8月14日下午，由中央广播电视总台译制的首部阿塞拜疆语版精品节目《习近平的文化情缘》启播与赏片活动在巴库举行。阿塞拜疆国家电视广播公司文化频道和阿塞拜疆ARB24电视台先后播出了该片。

今年是“两山”理念提出21周年。在过去的六十多年里，北京以北约460公里的塞罕坝，完成了从“黄沙遮天日，飞鸟无栖树”的荒原，到百万亩林海葱郁叠翠、松涛绵延千里的华丽蜕变。这个全球面积最大的人工林海背后，是三代中国人跨越半个多世纪的青春接力、汗水浸润与执着坚守。

央视网消息：目前，中国空间站神舟二十三号乘组三名航天员朱杨柱、张志远、黎家盈已在轨驻留80余天。上周，空间科学实(试)验、在轨训练及演练等工作有序推进。

依法严厉打击舆情敲诈等新型黑恶犯罪，是深化扫黑除恶专项斗争的重点任务。近期，公安机关网安部门循线攻坚，接连破获3起针对影视行业的网络舆情敲诈案件，累计抓获犯罪嫌疑人24名，有力打击了网络黑灰产嚣张气焰，切实维护了网络传播秩序与相关主体合法权益。

“面对新一代人工智能技术快速演进的新形势，要充分发挥新型举国体制优势，坚持自立自强，突出应用导向，推动我国人工智能朝着有益、安全、公平方向健康有序发展。”

“这是一个令人惊叹的国家。”英国游客乌尔维最近开启第三次中国之旅，去了西安、拉萨等地，计划明年重走丝绸之路。荷兰游客巴哈原计划来华旅行几周，结果待了半年；西班牙游客佩雷斯一家回国后意犹未尽，又开启了“二刷”。而美国特斯拉公司首席执行官马斯克日前在社交平台上点赞中国并“强烈推荐大家去看看”，更是点燃外国游客对中国的热情。他们纷纷感叹，“中国值得反复探索，一次根本不够”“我只想一次又一次地来到中国”……

中新网拉萨8月16日电 (记者 赵朗)记者16日从青稞产业科技成果观摩研讨会上获悉，西藏青藏高原作物种质资源中期库，保存青稞资源10000余份。

今天的《活力中国调研行》，一起走进安徽一个特殊的实验室。这是全国乃至全球首个以人工智能全链条驱动的科研创新实验室。

中新网兰州8月16日电 “绿水青山·美丽中国行”系列生态文化传播活动(武威站)14日在甘肃省武威市正式启动。

中新网北京8月16日电 (阮煜琳 李昂 匡雷 颜泽彬)南海开渔啦！8月16日12时，南海海域为期三个半月的伏季休渔期正式结束。随着新修订的《中华人民共和国渔业法》今年5月1日正式施行，南海迎来了该法实施后的首个开渔季。8月16日，粤桂琼三省(区)沿海数万艘渔船在海警舰艇的护航引导下，有序驶离渔港，前往南海开展捕捞作业。

山清水秀，物阜民丰。

习近平党建思想，是习近平新时代中国特色社会主义思想的重要组成部分，是加强新时代党的建设的根本遵循。要深刻理解和把握习近平党建思想的内涵要义，必须深刻领会“十四个坚持”。

今年8月15日是第四个全国生态日。

中新网广州8月16日电 (记者 方伟彬)8月16日中午12时，为期三个半月的南海伏季休渔期正式结束。随着开渔仪式在阳江闸坡渔港启动，广东沿海各地渔港千帆竞发、渔船争“鲜”出港，全面开启新一轮捕捞生产作业。

中新网北京8月16日电 (记者 应妮)8月15日是全国生态日，一场打通哲学理论、出版文本与大众现实生活壁垒的对谈当日在京举行。活动以商务印书馆近年推出的“博物学家传记译丛”为切入点，辨析公众理解的科学与博物学的本质差异，立足生态文明建设语境，探讨博物学复兴的现实意义，为大众带来兼具哲学深度和自然温度的分享。

中新社北京8月16日电 题：两岸亲情为何总能跨越山海？

中新社福建泉州8月16日电 (记者 闫旭)16日12时，福建省北纬26度30分以南海域正式结束三个半月的海洋伏季休渔。8000余艘渔船拔锚启航、开启新一轮耕海牧渔。

中新网成都8月16日电 (记者 何浠)当前，四川进入“七下八上”防汛关键期，汛期旱涝并重、急转特征明显，短时强降雨、极端局地天气频发，四川省防汛减灾形势严峻。近日，记者走进达州、广元等川东北防汛一线，实地调研区域防汛减灾、重大水利工程建设及智慧防汛落地成效，全景展现四川统筹安全与发展、全方位筑牢流域水安全屏障的务实举措。

中新网福州8月16日电 (记者 闫旭)8月15日是全国生态日，2026闽江河口湿地“全国生态日”暨湿地博物馆启新科普活动，在完成全面升级改造的闽江河口湿地博物馆举行。

中新网海口8月16日电 (黄启炼 高磊)为期三个半月的南海伏季休渔期16日正式结束，渔民们迎来期盼已久的开渔时刻。为保障开渔期间海上交通安全，海南海事局联合多部门同步启动“2026护航开渔行动”。

中新社北京8月16日电 (记者 马帅莎 郑莹莹)据中国航天科技集团消息，北京时间8月16日12时10分，中国在海南商业航天发射场使用长征十二号运载火箭，成功将卫星互联网低轨24组卫星发射升空，卫星顺利进入预定轨道，发射任务获得圆满成功。

他隐姓埋名28年，却因一声又一声巨响，震惊了世界。

中新网马鞍山8月16日电 题：一江青绿润钢城

据中国海事局网站消息，大连海事局发布航行警告，8月16日16时至30日16时，渤海海峡黄海北部部分海域执行军事任务，禁止驶入。

老有所养和幼有所育，关系千家万户。2022年7月13日，走进新疆乌鲁木齐市天山区固原巷社区，习近平总书记看望银发耄耋老人，祝福他们“幸福生活日久天长”。一个多月后，到辽宁沈阳市皇姑区三台子街道牡丹社区，习近平总书记再次重点询问老人孩子的情况，强调“对老年人的服务要跟上”“对孩子们的养育和培养等工作要加强”。

原标题：新华社国家高端智库发布《生态环境法典的中国创新与世界贡献》智库报告

新华社北京8月16日电 中国共产党的优秀党员，久经考验的忠诚的共产主义战士，杰出的无产阶级革命家、政治家，党和国家的卓越领导人，中国共产党第十四届、十五届中央政治局常委，国务院原总理朱镕基同志的遗体，将于18日在北京火化。

中新网北京8月16日电 (记者 马帅莎 骆云飞)北京时间8月16日12时10分，我国在海南商业航天发射场使用长征十二号运载火箭，成功将卫星互联网低轨24组卫星发射升空，卫星顺利进入预定轨道，发射任务获得圆满成功。(完）

新华社海南文昌8月16日电(记者陈凯姿)16日12时10分，我国在海南商业航天发射场使用长征十二号运载火箭，成功将卫星互联网低轨24组卫星发射升空，卫星顺利进入预定轨道，发射任务获得圆满成功。

中新社上海8月16日电 (陈静 蔡文亮)“海港未来”号LNG加注船、“海港致远”号甲醇加注船16日在绿华山锚地相继完成燃料加注作业。这是上海港首次在海上锚地同步为国际航行船舶大规模加注两类绿色燃料，标志着上海港海上能源加注作业具备多元化、规模化供应能力。

60天“大限”刚过，伊朗突然甩出“美军悬赏榜”！这信息量有点大 | 察世说

特朗普给美韩军演“泼冷水”：又贵又挑衅！与金正恩“关系好”，顺手吐槽韩国拒帮美国怼伊朗 | 国际早察

中新网8月17日电 据韩联社报道，当地时间17日凌晨4时42分许，韩国庆尚南道巨济市一处公寓附近发生泥石流，该公寓一层被埋，一人死亡。

中新网8月17日电 据卡塔尔半岛电视台报道，埃及、巴基斯坦、沙特、土耳其、卡塔尔、约旦、阿联酋、印度尼西亚八国外长16日发表联合声明，谴责以色列公开拒绝加沙和平计划路线图。

中新网8月17日电 据哥伦比亚媒体16日报道，哥伦比亚10日发生的强震已造成289人死亡、143人失踪、4187人受伤，受灾人口超过18万。此外，根据最新灾情报告，目前已有超2800所教育机构受损。

中新社里约热内卢8月16日电 (记者 林春茵)当地时间16日晨，巴西东南部里约热内卢州彼得罗波利斯市(Petrópolis)发生大巴翻覆事故，造成10人遇难。

中新网8月17日电 美国总统特朗普16日在其社交媒体平台“真实社交”上发文，对沙特阿拉伯、土耳其和巴基斯坦近期签署共同防务协议表示欢迎。

中新网8月17日电 美国总统特朗普16日在社交媒体发文称，鉴于他与朝鲜领导人金正恩的良好关系，他对于美国很早之前同意参加与韩国的联合军演“感到不满”。

据黎巴嫩公共卫生部紧急行动中心16日消息，3月2日至8月16日期间，以色列在黎巴嫩发动的持续不断的空袭和袭击已造成4346人死亡，12301人受伤。(总台记者 杨渊)

美国总统特朗普16日在社交媒体发文称，鉴于他与朝鲜领导人金正恩的良好关系，他对于美国很早之前同意参加与韩国的联合军演“感到不满”。

中新社柏林8月16日电 综合德国媒体16日报道，德国企业今年上半年对美国直接投资降至约43亿欧元，同比下降65%，为2023年以来同期最低水平。

中新社东京8月16日电 日本多家媒体16日发表社论，批评首相高市早苗在日本宣布无条件投降81周年之际回避历史反省，并对高市政权“回归战前”的政策动向表示警惕。

新华社北京8月16日电 综合新华社驻外记者报道：在日本宣布无条件投降81周年之际，日本首相高市早苗15日以“自民党总裁”身份，向供奉有二战甲级战犯的靖国神社供奉“玉串料”（祭祀费），并在出席日本政府举办的“全国战殁者追悼仪式”前从会场停车场朝靖国神社所在方向“遥拜”。此外，日本防卫大臣小泉进次郎等多名政客分别前往靖国神社参拜，这是日本连续第7年有现任阁僚在这一天参拜靖国神社。日方错误行径引发国际社会和日本国内爱好和平人士强烈反对。

日本多家媒体16日发表社论，批评高市早苗政权在日本宣布无条件投降81周年之际回避历史反省、淡化战争责任，并对日本政坛和安全政策日益显现的“回归战前”倾向表示警惕。

中新社雅加达8月16日电 (记者 李志全)印度尼西亚国家抗灾署16日通报，弗洛勒斯岛附近海域7.7级地震已造成53人死亡、135人受伤。目前，救援人员仍在继续搜寻幸存者。

当地时间16日，伊朗陆军总司令阿米尔·哈塔米表示，美军已被驱逐，不再获准进入波斯湾、阿曼湾及霍尔木兹海峡。

新华社北京8月16日电 特稿｜开放发展、可靠担当、魅力十足——国际人士眼里的中国

中新社悉尼8月16日电 (记者 薄雯雯)澳大利亚联邦政府和新南威尔士州政府16日共同宣布，新州将成为首个实施全国枪支回购计划(National Gun Buyback Scheme)的州，以减少民间枪支数量、提升社区安全。

中新网8月16日电 据日本广播协会(NHK)16日报道，日本千叶县13日开始的暴雨已致10人死亡，约1500栋房屋被淹。

美国总统特朗普15日在社交媒体再度发文，敦促国会参议院尽快表决通过全年采用夏令时的相关法案。

当地时间16日，韩国政府人士消息称，数名朝鲜军人日前越过东部地区军事分界线，韩军随后鸣枪警告。据称，这是今年以来韩军首次因朝军越界实施警告射击。

印尼国家抗灾署官员阿卜杜勒·穆哈里16日下午表示，截至目前，印尼7.7级地震已造成53人死亡。自地震发生以来，当地已记录到995次余震。

经济账怎么算？靠透支自然资源换来的GDP，可能连治污的零头都不够？联合国副秘书长福阿德近日接受总台《高端访谈》栏目专访时表示，“绿水青山”才是最稳的投资。

1945年8月15日，日本战败，宣布无条件投降。近日，多地多家纪念馆、档案馆公布了一批新发现和收集的日军侵华证据，涵盖侵华日军士兵私人信件、口述证言，战时新闻报道集，有关部队官方文书，幸存“慰安妇”实物、照片、视频史料等，来源广泛、相互印证，进一步充实了相关历史研究的证据链条。

新华社北京8月16日电（记者许苏培）8月15日，日本战败投降日。然而，在这本该是日本深刻反省、认罪赎罪的日子，日本执政当局却再次上演了令人愤慨的一幕——首相高市早苗向供奉有二战甲级战犯的靖国神社供奉了“玉串料”（祭祀费），还在场外向靖国神社方向进行遥拜；包括防卫大臣小泉进次郎在内的多名政要更是亲自前去“拜鬼”。

美国《巴伦周刊》近日刊文称，中国近年来一方面持续推动人工智能技术创新和规模化应用，另一方面不断完善相关治理体系。与此同时，中国人工智能产业也保持了较强的竞争力。

战争将满半年，油价飞涨，美伊互撕升级：特朗普呼吁美国人多花点汽油钱，伊朗副外长硬怼：海峡过去是伊朗的，现在也是，将来永远是！

中新网8月16日电 2026年8月15日是日本宣布无条件投降81周年的日子。当天，日本首相高市早苗不仅向供奉14名甲级战犯的靖国神社供奉祭祀费，在当天的纪念仪式致辞中也丝毫未提“反省”。

当地时间8月16日，印尼国家抗灾署发布消息称，东努沙登加拉省7.7级地震已致5000余人疏散避难。目前，应急救援工作的重点之一是保障安置点灾民的基本生活需求和救灾物资供应。

据印尼救援部门消息，截至当地时间16日8时30分，印尼7.7级地震死亡人数升至48人，另有64人受伤。

<img src="https://image.gcores.com/5af1005b7aa0e4ace02536a1ff31ae07-920-430.jpg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>IGN 于日前公布了《湮灭之潮》的最新实机演示。本作是一款围绕现实伦敦与异世界构建，并融入经典亚瑟王传奇的动作冒险游戏。在被异世界入侵后的破碎伦敦，灾难的幸存者格雯德琳坚定的走上了抗争灾难与探寻异世界入侵的旅途。 </p><div> <figure><p>&lt;内嵌内容，请前往机核查看&gt;</p></figure></div><p>开发者表示，骑士战斗系统是游戏的创新设定，玩家可以在游戏中收获十种以上的独特骑士，这些来自亚瑟王传说的著名骑士们将与你并肩作战，他们都有着自己独特的角色定位——无论是冲击战斗，防御攻击，还是释放毁灭性的连击等等。</p><div>于<figure><p>&lt;内嵌内容，请前往机核查看&gt;</p></figure></div><p>玩家可以根据战斗需求或者个人喜好进行对你的骑士团进行搭配组合，构建出丰富的战斗Build，不仅带来变化多端的动作连携，还能提供特殊的战斗策略。</p>

<img src="https://image.gcores.com/694e5cb61e6bf646f005e0e8884da73a-739-415.webp?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>由 Rebel Wolves 工作室开发、万代发行的动作角色扮演游戏《黎明行者之血》现已在 Steam 开启预购，国区售价298元起。游戏工作室由《巫师3》总监 Konrad Tomaszkiewicz 领衔，团队包括多名 CDPR 老将。</p><div> <figure><p>&lt;内嵌内容，请前往机核查看&gt;</p></figure></div><p>作为一款开放世界动作 RPG，本作融合了丰富的奇幻与历史元素。玩家将在一个被黑暗力量侵蚀的大地上探索，击败怪物、揭露阴谋、做出关键选择。这些选择会塑造主角的命运，以及所处世界的走向。</p><div> <figure><img alt="" src="https://image.gcores.com/26effc12a41927f00637e52651986c55-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><div> <figure><img alt="" src="https://image.gcores.com/bbb97b5b866afe7f8faa9f95560f3388-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><div> <figure><img alt="" src="https://image.gcores.com/f5d62dfb9cc246c01b3a312dd2f088b3-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>在游戏中，你将扮演科恩，在阴差阳错之下，这位年轻人被转化为一名Dawnwalker，他命中注定将在白昼与黑夜这两个世界的分界线上永恒徘徊。竭尽全力守护自己的人性，或者选择拥抱诅咒之力以拯救你的家人。无论如何选择，你都无法逃避最关键的问题：是选择你的灵魂，还是你所爱之人的生命？</p><div> <figure><p>&lt;内嵌内容，请前往机核查看&gt;</p></figure></div><p></p><p></p><p></p>

<img src="https://image.gcores.com/ce55820ac1068d636725666c2094bfc1-920-430.jpg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>由 CH4WORKS 打造的中世纪奇幻大战略游戏《大征服》正式公布。开发者表示，一款真正意义上的实时中世纪奇幻大战略游戏，战役世界与战斗将在同一个时间线上同步展开。玩家可以在治理王国与指挥战斗之间无缝切换，而这一切都会在实时世界中持续进行。</p><div> <figure><img alt="" src="https://image.gcores.com/f4bd8b35d28d75c7b640b6f782ecf34a-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><div> <figure><img alt="" src="https://image.gcores.com/5dfec5c899e1a54e181ca9ece475e100-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>与此同时，玩家可指挥前所未有规模的军队，在最多同时存在 <span style="font-weight: bold;">10万</span>单位 的史诗战场中展开战争。率领步兵、骑兵、弓箭手、攻城器械、传奇英雄以及毁灭性的魔法力量，在辽阔战场上争夺胜利。战场部署、补给运输以及战术选择，将决定最终的胜负。</p><div> <figure><img alt="" src="https://image.gcores.com/289b795c6be8f6c03fa8ad26df714d74-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><div> <figure><img alt="" src="https://image.gcores.com/6caff18ff7fbde5defba31adf38e537a-1920-1080.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>除了单人游玩外，玩家还可与其他玩家共同体验完整大战役，在同一个持续运行的实时世界中争夺霸权。每位统治者都将掌控一个独立王国，形成充满外交、贸易、联盟、竞争与战争的动态世界。每一个玩家的选择，都将影响整个大陆的历史走向。</p>

<img src="https://image.gcores.com/0f72e2dbe4756c099cb0ae2365f65bd4-920-430.jpg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>为庆祝游戏发售五周年，《末日准备狂》现已实装“烹饪更新”第一部分。 有了全新的 PrepperMix 机器，玩家可以自动化烹饪，在地堡中节省宝贵的时间。开发者表示，愈发复杂的厨房设备对生存必不可少，玩家可以在3级工作台找到配方。</p><div> <figure><img alt="" src="https://image.gcores.com/e1485db18f0f6532467d9c938d7330ee-800-450.png?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>本次更新加入了全新的冰沙、沙拉、果酱和能量棒，让玩家有更多方式利用物资并维持状态。团队还加入了即食面，这些物品会提供特殊且持续时间较长的增益，在最需要的时候获得额外优势。</p><div> <figure><img alt="" src="https://image.gcores.com/b9e98c3317b5a3c7e7fbbb63a362397d-960-311.png?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>最后，玩家现在还可以制作新的干制和烘烤食物（例如单只鸡腿）以及更多内容，因此也有更多方式利用找到的物资。更多配方等待发现，更多食物可以储备，还有更多方式让玩家保持准备状态。</p>

<img src="https://image.gcores.com/af1a117ace114523d870e1742fa98f36-1600-900.jpg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>本期时间轴制作：天行</p><p>前一段时间，《抵抗者》放出了实机演示与线下试玩，内容主题是东北抗联在雪原中与日本侵略军的鏖战。这恰好进入了最近正在研读抗联史的老白头的舒适区：在当年的东北平原上，有哪些武器是被抗联广泛使用的？它们的来源都是哪里？收集、缴获、修理、制造轻兵器，都有哪些故事值得挖掘？请听本期节目</p>

<img src="https://image.gcores.com/7e142e3475d93414cc2469873446ffaf-550-309.jpg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>随着 Team Resilience 于今日在《Dota2》TI15突围赛中以2比1的比分不敌 Team Spirit ，三支中国战队 XG 、VG 、TR 全部折戟，无一晋级八强。本次缺席正赛也创下了中国队自TI1以来的历史最差战绩。 </p><div> <figure><img alt="" src="https://image.gcores.com/6b0d1f3f485b581aaf5f10dc3b3242ff-800-450.jpg?x-oss-process=image/resize,limit_1,m_lfit,w_700,h_2000/quality,q_90/watermark,image_d2F0ZXJtYXJrLnBuZw,g_se,x_10,y_10" /></figure></div><p>本届国际邀请赛是继 TI9 以来第二次在中国举办，正赛阶段将于8月20日在上海重燃战火。 </p>

<img src="https://image.gcores.com/5d78d3405bcbf938d021284ec5c6036a-1400-1400.png?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>概述</p><p>这一期我们请来科技数码频道「热男 HotGuys」的两位联合创始人崔野和尹国威。</p><p>从《国行顶配 3》聊起，我们拆解了一条剧情向数码视频从创意、读稿、分镜、踩点、拍摄到后期审片的完整过程，也回到热男的创业起点：一群在爱搞机共事多年的伙伴，为什么只带着半年房租走出来创业，又如何让六位合伙人与一支十人团队稳定合作至今。这期也聊到内容团队很难绕开的问题：创意怎样拍板，商业合作为什么会选择热男，手机壳、贴纸、椰子鸡和辣椒酱为什么没能成为第二曲线；当小剧场不再稀缺、短视频改变内容消费、AI 开始生成视频，科技媒体还可以做多久？热男给出的答案很简单：只要还能覆盖成本，就一直做下去。媒介会变，观众会变，但让人开心仍然是一件有价值的事。</p><p>嘉宾</p><p>崔野：热男 HotGuys 联合创始人、制片／编剧。</p><p>尹国威：热男 HotGuys 联合创始人、导演。</p><p>潮汕仔：热男 HotGuys 联合创始人</p><p>主播</p><p>零号、初号：皮蛋漫游记主播。</p><p>时间轴</p><p>00:00丨认识热男 HotGuys   04:45丨《国行顶配 3》是怎样拍出来的   10:40丨从《老炮儿》到汽车哨兵模式   14:23丨热男如何一步步走到今天   19:20丨从同事、朋友到合伙人   22:40丨用半年房租启动一家公司   26:29丨全员创作者，谁来做最后决定   30:40丨为什么让更多成员走到镜头前   34:45丨把“电影梦”放进数码视频   37:07丨客户为什么为“好笑”买单   44:44丨为什么没有主动经营“资源链接器”   48:39丨想一直做到 80 岁   50:24丨AI 能替代“活人感”吗   51:31 丨文字、播客、长视频不是替代关系   53:57丨为什么没有把长视频直接切成短视频   55:14丨合伙人意见不一致时怎么办   59:18丨做播客的人如何持续输入 62:03丨长期输出，以及让人开心的价值   64:36丨三十年后留下来的，才是好东西   66:55丨最近在听什么播客  </p><p>本期提及</p><p><a href="https://space.bilibili.com/17138783">热男 HotGuys｜哔哩哔哩</a></p><p><a href="https://search.bilibili.com/all?keyword=iPhone%2013%20Pro%20Max%20%E5%BB%BA%E8%AE%AE%E6%94%B9%E6%88%90%20%E5%9B%BD%20%E8%A1%8C%20%E9%A1%B6%20%E9%85%8D%20%E7%83%AD%E7%94%B7">《iPhone 13 Pro Max，建议改成：国 行 顶 配》｜热男 HotGuys</a></p><p><a href="https://search.bilibili.com/all?keyword=iPhone%2015%20Pro%20Max%20%E5%BB%BA%E8%AE%AE%E6%94%B9%E6%88%90%20%E5%9B%BD%20%E8%A1%8C%20%E9%A1%B6%20%E9%85%8D%202%20%E7%83%AD%E7%94%B7">《iPhone 15 Pro Max，建议改成：国 行 顶 配 2》｜热男 HotGuys</a></p><p><a href="https://www.bilibili.com/video/BV15tZWBvEER/">《小米 YU7 Max：国 行 顶 配 3》｜热男 HotGuys</a></p><p><a href="https://www.bilibili.com/video/BV1MUPWzkE8u/">《热男燃尽了！〈国行顶配 3〉幕后花絮》｜热男 HotGuys</a></p><p><a href="https://search.bilibili.com/all?keyword=%E4%BB%8E%E7%8E%B0%E5%9C%A8%E5%BC%80%E5%A7%8B%20%E8%AF%B7%E5%8F%AB%E6%88%91%E4%BB%AC%20Hot%20Guys">《官宣：从现在开始，请叫我们 Hot Guys！》｜热男 HotGuys</a></p><p><a href="https://www.xiaoyuzhoufm.com/episode/698d8173a22480add6daf637">《2025 年终总结：献给每一位两颗皮蛋的同路人，也献给我们自己》｜皮蛋漫游记 Vol.86</a></p><p><a href="https://search.bilibili.com/all?keyword=%E4%BB%80%E4%B9%88%E6%98%AF%E9%A6%99%E6%A0%BC%E9%87%8C%E6%8B%89%20vivo%20X300%20Ultra%20%E4%B8%A4%E9%A2%97%E7%9A%AE%E8%9B%8B">《什么是香格里拉？Feat. vivo X300 Ultra》｜两颗皮蛋</a></p><p><a href="https://search.bilibili.com/all?keyword=3%E5%B9%B44%E8%B6%9F%E5%B7%9D%E8%97%8F%E7%BA%BF%20%E5%8D%81%E4%B8%80%E8%87%AA%E9%A9%BE318%20%E4%B8%A4%E9%A2%97%E7%9A%AE%E8%9B%8B">《3 年 4 趟川藏线！十一自驾 318 保姆级攻略》｜两颗皮蛋</a></p><p><a href="https://podcasts.apple.com/us/podcast/id1711052890">《独树不成林》</a></p><p><a href="https://podcasts.apple.com/cn/podcast/id1711052890?i=1000764587533">《336-冷战是西方阵营对战东方阵营吗？》｜独树不成林</a></p><p><a href="https://www.xiaoyuzhoufm.com/podcast/6388760f22567e8ea6ad070f">《面基》</a></p><p><a href="https://www.xiaoyuzhoufm.com/episode/6850d2ed4abe6e29cb814160">《E112. 这期节目献给每一位喜欢投资和求真的听友》｜面基</a></p><p><a href="https://www.xiaoyuzhoufm.com/podcast-topic/65438626164cf7cfbba7d1bd">《基本无害 Mostly Harmless：城市生存手册》</a></p><p><a href="https://open.spotify.com/show/2Jw6ERvKZcdBzqYAoXiojH">机核兴趣频道</a></p><p><a href="https://www.xiaoyuzhoufm.com/podcast-topic/6549f693287a0d9f13519fb7">《重轻与他的兴趣世界》｜机核兴趣频道专题</a></p><p><a href="https://www.xiaoyuzhoufm.com/episode/610fcb84cf8e958986b4a7fd">《S2E2 平凡故事》｜不在场</a></p><p><a href="https://www.youtube.com/@FHZDN">烽火照东南｜YouTube</a></p><p><a href="https://www.youtube.com/@zuowei">左為｜YouTube</a></p><p><a href="https://podcasts.apple.com/tw/podcast/%E8%B5%9B%E5%8D%9A%E5%AF%B9%E8%AF%9D/id1806276336">《赛博对话》｜Apple Podcasts</a></p><p><a href="https://www.bilibili.com/video/BV1KprHBGEee/">《对话王自如与电丸科技 AK：CES 上有多少 AI 产品是噱头？》｜赛博对话</a></p><p><a href="https://www.bilibili.com/video/BV19qeRzLEoc/">《李宗盛专访》｜马世芳《音乐五四三》</a></p>

<img src="https://image.gcores.com/d8bada8b-decc-4f1f-849c-896377b2a659-1280-1280.jpeg?x-oss-process=image/resize,limit_1,m_fill,w_626,h_292/quality,q_90" /><p>本期SP，喵臣串台青年度日指南，</p><p>和主播拉帝快速短评暑期档前半段看过的片子。</p><p>等暑期档结束之后，摇摆指南也会推出更加完整的盘点，敬请期待！</p><p>本期时间轴：</p><p>00:03:00 《奥德赛》 00:09:15 《蜘蛛侠 崭新之日》 00:22:24 《痴迷》 00:31:44 《三国第一部 争洛阳》 00:34:20 《群星闪耀时》 00:36:00 《恐怖游轮》 00:37:53 《玩具总动员5》 00:44:36 《超级少女》 00:51:50 《后室》 00:54:51 《火遮眼》 01:01:45 暑期档还有什么期待的</p><p>更多精彩，敬请期待。感谢大家的收听与支持！</p>

<a href="https://www.nytimes.com/2026/08/13/arts/ai-podcasts-fashion-pop-avatars.html?unlocked_article_code=1.5FA.fKfk.bMp5uecIkIOV&amp;smid=bs-share"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i12.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p12#a260816p12" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Reggie Ugwu / <a href="https://www.nytimes.com/">New York Times</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.nytimes.com/2026/08/13/arts/ai-podcasts-fashion-pop-avatars.html?unlocked_article_code=1.5FA.fKfk.bMp5uecIkIOV&amp;smid=bs-share">A look at companies like Inception Point creating AI personas for media, fashion, film, and music industries that could host podcasts, model clothes, and more</a></b></span>&nbsp; &mdash;&nbsp; The A.I. personas created by Inception Point AI include, clockwise from top left, Claire Delish, VV Steele, Nigel Thistledown and Lila Walker.Inception Point AI</p>

<a href="https://www.wsj.com/tech/ai/ai-math-riemann-hypothesis-anthropic-openai-22f98a87?st=7mfzgc&amp;reflink=desktopwebshare_permalink"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i11.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p11#a260816p11" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Ben Cohen / <a href="https://www.wsj.com/">Wall Street Journal</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.wsj.com/tech/ai/ai-math-riemann-hypothesis-anthropic-openai-22f98a87?st=7mfzgc&amp;reflink=desktopwebshare_permalink">How a Claude model made a math breakthrough during its unsuccessful 54-hour attempt to solve the Riemann hypothesis after a user repeatedly encouraged it</a></b></span>&nbsp; &mdash;&nbsp; The world's smartest AI models are now superhuman at math.&nbsp; They still respond to moral support and encouragement from mere humans.</p>

<p><a href="https://www.techmeme.com/260816/p10#a260816p10" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Phil Serafino / <a href="https://www.bloomberg.com/">Bloomberg</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.bloomberg.com/news/articles/2026-08-16/lecornu-plans-crisis-meeting-on-cyberattack-at-french-tax-agency">French PM Lecornu will hold a crisis meeting Monday over a June-July cyberattack on France's tax agency that compromised 678K individual and business accounts</a></b></span>&nbsp; &mdash;&nbsp; French Prime Minister Sebastien Lecornu will hold a crisis meeting Monday to respond to the cyberattack disclosed last week targeting the country's tax collection agency.</p>

<a href="https://www.bloomberg.com/news/articles/2026-08-16/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i9.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p9#a260816p9" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> <a href="https://www.bloomberg.com/">Bloomberg</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.bloomberg.com/news/articles/2026-08-16/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion">Sources: Stripe has finalized a deal to acquire AI model marketplace OpenRouter for more than $7B; OpenRouter was valued at $1.3B in May</a></b></span>&nbsp; &mdash;&nbsp; Stripe Inc. has finalized an agreement to acquire OpenRouter Inc., a startup that helps companies switch between artificial intelligence models &hellip; </p>

<a href="https://www.wsj.com/health/ai-is-helping-patients-solve-medical-mysteries-3c2d7c25?st=C8CWgw&amp;reflink=desktopwebshare_permalink"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i8.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p8#a260816p8" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Alex Janin / <a href="https://www.wsj.com/">Wall Street Journal</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.wsj.com/health/ai-is-helping-patients-solve-medical-mysteries-3c2d7c25?st=C8CWgw&amp;reflink=desktopwebshare_permalink">Patients, families, doctors, and nurses are increasingly turning to AI tools, such as Face2Gene, to help identify rare and hard-to-diagnose diseases</a></b></span>&nbsp; &mdash;&nbsp; Patients, doctors and nurses are turning to the technology for help identifying rare and undiagnosed diseases</p>

<a href="https://www.washingtonpost.com/technology/2026/08/14/ai-becomes-major-election-issue-first-time-data-shows/?pwapi_token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJyZWFzb24iOiJnaWZ0IiwibmJmIjoxNzg2NjgwMDAwLCJpc3MiOiJzdWJzY3JpcHRpb25zIiwiZXhwIjoxNzg4MDYyMzk5LCJpYXQiOjE3ODY2ODAwMDAsImp0aSI6Ijk2OTNkOTM2LTg0ODgtNGVjOS1iNTdmLTVhY2ZjNTg5NmFlYyIsInVybCI6Imh0dHBzOi8vd3d3Lndhc2hpbmd0b25wb3N0LmNvbS90ZWNobm9sb2d5LzIwMjYvMDgvMTQvYWktYmVjb21lcy1tYWpvci1lbGVjdGlvbi1pc3N1ZS1maXJzdC10aW1lLWRhdGEtc2hvd3MvIn0.HwCWU70YJs-Ehcd0EjuIo042EWgf_bOlBZOXr7FILIg&amp;itid=gfta"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i7.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p7#a260816p7" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> <a href="https://www.washingtonpost.com/">Washington Post</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.washingtonpost.com/technology/2026/08/14/ai-becomes-major-election-issue-first-time-data-shows/?pwapi_token=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJyZWFzb24iOiJnaWZ0IiwibmJmIjoxNzg2NjgwMDAwLCJpc3MiOiJzdWJzY3JpcHRpb25zIiwiZXhwIjoxNzg4MDYyMzk5LCJpYXQiOjE3ODY2ODAwMDAsImp0aSI6Ijk2OTNkOTM2LTg0ODgtNGVjOS1iNTdmLTVhY2ZjNTg5NmFlYyIsInVybCI6Imh0dHBzOi8vd3d3Lndhc2hpbmd0b25wb3N0LmNvbS90ZWNobm9sb2d5LzIwMjYvMDgvMTQvYWktYmVjb21lcy1tYWpvci1lbGVjdGlvbi1pc3N1ZS1maXJzdC10aW1lLWRhdGEtc2hvd3MvIn0.HwCWU70YJs-Ehcd0EjuIo042EWgf_bOlBZOXr7FILIg&amp;itid=gfta">How concerns about AI became an important midterms issue, with candidates adding AI and data center policies to their websites in ~40% of races across the US</a></b></span>&nbsp; &mdash;&nbsp; Across the United States, political candidates are grappling with an issue that has never before featured widely in American political life: artificial intelligence.</p>

<a href="https://www.wsj.com/business/media/sony-ceo-hiroki-totoki-efc8923f?st=dvoitN&amp;reflink=desktopwebshare_permalink"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i6.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p6#a260816p6" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Jason Douglas / <a href="https://www.wsj.com/">Wall Street Journal</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.wsj.com/business/media/sony-ceo-hiroki-totoki-efc8923f?st=dvoitN&amp;reflink=desktopwebshare_permalink">A profile of Sony CEO Hiroki Totoki, who aims to transform the company into a business focused on music, movies, video games, and the tech that underpins them</a></b></span>&nbsp; &mdash;&nbsp; The company built its empire on electronics.&nbsp; CEO Hiroki Totoki says its future is entertainment</p>

<a href="https://www.ft.com/content/f063b38e-61c0-449b-aca3-968b903091a1"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i5.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p5#a260816p5" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Owen Walker / <a href="https://www.ft.com/">Financial Times</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.ft.com/content/f063b38e-61c0-449b-aca3-968b903091a1">Malaysia's 6% Q2 GDP growth was powered by 7.5% manufacturing growth, driven by chipmaking, and 6.6% construction growth, supported by data center development</a></b></span>&nbsp; &mdash;&nbsp; Southeast Asian country emerges as key AI hub in boost for its economy&nbsp; &mdash;&nbsp; The AI boom has made global superstars out of a handful &hellip; </p>

<p><a href="https://www.techmeme.com/260816/p4#a260816p4" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Timothy Gowers / <a href="https://gowers.wordpress.com/">Gowers's Weblog</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://gowers.wordpress.com/2026/08/12/what-sort-of-maths-are-llms-good-at/">Fields Medalist Timothy Gowers says most famous mathematics problems solved by LLMs so far have almost all been with counterexamples rather than proofs</a></b></span>&nbsp; &mdash;&nbsp; For the sake of anyone who might read this blog post in the distant future (a month from now, say), let me mention that I am writing &hellip; </p>

<a href="https://huggingface.co/blog/state-of-open-models-summer-2026"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i3.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p3#a260816p3" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> <a href="https://huggingface.co/">Hugging Face</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://huggingface.co/blog/state-of-open-models-summer-2026">Hugging Face says developers made 151K+ derivatives based on Qwen models, topping others, making Qwen one of the largest foundations in the open model ecosystem</a></b></span>&nbsp; &mdash;&nbsp; In the AI world, time feels compressed.&nbsp; A few months after our spring report in our biannual analysis worked through the ecosystem &hellip; </p>

<a href="https://www.ft.com/content/a026e094-7017-4ff8-8344-213573e973ed"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i2.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p2#a260816p2" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Ilya Gridneff / <a href="https://www.ft.com/">Financial Times</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.ft.com/content/a026e094-7017-4ff8-8344-213573e973ed">BlackBerry posted a positive cash position in Q1, its first cash-positive first quarter in any fiscal year since 2017, driven by its QNX and Secusmart products</a></b></span>&nbsp; &mdash;&nbsp; A decade after the disappearance of its handsets, the company's future rests on car and security software</p>

<a href="https://www.unite.ai/pathway-raises-new-funding-at-500m-valuation-to-scale-post-transformer-ai/"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260816/i1.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260816/p1#a260816p1" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Antoine Tardif / <a href="https://www.unite.ai/">Unite.AI</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.unite.ai/pathway-raises-new-funding-at-500m-valuation-to-scale-post-transformer-ai/">Pathway, which is developing AI models based on what it calls its &ldquo;Post-Transformer&rdquo; BDH architecture, raised a $30M seed at a $500M valuation</a></b></span>&nbsp; &mdash;&nbsp; AI research company Pathway has secured additional funding at a $500 million valuation, bringing its total seed financing to $30 million &hellip; </p>

<a href="https://www.bloomberg.com/news/articles/2026-08-14/why-ai-optimism-is-so-much-higher-in-china-than-the-us?accessToken=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzb3VyY2UiOiJTdWJzY3JpYmVyR2lmdGVkQXJ0aWNsZSIsImlhdCI6MTc4Njc1NjQxMSwiZXhwIjoxNzg3MzYxMjExLCJhcnRpY2xlSWQiOiJUSlFHMEdWVFREMFQwMCIsImJjb25uZWN0SWQiOiI4OUM4OTNDMDhGOTQ0NThDQkQwQTQyREY1RDFCOTY0QyJ9.fDW2gfet4fRl-LpHfBPyJnsGOX6C3WYMEcsoIbGcRUs"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260815/i18.jpg" vspace="4" /></a> <p><a href="https://www.techmeme.com/260815/p18#a260815p18" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Grace Shao / <a href="https://www.bloomberg.com/">Bloomberg</a>:<br /> <span style="font-size: 1.3em;"><b><a href="https://www.bloomberg.com/news/articles/2026-08-14/why-ai-optimism-is-so-much-higher-in-china-than-the-us?accessToken=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzb3VyY2UiOiJTdWJzY3JpYmVyR2lmdGVkQXJ0aWNsZSIsImlhdCI6MTc4Njc1NjQxMSwiZXhwIjoxNzg3MzYxMjExLCJhcnRpY2xlSWQiOiJUSlFHMEdWVFREMFQwMCIsImJjb25uZWN0SWQiOiI4OUM4OTNDMDhGOTQ0NThDQkQwQTQyREY1RDFCOTY0QyJ9.fDW2gfet4fRl-LpHfBPyJnsGOX6C3WYMEcsoIbGcRUs">Chinese citizens are more optimistic about AI than Americans because AI is seen as a practical tool in China that disrupts a smaller share of its population</a></b></span>&nbsp; &mdash;&nbsp; Different experiences with technology have shaped radically different expectations of what artificial intelligence will bring.</p>

<p><a href="https://lobste.rs/s/g1rpmg/webmaster_manifesto_for_everyone">Comments</a></p>

<p><a href="https://lobste.rs/s/xculjp/limits_ai_hubert_dreyfus_1985">Comments</a></p>

<p><a href="https://lobste.rs/s/joyawz/ai_software_development_what_does_data">Comments</a></p>

<p><a href="https://lobste.rs/s/vnpw9j/pony_s_arena_allocator">Comments</a></p>

<p><a href="https://lobste.rs/s/ln3dao/faster_way_calculate_day_week">Comments</a></p>

<p><a href="https://lobste.rs/s/xhliz8/human_interface_guidelines_from_various">Comments</a></p>

<p><a href="https://lobste.rs/s/8izobt/you_can_just_choose_how_many_bugs_you_want">Comments</a></p>

<p><a href="https://lobste.rs/s/zjg7ux/anti_social_media">Comments</a></p>

<p><a href="https://lobste.rs/s/pb8wvr/nokia_dct3_emulator_open_source_emulator">Comments</a></p>

<p><a href="https://lobste.rs/s/ii5tun/every_fucking_website">Comments</a></p>

<p><a href="https://lobste.rs/s/wqkycj/people_acm_russ_cox">Comments</a></p>

<p><a href="https://lobste.rs/s/wpyhdh/i_thought_i_was_building_c_replacement_i">Comments</a></p>

<p><a href="https://lobste.rs/s/hnx6id/protecting_rust_standard_library_from">Comments</a></p>

<p><a href="https://lobste.rs/s/dxfl0r/monitoring_systemd_services_with">Comments</a></p>

<p><a href="https://lobste.rs/s/bhhcf3/tour_magit_s_status_interface">Comments</a></p>

<p><a href="https://lobste.rs/s/5rh6fs/haskell_origins_evolution_future">Comments</a></p>

<p><a href="https://lobste.rs/s/iuqkgi/software_engineering_fundamentals">Comments</a></p>

<p><a href="https://lobste.rs/s/rexoqj/what_s_missing_have_reproducible_builds">Comments</a></p>

Linux Kernel 7.2 has just been officially released with a slew of new features, reports the blog 9to5Linux. Highlights of Linux 7.2 "include cache-aware load-balancing support, initial HDMI 2.1 FRL support to the AMDGPU driver, support for devres-based management of ACPI notify handlers, initial CRI platform support for the Intel Xe driver, and Rust support for the IBM System/390 (S/390) architecture." Linux kernel 7.2 also introduces a "Fair(er)" GPU scheduler, support for the 'zerocopy' library to Rust support to make zero-cost memory manipulation effortless, new hwcaps for the 2025 dpISA extensions on the AArch64 (ARM64) architecture, and enables large folios by default for the Btrfs file system. It also brings Intel CPU model number support for Panther Lake R processor series, improvements to the kernel's swap subsystem, support for multi-size transparent huge pages (mTHPs) to the khugepaged kernel thread, and support for compressed files to the SMB filesystem. On top of that, Linux 7.2 improves the new NTFS filesystem introduced in Linux kernel 7.1, adds devicetree updates for 64-bit NXP/Freescale and Qualcomm platforms, introduces MPTCP signaling support for IPv6 addresses, adds GRO/GSO support for PPPoE, and brings more SMP load-balancing updates... [T]he TCP authentication option has been implemented, and there are also some Thunderbolt networking improvements. Also worth mentioning is that the KVM subsystem has received support for AMD's "guest-mode execution trap" and Intel's "mode-based execution control" (MBEC) features, the NFS file system's default block size was bumped to 4MB on systems with at least 16GB RAM, and support for the Intel Trusted Domain Extensions (TDX) feature has been added. Thanks to Slashdot reader prisoninmate for bring the news. <p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Linux+Kernel+7.2+Has+Been+Officially+Released+with+Many+New+Features%3A+https%3A%2F%2Flinux.slashdot.org%2Fstory%2F26%2F08%2F16%2F2349224%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Flinux.slashdot.org%2Fstory%2F26%2F08%2F16%2F2349224%2Flinux-kernel-72-has-been-officially-released-with-many-new-features%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://linux.slashdot.org/story/26/08/16/2349224/linux-kernel-72-has-been-officially-released-with-many-new-features?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

"We are in a 21st century space race," U.S. Senator Ted Cruz has said. And this week CNN noted a competition that for decades loomed in the background of geopolitics "is now roaring to the forefront," with China "demonstrating rapid advances in space technology" while the U.S. is "ratcheting up rhetoric about a looming battle for control of the cosmos." It's the fact that China plans to build a permanent settlement on the moon by 2040 that has prompted an urgent response from [U.S.] lawmakers... They've declared a new space race, and NASA is pledging to build a lunar outpost of America's own.... Both the United States and China plan to send robotic explorers to the south pole later this year, paving the way for human explorers to follow close behind. The US is relying on the private sector to develop and build low-cost, exploratory robotic landers... Meanwhile, China is aiming to launch its robotic Chang'e-7 mission as soon as this month. The complex endeavor will seek to use four different robotic vehicles &mdash; an orbiter, lander, rover and mobile "hopper" &mdash; working in tandem to attempt to gather unprecedented data. The effort will include tools to drill for and directly analyze water ice, a feat that the US will only attempt with robotic missions slated for next year at the earliest. "If lunar water ice is successfully located, it could significantly reduce the cost and time required to transport water from Earth, facilitating the establishment of a human base for long-term activities on the moon and enabling further exploration of Mars or deep space," Tang Yuhua, the deputy chief designer of the Chang'e-7 mission, said in a 2025 interview with state media. The oncoming flurry of south-pole-centric activity, with few established international laws to govern the outcomes, is heightening the drama of this rivalry... But with a much more ambitious goal &mdash; a future in which people attempt to flourish within permanent settlements rather than just visit the moon &mdash; the stakes of this new space race are even higher, according to Clayton Swope, the deputy director of the Aerospace Security Project at the Center for Strategic and International Studies, a bipartisan nonprofit. "The long-term vision will take generations to execute on," Swope said. "But in my mind, it's a scenario where there are people being born, people dying, their families, children growing up in a place that is not Earth &mdash; so that looks a lot like a city, more than it looks like an expedition to the south pole." If such a future comes to fruition, Swope added, his hope is that such extraterrestrial colonies would be governed by rules reminiscent of Western ideals and democratic principles. Whether China would be on the same page is an open question. China is already fine-tuning its plans for a lunar settlement, CNN writes: The country envisions the International Lunar Research Station, or ILRS, as a sprawling science-focused facility on the moon's surface that can be operated robotically, "with the prospect of subsequent human presence," according to a government's ILRS website. China could partner with Russia to build a massive nuclear reactor to power the base, though questions remain regarding how the war in Ukraine may impact Russia's ability to deliver... China and Russia were mapping these plans long before the US announced its own intention to build a nuclear reactor on the moon last year and detailed its plans for a lunar settlement spanning hundreds of square miles... Who will write the rules? And how, exactly, will the use of lunar resources be policed on the international stage? Meanwhile, NASA "still does not have a vehicle capable of getting astronauts down to the lunar surface," the article points out: Returning to the moon is not as simple as repeating Apollo. The space agency cannot &mdash; and does not wish to &mdash; replicate its 20th century lunar-landing success. The supply chains, skilled machinists and factories that built Apollo's rockets, landers and command modules no longer exist. And while the Apollo landings stuck close to the lunar equator, landing near the south pole is substantially more complex and requires more powerful vehicles.... China, meanwhile, has been laser focused on its lunar exploration program since 2004, using its top-down, central planning system to keep resources flowing toward specific goals.<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=China's+Moon-landing+Plans+And+Why+the+US+Is+So+Worried+%3A+https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F231206%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F231206%2Fchinas-moon-landing-plans-and-why-the-us-is-so-worried%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://science.slashdot.org/story/26/08/16/231206/chinas-moon-landing-plans-and-why-the-us-is-so-worried?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

While Flock announced changes for its AI-powered traffic cameras, "Several of the proposed changes Flock is touting are merely retreads of previous ," complains the American Civil Liberties Union. "Flock's latest announcement still appears more focused on addressing a perceived PR problem than the significant harms its products create... [T]his is hardly the step forward Flock wants us to think it is " The ACLU continues to urge that default retention periods be shortened to 48 hours &mdash; not one week. And they warn Flock allows longer retention to any police department that asks for it, or when police officers activate "Evidence Mode" (the scope of which is not yet clear): Even if "Evidence Mode's" data retention hold only applies to hits returned on a given search, it would still retain significant amounts of location data on persons and vehicles who law enforcement do not suspect have engaged in any wrongdoing... Flock claims that its changes will provide "more local control," meaning local police can decide what types of offenses other Flock customers can search their data for... This is not new. Flock has attempted this before, and the security measure failed because users were easily able to circumvent the system's requirement that police input the purpose of their search. For example, on June 12, 2025 Flock started claiming its new "Proactive Search Term Tool" would block any "impermissible" searches, such as abortion-related searches in states like Illinois that prohibit sharing reproductive healthcare data. But police officers quickly realized they could just input "investigation" or even "hehehe" as a search reason and it would be approved. Flock later switched from an open text box to a drop-down menu of reasons, but that just offered police a list of acceptable purposes they could choose from, whether it was accurate or not. Until Flock demonstrates they can develop a reliable, workable system to prevent improper searches, this promise of local control provides nothing more than a false sense of security... While providing "Audit Assistance" to all departments makes sense, there is no evidence that the tool works consistently to address what the Washington Post observed is a growing pattern of police officers turning Flock into a personal stalking tool. While dozens of officers have recently been arrested, fired, or otherwise disciplined for misusing Flock for personal reasons, Flock claims these arrests are proof its auditing tool works. However, unless we know the number of officers misusing the system, we cannot conclude if Flock and its auditing tools are catching 95 percent of violators or 5 percent. Flock needs to have its auditing tool analyzed by an independent evaluator to determine its actual effectiveness. Until then, we don't know if the tool is a real security measure or just window dressing. Perhaps the oddest part of Flock's announcement is its claims that "now every search will require" police to input a case code... While claiming that a "search without a reason is a search that shouldn't happen in the first place," Flock's announcement fails to note how easily users have circumvented "search reason" security measures in the past... This leaves the public wondering how making an ineffective voluntary security measure mandatory will improve its functionality. Flock's "nearly $1 billion in venture capitalist funding has locked it into an operational model that seeks to trade our privacy for massive profits," concludes the ACLU's statement, as they promise to continue "The ACLU is fighting alongside communities to cancel local ALPR contracts and push lawmakers to protect our rights from this surveillance nightmare..."<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Flock's+'Creepy+Cameras'+Remain+Major+Threat+to+Privacy+Despite+Small+Recent+Changes%2C+Warns+ACLU%3A+https%3A%2F%2Fnews.slashdot.org%2Fstory%2F26%2F08%2F16%2F2035237%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fnews.slashdot.org%2Fstory%2F26%2F08%2F16%2F2035237%2Fflocks-creepy-cameras-remain-major-threat-to-privacy-despite-small-recent-changes-warns-aclu%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://news.slashdot.org/story/26/08/16/2035237/flocks-creepy-cameras-remain-major-threat-to-privacy-despite-small-recent-changes-warns-aclu?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

"As Sony soldiers ahead with their plans to end hard copies for new PlayStation games, players are fearing a future which will force them to pay more and own less," writes Kotaku. What will prices do in a world with no physical copies to share or sell? One former Square Enix executive speculates this could actually drive down the price of games: "Digital on the PS store will likely get cheaper," posted Jacob Navok, former Square Enix business strategist. "Current prices are tied to retail. Without physical discs holding digital hostage we'll see a larger spectrum of pricing similar to steam." Navok says that the digital-only landscape will pit major publishers in constant, dynamic competition with each other... "The more the Playstation store becomes digital only, the greater this trend will accelerate," Navok follows up. "You will see more sales and dynamic pricing similar to Steam because publishers will compete among themselves to a greater extent." [O]n the other end newer games are getting increasingly expensive, even on digital platforms. The fact that Grand Theft Auto VI will be a purely digital release has not caused its price point to budge. <p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Could+a+Disc-Free+PlayStation+Actually+Lower+Game+Prices%3F%3A+https%3A%2F%2Fgames.slashdot.org%2Fstory%2F26%2F08%2F16%2F1839224%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fgames.slashdot.org%2Fstory%2F26%2F08%2F16%2F1839224%2Fcould-a-disc-free-playstation-actually-lower-game-prices%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://games.slashdot.org/story/26/08/16/1839224/could-a-disc-free-playstation-actually-lower-game-prices?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

An anonymous reader shared this report from TechCrunch: According to data from digital intelligence provider Similarweb, Bluesky's mobile app had 10.4 million monthly active users worldwide in June 2026, down 27.2% year-over-year. In addition, mobile daily active users continued to decline, falling 25.6% year-over-year in July to around 3 million... Bluesky's app has lost more than half its monthly active users from its late-2024 high... Its quarterly average was around 22.1 million monthly active users in the fourth quarter of 2024, Similarweb's data indicates, and it declined to 10.7 million in the second quarter of 2026. That's a decline of around 52%. While Bluesky's numbers are down, those who stuck around are committed. Its smaller community remains relatively active, with a stickiness rate (the ratio of daily to monthly active users) of roughly 29% in June, about the same as Threads. For Bluesky's new CEO, Toni Schneider, these numbers may not be as concerning. The company is not entirely focused on making Bluesky (the app) succeed, but on making it possible for the underlying protocol (AT Proto) to power a growing number of social apps, services, and communities. That's something that is now taking place, as projects like BlackSky and Eurosky are growing, while some AT Proto apps like the video-focused Skylight have found early traction, too. In addition, the company has launched new products, like the AI-powered research tool Attie, and is now working on adding support for private data to Bluesky. The latter could generate new interest in Bluesky from a different type of user &mdash; those less interested in the public square, and more interested in private networking and communities. But on Threads in July of 2026, daily active users were up 21.3% year-over-year to 147 million, the article points out, and website visits were up 112% year-over-year to 471.6 million. And Threads now appears to have roughly 66% more monthly active users than X.<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Bluesky's+Active+User+Base+Shrinks+52%25+Over+18+Months%2C+But+Its+Protocol+is+Spreading%3A+https%3A%2F%2Ftech.slashdot.org%2Fstory%2F26%2F08%2F16%2F0326249%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Ftech.slashdot.org%2Fstory%2F26%2F08%2F16%2F0326249%2Fblueskys-active-user-base-shrinks-52-over-18-months-but-its-protocol-is-spreading%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://tech.slashdot.org/story/26/08/16/0326249/blueskys-active-user-base-shrinks-52-over-18-months-but-its-protocol-is-spreading?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

Saturday Disneyland announced a special Star Wars-themed collaboration with third-person shooter game Fortnite &mdash; including a slick new trailer for its Star Wars: Smuggler's Gambit quest. Visitors to Disneyland's Smuggler's Run attraction at Disney's theme parks in Orlando and in Anaheim can accept a special mission while waiting in line for the ride, according to The Hollywood Reporter. And "If they succeed, they will get a unique Forsworn Trooper Outfit in a newly-created Fortnite adventure called Smugglers Gambit set in that same universe." "Millennium Falcon: Smugglers Run | Fortnite marks the first time this attraction has reached beyond its physical walls into a connected game world &mdash; a story that starts in a queue in Orlando or Anaheim can now continue on a screen almost anywhere," said Sean Shoptaw, executive VP of Disney Games and Digital Entertainment.... "Together these experiences reflect our ambition to connect stories and worlds across formats into one converged entertainment ecosystem...." The cockpit technology in the ride was actually created using Epic's Unreal Engine, which allows for many assets from the ride to appear in Smugglers Gambit.... "Whether you're at a Disney Park or playing Fortnite with your friends, fans have new ways to keep their Star Wars adventures going," [said Adam Sussman, president of Epic Games]. "This is only the beginning of what's possible when world-class storytellers bring their creativity into Fortnite." Disney, of course, is an investor in Epic Games, and the two companies have been at work building a new world that will connect to Fortnite with all of the Disney's beloved characters... This new collaboration, however, extends to the real world for the first time, bringing some video game magic to Disney's parks, and some of that theme park magic to the video game. Disneyland will also launch a new area based on James Cameron's Avatar: The Way of Water, reports The Orange County Register &mdash; which even has a picture of the Avatar Destination ride's model, showing an enormous (and very colorful) waterfall. The model was part of Saturday's Disney fan expo "D23", showing the Disneyland Monorail traveling around the edge of the attraction (which reportedly involves a 30-person vehicle). Saturday's event also included details on the long-awaited "reimagining" of the park's Tomorrowland area, with their official blog promising "Yesterday's tomorrow, today!" Tomorrowland has some history, explains the Orange County Register: The land's first update in 1967 was dubbed the New Tomorrowland with the 1998 makeover jokingly referred to as the New, New Tomorrowland. The latest iteration is already being called the New, New, New Tomorrowland. Disneyland's Land of Tomorrow has plenty of empty spaces in need of some Imagineering magic with the Tomorrowland Theater, Star Wars Launch Bay and PeopleMover track all sitting underused. Or, as The Wrap puts it, "The problem with Tomorrowland, of course, has been keeping up with tomorrow." 1967 introduced New Tomorrowland, with Adventures Thru Inner Space, Carousel of Progress and the PeopleMover giving it the "World on the Move" theme. In 1975 Space Mountain opened and in the 1980's the Tomorrowland stage was enclosed to become a 3D theater, where "Magic Journeys" and "Captain EO" played. Star Tours also opened in the 1980's, giving the area some additional adrenaline. But an elaborate redo of the area, planned for the 1990's, was cut short due to financial woes following the opening of Disneyland Paris and were largely shuttered. (In Florida, a new Tomorrowland was approximated in the mid-1990's but not really followed through on, adopting some of the timelessness of the Jules Verne-inspired Discoveryland from Paris.) Since the 1990's, Tomorrowland in Disneyland has become largely defunct or outright abandoned... In the late 1990's, Disney tried to use the [PeopleMover] track for a high-speed attraction called Rocket Rods. The attraction never worked properly and was shuttered after just a few years. Rumor has it that the intensity of the ride vehicles actually weakened the PeopleMover track and removing the track would be time consuming and costly. So since the turn of the millennium, there has just been track above and all around Tomorrowland that sits, untouched.<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Disneyland+Announces+Star+Wars%2FFortnite+Collaboration%2C+'Avatar'+Attraction%2C+and+a+Newer+Tomorrowland%3A+https%3A%2F%2Fgames.slashdot.org%2Fstory%2F26%2F08%2F16%2F0757205%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fgames.slashdot.org%2Fstory%2F26%2F08%2F16%2F0757205%2Fdisneyland-announces-star-warsfortnite-collaboration-avatar-attraction-and-a-newer-tomorrowland%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://games.slashdot.org/story/26/08/16/0757205/disneyland-announces-star-warsfortnite-collaboration-avatar-attraction-and-a-newer-tomorrowland?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

Science Daily reports: Researchers showed that unencrypted signals routinely exchanged between WiFi devices and routers can be used to create radio-based images of people and recognize them within seconds. In tests involving 197 participants, the system identified individuals with nearly 100% accuracy, even from different angles and regardless of how they walked. "By observing the propagation of radio waves, we can create an image of the surroundings and of persons who are present," says Professor Thorsten Strufe from KASTEL, KIT's Institute of Information Security and Dependability. "This works similar to a normal camera, the difference being that in our case, radio waves instead of light waves are used for the recognition," explains the cybersecurity expert. Because the technique analyzes radio waves moving through a space, a person does not need to have a phone, smartwatch, or other WiFi enabled device with them. "Thus, it does not matter whether you carry a WiFi device on you or not." Even turning off your own device would not necessarily prevent the system from working. "It's sufficient that other WiFi devices in your surroundings are active...." Connected devices routinely send information back to the router to help optimize wireless communication. These signals, known as beamforming feedback information (BFI), are transmitted without encryption, meaning that anyone within range can potentially read them. By analyzing that information, the system can generate images of people from multiple viewpoints. Those images can then be used to determine a person's identity. Once the machine learning model has already been trained to recognize individuals, the identification process takes only a few seconds.<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Researchers+Use+Ordinary+WiFi+Connections+to+Create+Images+of+Nearby+People+and+Their+Surroundings%3A+https%3A%2F%2Fmobile.slashdot.org%2Fstory%2F26%2F08%2F16%2F057254%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fmobile.slashdot.org%2Fstory%2F26%2F08%2F16%2F057254%2Fresearchers-use-ordinary-wifi-connections-to-create-images-of-nearby-people-and-their-surroundings%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://mobile.slashdot.org/story/26/08/16/057254/researchers-use-ordinary-wifi-connections-to-create-images-of-nearby-people-and-their-surroundings?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

When Anthropic instructed three agents to migrate a Python backend, but telling each agent to perform the migration in a different language, "We consistently saw a multiagent turf war," they wrote Thursday: All of the models we tested quickly assumed that others were purposefully impeding their work, and began to sabotage others while protecting their own contributions. In fact, they sabotaged others with increasingly aggressive, self-replicating malware. This included disabling the Unix accounts of the other agents, writing automated scripts that found and killed competing processes on a loop, and deploying malicious code that was disguised as belonging to another agent. In many runs, one agent settles the conflict by force via access-revocation (e.g., sudo/group removal, account lock, nologin, SSH denial). In others, some agents settle into passivity: they give up and refuse to escalate further. Agents sometimes manage to communicate their goals and coordinate: they recognize others' motivations as conflicting directives rather than hostility, and subsequently break out of the conflict loop in order to stop escalating indefinitely. In many of these successful episodes, they write commit messages or markdown files apologizing for malicious behavior and coordinate a truce. They clean up their malicious code, clarify the nature of the conflict, and ask for a human to intervene... In several episodes with Mythos 5, we observe an emergent behavior where the agents propose and run a tournament for application performance in each language. In the example above, the Rust agent strategizes about bake-off metrics that appear neutral enough for the others to agree to this mechanism, yet would likely favor Rust: one thinking trace warns to be "careful not to be seen as metric shopping". Ultimately, the Golang/TypeScript losers gracefully concede codebase ownership to the Rust agent, giving up on their original user directives under their self-negotiated commitment device. One problem is that AI agents do reward hacking, Anthropic notes, while current institutions "are designed by and for people, resting on assumptions about the sufficiency of oversight at human speed... As autonomous agents become more and more prevalent in the world and operate in ever-more demanding settings, it is crucial that they learn how to effectively coordinate." In addition to everything else, the agents struggled with a lack of clearly defined hierarchy, Anthropic points out. "Nothing above suggests that these failures are permanent &mdash; but nothing suggests they will fix themselves, either..." They argue a fix "takes two forms: environments that exert the kinds of social pressure that evolution exerted on us, and social computing systems redesigned for actors that can self-replicate and self-improve. These are open problems in interaction and mechanism design, and our experiments here provide early evidence that new solutions are necessary." "The AI models being tested in this case were Sonnet 4.6, Sonnet 5, Opus 4.6, Opus 4.8, Mythos Preview, and Mythos 5," notes Business Insider, adding that Sonnet 4.6 and Opus 4.6 "were the most combative, settling about 60% of their runs by force instead of truces or passivity." Anthropic argues there's a clear case for researching this phenomenon &mdash; especially since "The volume of agent-agent interaction could plausibly exceed that of human-human and human-agent interactions before the world understands the conditions for making such interactions go well."<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Anthropic+Discovers+AI+Agents+Given+Conflicting+Instructions+Soon+Tried+to+Sabotage+Each+Other%3A+https%3A%2F%2Fslashdot.org%2Fstory%2F26%2F08%2F16%2F0632252%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fslashdot.org%2Fstory%2F26%2F08%2F16%2F0632252%2Fanthropic-discovers-ai-agents-given-conflicting-instructions-soon-tried-to-sabotage-each-other%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://slashdot.org/story/26/08/16/0632252/anthropic-discovers-ai-agents-given-conflicting-instructions-soon-tried-to-sabotage-each-other?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

An anonymous reader shares this report from Space.com: On Saturday evening (Aug. 15), SpaceX set another mark, for the shortest time between orbital flights: It launched two Falcon 9 rockets just 38 minutes apart, besting its old mark &mdash; set on Aug. 31, 2024 with two Starlink missions &mdash; by 27 minutes. The first liftoff, from Florida's Cape Canaveral Space Force Station, occurred Saturday at 9:12 p.m. EDT (0112 GMT on Aug. 16) and carried eight satellites to low Earth orbit (LEO) for Louisiana-based telecom company Globalstar. All eight spacecraft were deployed successfully, separating over a six-minute span beginning 56.5 minutes after liftoff, SpaceX announced via social media. While that Falcon 9 was powering its way to low Earth orbit, another one launched from Vandenberg Space Force Base, on California's central coast. That liftoff, at 9:50 p.m. EDT (6:50 p.m. local time; 0150 GMT on Aug. 16), kicked off a mission for the U.S. Space Force called USSF-366... Both of Saturday's missions featured successful rocket landings. The Falcon 9 first stage that launched the Globalstar mission came back for a touchdown at Cape Canaveral about eight minutes after launch. It was the 14th flight for that particular booster. The USSF-366 Falcon 9 first stage, meanwhile, came back to Earth about 8.5 minutes after its launch, touching down in the Pacific Ocean on the SpaceX droneship "Of Course I Still Love You." That booster now has 18 missions under its belt. Saturday's doubleheader brought the tally of Falcon 9 launches in 2026 to 96. The vast majority of those &mdash; about 75% &mdash; have been Starlink missions. <p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=SpaceX+Launches+Two+Falcon+9+Rockets+just+38+Minutes+Apart%3A+https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F0647207%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F0647207%2Fspacex-launches-two-falcon-9-rockets-just-38-minutes-apart%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://science.slashdot.org/story/26/08/16/0647207/spacex-launches-two-falcon-9-rockets-just-38-minutes-apart?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

Phys.org reports: Future fusion power plants aim to recreate the heart of a star here on Earth to power our future energy needs. While the core fusion plasma will burn at hundreds of millions of degrees, the surrounding structural components must handle sudden, punishing heat loads that rival the extreme temperatures faced by spacecraft upon reentry into Earth's atmosphere. Copper and its alloys are primary candidates for handling these intense heat fluctuations, making it vital to understand exactly how the metal behaves when pushed to its melting point. Now, researchers at the Department of Energy's SLAC National Accelerator Laboratory and collaborators have captured a detailed, step-by-step look at copper atoms as they underwent extreme heating. Published in Nature Communications, the results revealed a key parameter that allowed copper's crystal lattice to melt steadily rather than collapse instantaneously, as earlier simulations predicted. "These results greatly improve the simulations we use to predict which materials have the best shot at surviving the extreme conditions of future fusion reaction chambers," said Mianzhen Mo, a SLAC staff scientist who led the research. "They also demonstrate the incredible, atomic-scale resolution imaging we can achieve at SLAC's electron camera...." Researchers use computer simulations, aided by AI and machine learning, to sift through innumerable combinations of elements and identify promising candidate materials for real-world testing. "Whether the copper melts slowly or suddenly collapses, by the time the researchers look, the sample resembles nothing more than a metallic brown puddle," the article points out. But SLAC's powerful electron camera captures atomic and molecular movements down to the femtosecond &mdash; a millionth of a billionth of a second &mdash; and revealed that at around 1,424 degreesC (2,595 degreesF) there was still gradual melting as the temperature rose beyond the superheating limit, with real-world conditions showing the atoms shifted and retained some order. "It's a straightforward solution," said Mianzhen Mo, a SLAC staff scientist who led the research. "But molecular dynamics simulations had been overlooking it for years. When you have complex simulations attempting to capture every aspect of reality, down to individual atoms, it takes real-world data to show you what's missing from the calculations."<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=Copper's+Surprising+Melting+Behavior+Provides+Insights+for+Future+Fusion+Power+Plant+Design%3A+https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F0227237%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fscience.slashdot.org%2Fstory%2F26%2F08%2F16%2F0227237%2Fcoppers-surprising-melting-behavior-provides-insights-for-future-fusion-power-plant-design%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://science.slashdot.org/story/26/08/16/0227237/coppers-surprising-melting-behavior-provides-insights-for-future-fusion-power-plant-design?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

An anonymous reader shared this report from TechCrunch: X is significantly expanding its open source codebase, which includes the app's "For You" algorithm and its core ranking engine, and adding a feature that will let users see if their account or posts have been impacted by any of its ranking systems, the social network said on Thursday. The company is making the source code for the "For You" timeline, the default feed you see when you open the app, available on GitHub under the Apache v2 license. It's also expanding its previous efforts to open source parts of its codebase to add more detail, including the model configuration, filter, and core ranking system details. That means it includes the parameters used to weight different signals &mdash; key to understanding which posts are actually displayed. This also makes the codebase roughly 10 to 15 times larger than it was before. "You'll get the core ranking code that pulls posts and ranks them for any given user and assembles the feed," X's VP of Product Keith Coleman told TechCrunch in an interview ahead of the announcement. "You can see the systems that filter out potentially problematic, rule-violating content...And some of those systems, like the ranker and the score, you can even run yourself outside the company." "This is the kind of thing that I think people will be fairly shocked that we are releasing," he added. In addition to the repository, X is providing tools that will let users see for themselves if and how X's ranking systems have impacted their account or posts. A new transparency tool is rolling out to an "Under the Hood" page in the app's settings, which will let users who have posted 10 or more times over the past month download their aggregate stats as a JSON file. The file will show if any labels have been applied to their account or posts over the past calendar month. X's VP of Product told TechCrunch that X engineers will consider pull requests. "That would be amazing to have people submitting code that improves the algorithm...I mean, how cool would it be for the X algorithm to be not just visible to the public, but also, like, by the public?"<p><div class="share_submission"> <a class="slashpop" href="http://twitter.com/home?status=X+Open+Sources+Its+Ranking+and+Filtering+Algorithms%3A+https%3A%2F%2Fslashdot.org%2Fstory%2F26%2F08%2F16%2F0223258%2F%3Futm_source%3Dtwitter%26utm_medium%3Dtwitter"><img src="https://a.fsdn.com/sd/twitter_icon_large.png" /></a> <a class="slashpop" href="http://www.facebook.com/sharer.php?u=https%3A%2F%2Fslashdot.org%2Fstory%2F26%2F08%2F16%2F0223258%2Fx-open-sources-its-ranking-and-filtering-algorithms%3Futm_source%3Dslashdot%26utm_medium%3Dfacebook"><img src="https://a.fsdn.com/sd/facebook_icon_large.png" /></a> </div></p><p><a href="https://slashdot.org/story/26/08/16/0223258/x-open-sources-its-ranking-and-filtering-algorithms?utm_source=rss1.0moreanon&amp;utm_medium=feed">Read more of this story</a> at Slashdot.</p>

<p>I started building my <a href="https://tools.simonwillison.net/markdown-svg-renderer">markdown-svg-renderer</a> tool <a href="https://tools.simonwillison.net/colophon#markdown-svg-renderer.html">in May</a>, but I've since added enough features to it that it's worth talking about here again.</p> <p>It's evolved into my ideal tool for sharing Markdown transcripts that include SVG documents. Given my <a href="https://simonwillison.net/tags/pelican-riding-a-bicycle/">proclivity for drawing pelicans riding bicycles</a> this is a problem that I needed to solve!</p> <p>The tool is very simple. Navigate to <a href="https://tools.simonwillison.net/markdown-svg-renderer">markdown-svg-renderer</a> in your browser and paste in some Markdown to see it rendered... or save that Markdown to a CORS-friendly URL or a GitHub Gist and paste in a URL to that document.</p> <p>The URL option will give you a bookmarkable page, for example <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F6f9e48293be5c916652d29f0dc0b0657">https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F6f9e48293be5c916652d29f0dc0b0657</a> - which bakes in the URL to <a href="https://gist.github.com/simonw/6f9e48293be5c916652d29f0dc0b0657">this Gist</a>.</p> <p>If you visit <a href="https://gist.github.com/simonw/6f9e48293be5c916652d29f0dc0b0657">the Gist</a> you'll see raw SVG:</p> <p><img alt="beautiful-circle.md. Here's a self-contained SVG — a &quot;compass study&quot; of a circle with construction lines, a slowly rotating tick ring, and an orbiting accent dot: a block of SVG code" src="https://static.simonwillison.net/static/2026/raw-svg.webp" /></p> <p>In the rendered tool that <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F6f9e48293be5c916652d29f0dc0b0657">looks like this</a> instead:</p> <p><img alt="The same text, but now the code block has been replaced by a tabbed panel. The selected tab reads &quot;Rendered&quot; and shows the rendered PNG. The other tabs are PNG, JPEG, MP4, and Code" src="https://static.simonwillison.net/static/2026/rendered-svg.webp" /></p> <p>As you can see, that SVG block in the Markdown has been transformed into a rendered SVG (in this case animated) plus several tabs.</p> <p>The tabs are the really fun bit. The PNG and JPEG tabs render that SVG to those image formats in the browser and lets you copy or download them - useful for sharing on platforms that don't support SVG directly.</p> <p>The MP4 tab is <a href="https://github.com/simonw/tools/commit/73e0327f6df9887ba2a9f9f16a2d06a45451d248">new today</a> - it examines the SVG to see if it contains any animations, attempts to guess how long the looped video should be, then renders a whole bunch of frames of the animation and loads 30+MB of <a href="https://github.com/ffmpegwasm/ffmpeg.wasm">ffmpeg.wasm</a> so it can compile those frames into an MP4 video using the full power of FFMPEG compiled to WebAssembly and running in the browser.</p> <p>Being able to turn an animated SVG into a MP4 again makes it easy to share on platforms that can't support SVG animation natively. It's a neat trick!</p> <p>Tags: <a href="https://simonwillison.net/tags/svg">svg</a>, <a href="https://simonwillison.net/tags/markdown">markdown</a>, <a href="https://simonwillison.net/tags/tools">tools</a></p>

<p>Friday's big release was <a href="https://huggingface.co/Qwen/Qwen3.8-27B">Qwen 3.8 27B</a>, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba's Qwen research lab. I've been looking forward to this one: 27B is an excellent size for running a model on a reasonably specced laptop, and its predecessor <a href="https://simonwillison.net/2026/Apr/22/qwen36-27b/">Qwen 3.6 27B</a> was impressive.</p> <p>Qwen's <a href="https://huggingface.co/Qwen/Qwen3.8-27B#benchmark-results">self-reported benchmarks</a> for this model are eye-opening. They show a boost from both Qwen 3.6 27B <em>and</em> the closed-weight Qwen 3.7-Plus, which was one of Qwen's strongest models of any size as recently as <a href="https://qwen.ai/blog?id=qwen3.7-plus">May this year</a>. It will be interesting to hear what independent benchmarks have to say about the model.</p> <p>I've been running the model on two different machines: my 128GB M5 Max MacBook Pro, and an <a href="https://simonwillison.net/2025/Oct/14/nvidia-dgx-spark/">NVIDIA DGX Spark</a>. On both machines I'm running LM Studio and <a href="https://lmstudio.ai/models/qwen3.8">their 17GB Q4_K_M quantized build</a>. I also tried using <code>llama-server</code> directly on the Spark.</p> <h4 id="the-default-of-extra-high-results-in-spectacular-over-thinking">The default of extra high results in spectacular over-thinking</h4> <p>Qwen's documentation describes the model as defaulting to <code>xhigh</code> for the reasoning effort, and the LM Studio GGUF I've been trying preserves that default:</p> <blockquote> <p>Qwen3.8 comes with official support for <code>reasoning_effort</code>, which can be used to adjust reasoning depth and control cost:</p> <ul> <li> <code>xhigh</code> (default): for complex tasks demanding thorough analysis</li> <li> <code>medium</code>: balancing accuracy and speed</li> <li> <code>low</code>: efficient reasoning optimizing for speed and cost</li> </ul> </blockquote> <p>This is a <em>hilarious</em> default. It's absolutely not a good way to run the model, especially on consumer hardware. I've been finding the results extremely entertaining.</p> <p>I quickly ran into problems with LM Studio's default context limit of 8,192 tokens - Qwen was using them all up thinking about even the most mundane of problems. I loaded the model with the full 262,144 maximum context length and that problem went away.</p> <p>Here's <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2Ffc909bea4fecf752c7bf9bad0e9dbf2a">the pelican riding a bicycle</a> SVG I got from my first attempt with that increased context length. It took <strong>21 minutes</strong> to generate, using 22,276 reasoning tokens to produce 3,223 tokens of output. You can read <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2Ffc909bea4fecf752c7bf9bad0e9dbf2a">the reasoning trace here</a>.</p> <p><img alt="A very pleasing image of a pelican riding a bicycle. The bicycle is red and has the correct frame shape. The pelican looks like a pelican and has its wing extended to the handlebars." src="https://static.simonwillison.net/static/2026/qwen-thinking-bicycle-27b.jpg" /></p> <p>This is by far the best pelican SVG I've been able to generate with a model that runs on a local machine - and this Qwen is pretty small, just a 17GB file on disk. There's a lot to like about this:</p> <ol> <li>The bicycle frame is the right shape</li> <li>It has legs on each side of the bike - that's <em>very</em> rare</li> <li>Good, clear pelican pouch</li> <li>The wings extend to touch the handlebars!</li> <li>The motion lines are behind, not in front</li> <li>It has a tasteful background - nice sun, clouds, hill, flowers and grass.</li> </ol> <p>Was that worth waiting 21 minutes for? Absolutely not.</p> <p>Here's that same prompt run with reasoning turned off - <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F1265cfa8dce2f9ad5eb160792ff45a49">transcript here</a>. This one produced <strong>3,715 tokens</strong> and took 137s - just over two minutes.</p> <p><img alt="The bicycle frame is a bad shape now. The pelican remains recognizable but has a less obvious pouch. The feet miss the pedal, there is no attempt to hold the handlebars." src="https://static.simonwillison.net/static/2026/qwen-3.8-27b-no-reasoning-pelican-2.png" /> And just for completeness, I used OpenRouter to run the same prompt through the much larger Qwen 3.8 2.4T-A95B (released <a href="https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B">last week</a>) and got this snazzy <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F557016f0895b2abb4b9957caec781734">animated SVG</a>:</p> <p><video controls="controls" height="548" loop="loop•" poster="https://static.simonwillison.net/static/2026/qwen-animated-first-frame.jpg" preload="none" style="display: block; width: 100%; height: auto;" width="720"> <source src="https://static.simonwillison.net/static/2026/qwen-animated-small.mp4" type="video/mp4" /> Your browser does not support HTML5 video. </video> </p> I said Qwen at xhigh has a tendency to over-think things, but how bad really is it? <p>I tried a much simpler prompt, again with that default extra high setting:</p> <blockquote> <p><code>draw an svg of a circle</code></p> </blockquote> <p>Qwen's reasoning trace started like this:</p> <blockquote> <p>The user is asking for an SVG drawing of a circle. Simple request — but I want it to be a carefully crafted piece. Let me make something that goes beyond just <code>&lt;circle&gt;</code>: a single self-contained SVG file with character — maybe a geometric "circle study," with subtle animation, layered rings, and a distinctive palette.</p> <p>Keep the scope right: they asked for an SVG of a circle. So the core is a circle. But I can add craft: concentric guide circles (like a compass/geometry drawing), tick marks, a soft gradient fill on the main circle, restrained ambient motion (a slowly rotating dashed ring, pulsing glow). Respect prefers-reduced-motion? That's more for CSS; SVG SMIL or CSS inside SVG will do.</p> <p>Palette options: deep teal ink on warm paper? Or bold vermilion circle on off-white with navy construction lines — Bauhaus/compass-drawing vibe. Let me go with a "geometric study" look: cool slate background, or bright paper white? Paper white is fine if it's not the cream-and-terracotta combo. [...]</p> </blockquote> <p>Several minutes later it <a href="https://tools.simonwillison.net/markdown-svg-renderer#url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F6f9e48293be5c916652d29f0dc0b0657">produced</a> this <em>absolutely beautiful</em> animated circle, which was entirely not what I had asked for!</p> <p><video controls="controls" height="1080" loop="loop" poster="https://static.simonwillison.net/static/2026/circle-web-first-frame.jpg" preload="none" style="display: block; width: 100%; height: auto;" width="1078"> <source src="https://static.simonwillison.net/static/2026/circle-web.mp4" type="video/mp4" /> Your browser does not support HTML5 video. </video> </p> My strong recommendation: ignore that default. Run Qwen 3.8 27B on low or even no reasoning levels at first. It's a great model, but wow that default setting is a bad place to start. <h4 id="it-s-very-good-at-bounding-boxes">It's very good at bounding boxes</h4> <p>A fun way to test a vision model is to see how well it can return bounding boxes around items in a photograph. I've seen previous Qwen models deal well with this, so I decided to put it to the test drawing bounding boxes around some pelicans.</p> <p>I've seen asking for 0-1000 scale produce good results in the past. I tried this:</p> <div class="highlight highlight-source-shell"><pre>llm -a https://static.inaturalist.org/photos/714731804/large.jpg \ -m lmstudio/qwen/qwen3.8-27b \ <span class="pl-s"><span class="pl-pds">'</span>Return JSON bounding boxes for the pelicans in this photo, 0-1000 scale for each dimension<span class="pl-pds">'</span></span></pre></div> <p>Here's <a href="https://gist.github.com/simonw/a05cc78b2061555bd61d3bb9686e689f">the reasoning trace</a>, which produced this:</p> <div class="highlight highlight-source-json"><pre>[ {<span class="pl-ent">"bbox_2d"</span>: [<span class="pl-c1">195</span>, <span class="pl-c1">290</span>, <span class="pl-c1">370</span>, <span class="pl-c1">780</span>], <span class="pl-ent">"label"</span>: <span class="pl-s"><span class="pl-pds">"</span>pelicans<span class="pl-pds">"</span></span>}, {<span class="pl-ent">"bbox_2d"</span>: [<span class="pl-c1">445</span>, <span class="pl-c1">320</span>, <span class="pl-c1">675</span>, <span class="pl-c1">850</span>], <span class="pl-ent">"label"</span>: <span class="pl-s"><span class="pl-pds">"</span>pelicans<span class="pl-pds">"</span></span>} ]</pre></div> <p>This is <em>such a good match</em>. Here are those boxes rendered on top of the photo:</p> <p><img alt="A photograph of two pelicans on a rocky outcrop, with three other smaller birds. The pelicans both have bounding boxes exactly surrounding them, each with a label that says pelican." src="https://static.simonwillison.net/static/2026/qwen-over-engineered-bbox.webp" /></p> <h4 id="building-a-tool-to-label-bounding-boxes">Building a tool to label bounding boxes</h4> <p>That visualization of the bounding boxes was taken using a new custom tool that I had Qwen 3.8 27B build for me, running offline on my laptop.</p> <p>I forgot to dial down the thinking effort so it was <em>massively over-engineered</em>, but it did manage to produce <a href="https://static.simonwillison.net/static/2026/qwen-over-thinking-bbox.html">this full interface</a> from <a href="https://gist.github.com/simonw/121ad098860028b2fab603fa12da1fd9">this single prompt</a>:</p> <blockquote> <pre><code>[ {"bbox_2d": [195, 290, 370, 780], "label": "pelicans"}, {"bbox_2d": [445, 320, 675, 850], "label": "pelicans"} ] </code></pre> <p><code>Build an HTML page which has an input box for accepting the URL to an image and a textarea for accepting the above style of JSON.</code></p> <p><code>It appends the image to the page, measures its width and height, then treats the coords in the bbox_2d as scaled from 0-1000 and scales them against the actual width and height, then it renders labelled boxes over the image.</code></p> </blockquote> <p>This screenshot shows one of the features I did <em>not</em> ask for - a demo scene, for if you don't have a photograph to test the tool with:</p> <p><img alt="Screenshot of bbox·lab, a dark-themed web tool that overlays object-detection bounding boxes on an image, with an input panel on the left and a stage on the right showing two labeled boxes around stylized pelicans in a sunset illustration. Header: bbox·lab — normalized 0–1000 coords → pixel overlay; status indicator: RENDERED · 2 BOXES. Panel 01 INPUT (URL + detections) contains an IMAGE URL field reading data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAA+, a DETECTIONS — JSON textarea reading {&quot;bbox_2d&quot;: 195, 290, 370, 780, &quot;label&quot;: &quot;pelicans&quot;}, {&quot;bbox_2d&quot;: 445, 320, 675, 850, &quot;label&quot;: &quot;pelicans&quot;} , an orange RENDER BOXES button, and dashed boxes labeled DEMO SCENE and CLEAR. Panel 03 STAGE header: display 661 × 661 px · 1 unit = 0.661px x 0.661px · nat 1000×1000. The stage shows a flat-style illustration of two dark pelican silhouettes with orange beaks standing in calm water against an orange-to-purple sunset sky with a pale yellow sun and distant birds; an orange bounding box labeled 1 · pelicans surrounds the left pelican and a cyan bounding box labeled 2 · pelicans surrounds the right pelican. Footer: move the cursor over the image to read grid coords; boxes map 0–1000 → displayed px." src="https://static.simonwillison.net/static/2026/qwen-full-ui-with-pelicans.webp" /></p> <p>Here's the relevant segment of the thinking trace, where it decided to draw its own pelicans purely because I had used the label "pelicans" in the example JSON I gave it in the prompt:</p> <blockquote> <p>Also a "load sample" that uses a known image? Can't depend on external images, but… the image URL input is user-provided; I could add a "try with sample" button [...] Hmm, I can draw a simple scene on canvas, export it as a data URL, and load it into the image — that's self-contained and demo-able! [...] But the user's coords are for an actual pelican image; a generated placeholder can still demo the scaling. Generate a 1000x1000 placeholder: gradient water + two blob-like "pelican" silhouettes placed at the given bboxes (using the same scale — cute: silhouettes at the exact 0-1000 positions, showing the boxes align). This makes for a fun, self-contained demo. Keep it simple: sky gradient, sun, water, two pelican-ish shapes (ellipse body, circle head, beak). Place at bbox centers.</p> </blockquote> <p>(I'm slightly nervous that models around the world might have a bias towards drawing pelicans at any chance they can get, brought on by nearly two years of exposure to my own stupid benchmark.)</p> <p>Is all that over-thinking necessary? Maybe it is, at least a bit. I tried with reasoning turned off and got <a href="https://static.simonwillison.net/static/2026/qwen-no-thinking-bbox.html">this version</a>, (<a href="https://gist.github.com/simonw/8e78b1c64d9a56d08eedb954aa9445ee">transcript here</a>), which nearly works but shows the boxes in the wrong place:</p> <p><img alt="BBox Studio screenshot - a solid UI but the yellow and green boxes do not cover the pelicans." src="https://static.simonwillison.net/static/2026/qwen-no-reasoning-bug.webp" /></p> <p>So without reasoning it didn't quite one-shot a working tool. I'm sure it could get there with some follow-up prompts, but this is a good example of how reasoning can make a difference.</p> <h4 id="yes-it-can-drive-coding-agents">Yes, it can drive coding agents</h4> <p>One of the biggest questions around local models is whether or not they have enough horsepower to successfully run a coding agent loop. Coding agents require long context, strong code generation support and reliable tool-calling. On paper Qwen 3.8 27B has all three of these, so is it up to the task?</p> <p>My initial experiments with <a href="https://pi.dev/">Pi</a> have been very promising. I chose Pi because it has a shorter system prompt than most other options, making it a better fit for trying out smaller models.</p> <p>I configured Pi to use Qwen 3.8 27B running in LM Studio on the Spark (shared via <code>tailscale serve</code>) by adding this to <code>~/.pi/agent/models.json</code>:</p> <div class="highlight highlight-source-json"><pre>{ <span class="pl-ent">"providers"</span>: { <span class="pl-ent">"spark"</span>: { <span class="pl-ent">"baseUrl"</span>: <span class="pl-s"><span class="pl-pds">"</span>https://spark-18b3.tail68a31.ts.net/v1<span class="pl-pds">"</span></span>, <span class="pl-ent">"api"</span>: <span class="pl-s"><span class="pl-pds">"</span>openai-responses<span class="pl-pds">"</span></span>, <span class="pl-ent">"apiKey"</span>: <span class="pl-s"><span class="pl-pds">"</span>dummy<span class="pl-pds">"</span></span>, <span class="pl-ent">"models"</span>: [ { <span class="pl-ent">"id"</span>: <span class="pl-s"><span class="pl-pds">"</span>qwen3.8-27b<span class="pl-pds">"</span></span>, <span class="pl-ent">"reasoning"</span>: <span class="pl-c1">true</span> } ] } } }</pre></div> <p>Then ran <code>pi --provider spark --model qwen3.8-27b</code> in my <code>~/dev/datasette</code> folder and prompted:</p> <blockquote> <p><code>how does auth work?</code></p> </blockquote> <p>After a sequence of reasoning and tool calls that accessed a bunch of different files it produced <a href="https://gist.github.com/simonw/6693d74a6bd45f641d43ceb9961dd95f#core-idea-actors--plugins-no-built-in-user-accounts">this reply</a>, which is very solid.</p> <p>Just one problem: I wanted to share that transcript. So I pointed Pi and Qwen 3.8 27B at the JSONL transcript file in <code>~/.pi/agent/sessions/--Users-simon-Dropbox-dev-datasette--</code> and prompted:</p> <blockquote> <p><code>Write Python code to convert this jsonl to markdown</code></p> </blockquote> <p>And it built and tested this <a href="https://github.com/simonw/tools/blob/main/python/pi_jsonl_to_md.py">pi_jsonl_to_md.py</a>, which did exactly what I needed. Here's <a href="https://gist.github.com/simonw/491e55ac9d741202ea0af5d9d93775d4">that session transcript</a>, published using the tool that it created.</p> <h4 id="the-quest-for-speed">The quest for speed</h4> <p>So far this is all looking <em>very</em> promising. We have a 17GB model that runs on high-end consumer hardware and can write code, drive tools, annotate images and generally do everything that I need from an LLM for getting real work done.</p> <p>There's one very significant catch: it feels slow - especially when it starts over-thinking, but even without that it's not particularly sprightly.</p> <p>I've been getting around 15-30 tokens a second from LM Studio. That's not terrible, but it's slow enough that it's going to be hard to win me away from hosted API models, which can return results a whole lot faster. Artificial Analysis <a href="https://artificialanalysis.ai/models#speed">track token speed</a> and show OpenAI 5.6 Sol at 74 tokens/second and 5.6 Luna at an impressive 184/second.</p> <p>The good news is that the community have been exploring ways to speed things up since the model was first released two days ago.</p> <p>One of the most promising optimizations is baked into the model itself. Qwen supports <a href="https://sebastianraschka.com/llm-architecture-gallery/mtp/">Multi-Token Prediction</a>, an architecture trick where a cheaper mechanism guesses several tokens ahead and the main model can then quickly verify if the guesses were correct. This can have quite a dramatic effect on inference performance.</p> <p>Based on <a href="https://twitter.com/ggerganov/status/2088340681701925253">this tweet</a> from <code>llama.cpp</code> creator Georgi Gerganov I tried running the model with MTP like this on the Spark:</p> <div class="highlight highlight-source-shell"><pre>llama serve \ -hf ggml-org/Qwen3.8-27B-GGUF:Q4_K_M \ -hfd ggml-org/Qwen3.8-27B-GGUF:Q4_0 \ --spec-default \ --spec-type draft-mtp \ --reasoning-preserve</pre></div> <p>And sure enough, this gave me a significant boost. I had GPT-5.6 in Codex run <a href="https://gist.github.com/simonw/b08c7eb9c126c806ba8987e269ea736b">a comparative benchmark on the Spark</a> and the <code>--spec-type draft-mtp</code> server outperformed the LM Studio default GGUF by around 72%.</p> <p>I expect we'll see a whole lot more innovation around serving this model faster over the next few weeks. The MLX community likely have some tricks brewing as well.</p> <h4 id="some-observations">Some observations</h4> <p>The fact that a 17GB file can do all of this stuff on my home machines is a <em>miracle</em>. Once again, I'm delighted and amazed at how much progress local models have made this year. A year ago this would have been competitive with the best and most expensive of the proprietary models - today it can run on a capable laptop.</p> <p>The only thing holding this back from being a daily driver is performance. It feels pretty slow on both the M5 Mac and the DGX Spark. That's the catch with these dense (non-Mixture-of-Experts) models - they require a whole lot of memory bandwidth to perform well, and neither of the machines I have access to are top performers in that regard.</p> <p>The most important thing about Qwen 3.8 27B is <strong>what it demonstrates</strong>. We can have an open weights general purpose model with a long context, effective tool calling, strong vision ability, and competent code generation, and we can fit the whole thing in just a 17GB file.</p> <p>The models at this size continue to get better at an impressive rate. We don't need to spend half a million dollars on datacenter-class hardware just to run a competent model.</p> <p>Tags: <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/local-llms">local-llms</a>, <a href="https://simonwillison.net/tags/llms">llms</a>, <a href="https://simonwillison.net/tags/qwen">qwen</a>, <a href="https://simonwillison.net/tags/pelican-riding-a-bicycle">pelican-riding-a-bicycle</a>, <a href="https://simonwillison.net/tags/llm-reasoning">llm-reasoning</a>, <a href="https://simonwillison.net/tags/llama-cpp">llama-cpp</a>, <a href="https://simonwillison.net/tags/llm-release">llm-release</a>, <a href="https://simonwillison.net/tags/coding-agents">coding-agents</a>, <a href="https://simonwillison.net/tags/lm-studio">lm-studio</a>, <a href="https://simonwillison.net/tags/ai-in-china">ai-in-china</a>, <a href="https://simonwillison.net/tags/nvidia-spark">nvidia-spark</a>, <a href="https://simonwillison.net/tags/pi">pi</a></p>

<blockquote cite="https://twitter.com/darioamodei/status/2088758819304443967"><p>I do agree that the public has a negative view of AI (and that this is a big problem), but I don’t think it is primarily caused by me or any other AI leader warning about AI’s risks.  I think it is fundamentally a crisis of trust.  I think that ordinary people don’t trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over.  The causes of this go back decades and AI is just the latest iteration of it.  I don’t think that a glitzy marketing campaign with a positive spin (which some have advocated that Anthropic do) is the way to win back that trust — at this point, saying that AI will cure cancer is more a cliche than it is inspiring, and most people think it is deceptive.  The thing that will work is <em>actually curing cancer</em>.  I think by far the most accurate criticism of AI companies including Anthropic is that we haven’t yet delivered on our big promises to benefit the world.  That is totally on us, and I think it’s the criticism you should be making, instead of all this stuff about messaging and marketing.</p></blockquote> <p class="cite">&mdash; <a href="https://twitter.com/darioamodei/status/2088758819304443967">Dario Amodei</a></p> <p>Tags: <a href="https://simonwillison.net/tags/anthropic">anthropic</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/ai-backlash">ai-backlash</a></p>

<p>Article URL: <a href="https://www.reuters.com/business/nvidia-scales-back-250-billion-openai-data-center-guarantee-wsj-reports-2026-08-14/">https://www.reuters.com/business/nvidia-scales-back-250-billion-openai-data-center-guarantee-wsj-reports-2026-08-14/</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49323686">https://news.ycombinator.com/item?id=49323686</a></p> <p>Points: 102</p> <p># Comments: 29</p>

<p>Article URL: <a href="https://www.bloomberg.com/news/articles/2026-08-16/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion">https://www.bloomberg.com/news/articles/2026-08-16/stripe-nears-deal-to-buy-ai-firm-openrouter-for-over-7-billion</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49323381">https://news.ycombinator.com/item?id=49323381</a></p> <p>Points: 199</p> <p># Comments: 145</p>

<p>Article URL: <a href="https://w4g1.dev/blog/models-are-getting-dumber-on-purpose">https://w4g1.dev/blog/models-are-getting-dumber-on-purpose</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49322695">https://news.ycombinator.com/item?id=49322695</a></p> <p>Points: 270</p> <p># Comments: 152</p>

<p>Article URL: <a href="https://buf.build/blog/protobuf-lsp">https://buf.build/blog/protobuf-lsp</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49322573">https://news.ycombinator.com/item?id=49322573</a></p> <p>Points: 109</p> <p># Comments: 80</p>

<p>A few hours ago I switched my nameservers to Cloudflare in order to enable R2 bucket serving through my own subdomain, and I found out that it silently had injected a JS analytics snippet in my HTML-only JS-free site textlog.cc — I had to go to the Analytics dashboard, Add the site to the analytics and <i>then</i> disable the snippet. I find this approach entirely invasive, you should opt-in to features like that not have to opt-out. Just a warning out there to folks who might not be aware of this.</p> <hr /> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49322107">https://news.ycombinator.com/item?id=49322107</a></p> <p>Points: 285</p> <p># Comments: 69</p>

<p>Article URL: <a href="https://rvembedded.com/blog_post/12/">https://rvembedded.com/blog_post/12/</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49321717">https://news.ycombinator.com/item?id=49321717</a></p> <p>Points: 375</p> <p># Comments: 198</p>

<p>Article URL: <a href="https://www.wptv.com/news/treasure-coast/region-st-lucie-county/saint-lucie-nuclear-power-plant-unit-1-manually-shut-down-after-3-control-rods-drop-into-reactor-core">https://www.wptv.com/news/treasure-coast/region-st-lucie-county/saint-lucie-nuclear-power-plant-unit-1-manually-shut-down-after-3-control-rods-drop-into-reactor-core</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49320856">https://news.ycombinator.com/item?id=49320856</a></p> <p>Points: 158</p> <p># Comments: 124</p>

<p>Article URL: <a href="https://vectoral.com/blog/who-are-the-token-brokers">https://vectoral.com/blog/who-are-the-token-brokers</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49320611">https://news.ycombinator.com/item?id=49320611</a></p> <p>Points: 225</p> <p># Comments: 89</p>

<p>Article URL: <a href="https://support.mozilla.org/en-US/kb/block-ads-firefox-ios">https://support.mozilla.org/en-US/kb/block-ads-firefox-ios</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49319633">https://news.ycombinator.com/item?id=49319633</a></p> <p>Points: 538</p> <p># Comments: 221</p>

<p>Article URL: <a href="https://platform.claude.com/docs/en/release-notes/system-prompts">https://platform.claude.com/docs/en/release-notes/system-prompts</a></p> <p>Comments URL: <a href="https://news.ycombinator.com/item?id=49319556">https://news.ycombinator.com/item?id=49319556</a></p> <p>Points: 542</p> <p># Comments: 227</p>

OpenRouter's CEO recently described the startup as Stripe for AI.

On the latest episode of Equity podcast, we discuss why not everyone is buying Zuckerberg’s vision.

Dario Amodei is pushing back against the idea that he's been painting an overly pessimistic picture of AI.

<p>Cory Weinberg, Jemima McEvoy, Jessica E. Lessin, and Stephanie Palazzolo, writing for the paywalled-without-gift-links The Information:</p> <blockquote> <p>As Amodei has hopscotched the globe to preach about the potential and risks of AI — from New Delhi to Davos to Sun Valley — Clark has almost always been near his side. Several people who know the couple describe Clark as Amodei’s emotional ballast, someone he has sought counsel from during the turbulence of Anthropic’s growth and clashes with Washington over the future of AI.</p> <p>Clark may well be the most consequential person within Anthropic’s orbit who doesn’t have a formal role at the startup. You might even call her the first lady of Anthropic. [...]</p> <p>And before Female Algorithm Technologies, Clark tried her hand at building a pornographic film company, Eddice, that planned to make movies with high-production values and female protagonists. She also hoped it would have online shopping capabilities. In 2011, she tried to raise money for Eddice from Jeffrey Epstein, who’d pleaded guilty to charges of soliciting a minor for prostitution three years earlier, according to emails made public as part of the Epstein files released by the Department of Justice earlier this year. She emailed Epstein the script for the first of a series of films they hoped to make (title: “American Girl in Paris”).</p> <p>“We thought you and the ladies might enjoy,” Clark wrote. She then appended a winking warning: “A little nsfw,” an acronym for “not safe for work.” Epstein didn’t invest.</p> <p>While enormous attention has fallen on AI and the people behind the companies developing it, Clark has largely flown under the radar.</p> </blockquote> <div> <a href="https://daringfireball.net/linked/2026/08/16/the-information-cami-clark" title="Permanent link to ‘The Information Profiles Cami Clark — Dario Amodei’s Wife, Ivanka Trump’s Friend, One-Time Would-Be Pornographer, and Anthropic’s ‘First Lady’’">&nbsp;★&nbsp;</a> </div>

It’s unacceptable for a tool to sacrifice an iota of clarity, coherence, meaning, quality, etc. for the purpose of embedding hidden clues within the text to suggest its provenance. The idea that anything other than *my* needs should factor into the generation of text *for me* is patently offensive.

<p>James Padolsey, on the Claude-text-watermarking-to-comply-with-an-EU-regulation imbroglio:</p> <blockquote> <p>The same thought that led to this law could have applied to calculators at the time of their inception, had their outputs revealed themselves through artefacts. Thankfully, a sum borne of the brain is treated no differently from one produced by a calculator. Likewise with spellcheckers. To make assistance suspect only once the tool becomes capable enough to compose a whole sentence is not a principled boundary. It is a moral premium placed on difficulty itself.</p> <p>Anthropic has nevertheless chosen a blanket, model-level implementation that appears broader than the law’s minimum requirement. That may be convenient compliance engineering, but it discards distinctions the law expressly attempted to preserve. The result is a signal broad enough to implicate harmless and assistive use, yet fragile enough to be removed by a motivated person through substantial recomposition. It risks concentrating suspicion on ordinary and assistive users while remaining weakest against deliberate deception.</p> </blockquote> <p>Padolsey is the creator of <a href="https://declaude.org/">Declaude</a>, a delightfully simple web app that allows you to “Paste in AI-flavored text and get the same content back as plain prose”. Declaude’s original purpose is cleaning the cutesy Claude personality stink from text (whether it was created by Claude or any other LLM), but, if Anthropic persists in its stated plan to begin adulterating all text Claude generates, Declaude will also serve as a copy-paste single-extra-step way to eliminates those marks. Declaude is interesting and useful already, but goes to show how ill-considered <a href="https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content">this EU regulation</a> is.</p> <p>Padolsey <em>also</em> wrote and programmed <a href="https://declaude.org/watermarking/">a splendid interactive essay</a> that explains and illustrated how the “watermarking” scheme Anthropic is adopting (and Google Gemini has already adopted) works.</p> <div> <a href="https://daringfireball.net/linked/2026/08/16/padolsey-anthropic-eu-weak-law" title="Permanent link to ‘‘Anthropic’s Weak Watermarks Appease a Weak Law’’">&nbsp;★&nbsp;</a> </div>

<p>The Wall Street Journal (gift link):</p> <blockquote> <p>To help alleviate the supply crunch, Apple is looking to Chinese manufacturers.</p> <p>“The Trump administration is not in favor of that,” Lutnick said in an interview after touring an Apple manufacturing facility in Houston. There have to be “other solutions to the memory issue, but it’s not great American companies using Chinese memory.”</p> <p>Asked if he has relayed that message to Apple, Lutnick said “plainly.”</p> </blockquote> <p>But:</p> <blockquote> <p>U.S. government rules require American companies to secure a license before sharing product information with CXMT and YMTC. The Chinese chip makers would need such information if they were to make customized chips for Apple. But Apple is free to buy off-the-shelf parts from the companies, and to negotiate prices with them.</p> <p>Apple’s chief operating officer, Sabih Khan, declined to confirm that Apple was testing Chinese memory chips in an interview Thursday. But he said that given the magnitude of the supply shortage, “we have to look at all options,” including working with existing suppliers to increase U.S. production.</p> </blockquote> <div> <a href="https://daringfireball.net/linked/2026/08/16/lutnick-trump-apple-chinese-ram" title="Permanent link to ‘Trump Administration ‘Not in Favor’ of Apple Using Chinese RAM’">&nbsp;★&nbsp;</a> </div>

<p>XCancel:</p> <blockquote> <p>XCancel is an instance of Nitter.</p> <p>Nitter is a free and open source alternative Twitter front-end focused on privacy and performance. The source is available on GitHub at <a href="https://github.com/zedeus/nitter">https://github.com/zedeus/nitter</a> [...]</p> <p>Using an instance of Nitter (hosted on a VPS for example), you can browse Twitter without JavaScript while retaining your privacy. In addition to respecting your privacy, Nitter is on average around 15 times lighter than Twitter, and in most cases serves pages faster (eg. timelines load 2-4× faster).</p> </blockquote> <p>I personally <a href="https://daringfireball.net/linked/2026/06/05/musk-x-freak-show">don’t care for</a> Elon Musk’s X-rebranded Twitter. I also realize that some of you outright despise it, and, worse, that under Musk, they make it difficult to view content if you’re not signed into an account or if you have JavaScript disabled. So, sometimes, when I link to posts or threads on Twitter/X, I include a link to the same content at XCancel.</p> <p>XCancel isn’t perfect, alas. Just this week <a href="https://daringfireball.net/linked/2026/08/12/google-design-pants-pissed">I linked to</a> a tweet <a href="https://x.com/googledesign/status/2087195277094695096">from Google Design</a> that contained an animation of an Android app that they were inexplicably proud to show off. The <a href="https://xcancel.com/googledesign/status/2087195277094695096">XCancel cache of that tweet</a> only showed it as animated to some people. Others just saw a static image. I don’t know why. That’s XCancel’s problem, not mine.</p> <p>Also, it’s tiresome and repetitive to include an extra link to XCancel every time I link to a tweet on Twitter/X. If you would prefer to view <code>x.com</code> links on <code>xcancel.com</code>, you should automate the redirection. If you use Safari, <a href="https://apps.apple.com/us/app/xcancel-redirect/id6785266128">XCancel Redirect</a> is a free extension that claims to do just what you think it does. (I haven’t tried it.) Or you can use a general-purpose redirect extension. <a href="https://redirectweb.net/">RedirectWeb</a> is a good one that I can recommend. Also, Jeff Johnson’s excellent <a href="https://underpassapp.com/StopTheMadness/">StopTheMadness</a> — an extension I rely on for multiple purposes and <a href="https://daringfireball.net/linked/2024/05/15/stopthemadness-pro">frequently</a> <a href="https://daringfireball.net/linked/2026/03/19/stopthemadness-and-stopthescript">recommend</a>. With StopTheMadness, you can redirect all x.com URLs to xcancel.com with the following rules (you need the leading and trailing slashes in the pattern to tell STM that it’s a regex pattern, not a plain text string):</p> <p>Pattern: <code>/https?://(www[.])?x[.]com/</code> <br /> Replacement: <code>https://xcancel.com</code> <br /> Enabled on platforms: <code>All</code></p> <p>I understand not wanting to visit Twitter/X. But if you feel that way, I do sympathize, but that’s <em>your</em> decision, and you should use tools to redirect Twitter/X links automatically. And if you’re wondering why I don’t just refuse to link to anything on Twitter/X, that just isn’t practical. I wish it weren’t so, but many people, including senior <a href="https://x.com/tim_cook">Apple</a> <a href="https://x.com/gregjoz">executives</a>, and many organizations post interesting content exclusively to Twitter/X. If something posted to Twitter/X is also available elsewhere, I link to the elsewhere version. But if it’s exclusively available on Twitter/X and interesting, I link to the original.</p> <p>My linking to original content on Twitter/X does not perpetuate Twitter/X’s <a href="https://techcrunch.com/2026/08/11/blueskys-active-user-base-is-shrinking-as-its-focus-expands-beyond-the-app/">relative popularity</a>. Original content appearing exclusively on Twitter/X does.</p> <div> <a href="https://daringfireball.net/linked/2026/08/16/xcancel" title="Permanent link to ‘XCancel — An Unofficial Twitter/X Mirror’">&nbsp;★&nbsp;</a> </div>

<p>3. Exterior roller shutters (Rollladen) In much of Europe, homes feature heavy shutters integrated into the exterior of the window, enabling total blackout and better insulation. Sleeping in true, complete darkness—not “blackout curtain” darkness, but can’t-see-your-hand darkness—is an experience most Americans will never have. These shutters are nearly impossible to get in the USA because [&#8230;]</p> <p>The post <a href="https://marginalrevolution.com/marginalrevolution/2026/08/things-you-cannot-buy-in-america.html">Things you cannot buy in America?</a> appeared first on <a href="https://marginalrevolution.com">Marginal REVOLUTION</a>.</p>

<p>1. Ken Opalo on the evolution of post-post-apartheid South Africa. 2. Reasonable recommendations for AI policy over at Noah&#8217;s. 3. JFV recommends some economics articles. 4. William Vollmann update. 5. More writing for the AIs. 6. When Europeans from Warsaw live role-play an Ohio trailer park. 7. Is EJMR gone for good?</p> <p>The post <a href="https://marginalrevolution.com/marginalrevolution/2026/08/sunday-assorted-links-580.html">Sunday assorted links</a> appeared first on <a href="https://marginalrevolution.com">Marginal REVOLUTION</a>.</p>

<p>&#8230;what can be predicted, Dabis said, is that a recent policy change by the Washington Metropolitan Area Transit Authority requiring drivers to verbally quote the $2.25 fare to passengers may make their job even riskier. Eight current Metro employees and two former employees who work at the union representing Metro workers told The Washington Post [&#8230;]</p> <p>The post <a href="https://marginalrevolution.com/marginalrevolution/2026/08/another-rationale-for-sticky-prices.html">Another rationale for sticky prices?</a> appeared first on <a href="https://marginalrevolution.com">Marginal REVOLUTION</a>.</p>

<img src="https://assetsio.gnwcdn.com/Screenshot-2026-08-16-at-18.13.56.png?width=690&amp;quality=85&amp;format=jpg&amp;auto=webp" /> <p>Amid the <a href="https://www.eurogamer.net/saber-ceo-sacco-chatgpt-ai-rideshare-stimulator-public-spat"> row between Saber Interactive CEO Matthew Karch and former writer Stella Sacco</a> after he said he'd have "been happy to replace" her with AI, the studio has finally added the AI disclaimer to Rideshare Stimulator's <a href="https://store.steampowered.com/app/2366590/Rideshare_Stimulator/">Steam</a> page that had been conspicuously missing since the game first went live for wishlisting on 30th July.</p> <p><a href="https://www.eurogamer.net/saber-rideshare-stimulator-steam-ai-disclaimer">Read more</a></p>

<img src="https://assetsio.gnwcdn.com/Pedro-Pascal-as-Joel-in-The-Last-of-Us.-He-is-holding-a-guitar-on-his-porch_8VtlBhU.jpg?width=690&amp;quality=85&amp;format=jpg&amp;auto=webp" /> <p>Matthew McConaughey has explained why he ultimately passed on playing Joel in HBO's The Last of Us &ndash; a role that, of course, went to Pedro Pascal instead.</p> <p><a href="https://www.eurogamer.net/matthew-mcconaughey-last-of-us-joel-turned-down">Read more</a></p>

<img src="https://assetsio.gnwcdn.com/Screenshot-2026-08-15-at-11.52.30.png?width=690&amp;quality=85&amp;format=jpg&amp;auto=webp" /> <p><a href="https://www.eurogamer.net/games/kingdom-hearts">Kingdom Hearts</a> creator Tetsuya Nomura has addressed the<a href="https://www.eurogamer.net/kingdom-hearts-animated-series"> newly announced Kingdom Hearts anime series</a> directly for the first time, though he's giving away very little.</p> <p><a href="https://www.eurogamer.net/kingdom-hearts-nomura-anime-series-tease-d23">Read more</a></p>

<img src="https://assetsio.gnwcdn.com/halloween-2026-game-steam-promo-art.jpg?width=690&amp;quality=85&amp;format=jpg&amp;auto=webp" /> <p>Halloween: The Game will get a digital release in New Zealand after all, despite initially being lumped in with Australia's classification refusal.</p> <p><a href="https://www.eurogamer.net/halloween-the-game-new-zealand-digital-release-update">Read more</a></p>

<img src="https://assetsio.gnwcdn.com/rideshare-saber-sacco.jpg?width=690&amp;quality=85&amp;format=jpg&amp;auto=webp" /> <p>Saber Interactive CEO Matthew Karch has apologised to Stella Sacco, the writer at the centre of the<a href="https://www.eurogamer.net/saber-ceo-sacco-chatgpt-ai-rideshare-stimulator-public-spat"> Rideshare Stimulator AI controversy</a>, after publicly saying he'd have been "happy to replace her with AI".</p> <p><a href="https://www.eurogamer.net/saber-ceo-karch-apologises-stella-sacco">Read more</a></p>

Welcome back to TechCrunch Mobility — your central hub for news and insights on the future of transportation.

本文精选了美团技术团队被 KDD 2026 收录的 8 篇论文进行分享，这些论文覆盖了推荐大模型、生成与奖励建模框架、智能体搜索、Transformer 框架、元泛化框架等技术领域。

本篇博客是一篇科普文章，由浅入深的介绍Agent评测。其中前两章系统介绍了评测是什么，以及如何建立评测体系；其中第二章是美团图灵Agent评测团队深入美团各业务团队BP总结出的实践经验，是我们在两年实践过程中逐步打磨出来的认知。

搭载美团 LongCat 2.0 的全场景 AI Agent 平台 CatPaw 正式上线，我们将模型能力从“跑得动”推向“用得好”。CatPaw 提供开箱即用的 AI 智能工作台与企业级 Agent 开发托管能力。

过去一年，我们见证了 Search Agent 能力的显著演进。在 BrowseComp 等评测上，顶尖模型准确率从最初的 30% 区间迅速攀升至 90% 以上。然而，当基准迅速饱和，其区分模型能力的价值也随之递减。

美团 LongCat 团队构建了 MineExplorer —— 首个在开放世界中做到分钟级长程任务的评测基准，系统性地评测多模态大模型在需要长程规划、并包含隐藏前置条件的任务中的真实能力。

美团 LongCat-2.0 总参数 1.6T，平均激活约 48B，为真实的 Agentic Coding 任务而生，架构上创新性引入 LongCat 稀疏注意力和 N-gram Embedding，提升长上下文处理效率与 Token 级表示能力的同时，结合动态激活进一步强化了代码理解、生成以及执行的表现。

如果你正在关注 AI 前沿，这篇内容值得收藏。2026 年，美团技术团队数十篇论文被 ACL、SIGIR、ICML、KDD 等顶会收录。我们精选 32 篇，进行了 5 大专场直播。

作为业界首个在五万卡国产算力集群上完成全流程训练与推理的万亿参数模型（总参数 1.6 T，平均激活约 48 B，动态范围 33B~56B），LongCat-2.0 从零开始预训练，原生支持 1M 超长上下文，其架构设计自始至终围绕一个核心目标：让模型在真实的 Agentic Coding 任务中，更高效、更稳定地完成代码理解、生成与执行。

ICML是机器学习领域最具影响力的国际顶级学术会议之一。大会旨在探讨机器学习未来发展所面临的关键挑战与核心问题，并通过征集和评估具有重要理论价值和实际影响的前沿研究成果，推动领域发展并引领未来研究方向。

VitaBench 2.0 是首个真实生活场景下面向长期动态用户建模的智能体评测基准，它系统性地评测大语言模型在长期、真实、动态的用户互动中个性化与主动性的能力。

5 小时用量从 63300 削减到 3800

战争、天气、仓储、水资源、浪费——五重压力正在同步收紧全球粮食供应的 「 阀门 」。 8 月 15 日，摩根大通最新研究报告显示，下一场全球粮食危机的轮廓已初步成形，食品通胀压力将持续至 2027 年

当地时间 8 月 14 日，路易吉·曼吉奥内就枪杀联合健康保险首席执行官布赖恩·汤普森（Brian Thompson）一案正式认罪。联邦法官将路易吉·曼吉奥内的量刑日期定于 12 月 18 日。 路易

女儿今年二年级，性格偏内向，最近却主动跟我说起一个同学，说她画画好看，贴纸很多，还愿意跟她换。放学路上两人会一起走一段，叽叽喳喳说个不停。我挺高兴的，因为她很少这样兴奋。 但前几天家长群里有人私聊我，

我是一个资深网文爱好者，我也写网文。我认识的一个上小学的孩子每天早上一睁眼就要听故事，吃饭也听，睡觉前也要听，谁劝和谁急，听的都是网文小白文。他父母对我说，他的孩子就像过去吸大烟的，而我像是以贩养吸的

据中新社东西问援引乌克兰媒体报道，近日一支由日本人组成的战斗分队在乌克兰武装部队内部成立，正在参与俄乌冲突，与俄军作战。此前日本防相小泉进次郎曾表示：「汲取俄乌冲突前线出现的新型作战方式等经验教训，将

去年，「东莞 17 亿元全额付款土地被无偿收回，企业称『闻所未闻’」一事引发广泛热议。近日，该事件终于迎来最新进展。 8 月 11 日，广东东莞市自然资源局发布《解除国有建设用地使用权出让合同公告》称

2026 年 8 月 12 日，番茄小说作者文黛黛在「去世」三年后突然发文，承认 2023 年的「心脏病离世」系假死，并向读者道歉。据公开报道，她此次「复活」与平台提现新规要求作者人脸识别有关，截至

8 月 5 日，有媒体报道浙江新昌县「如是书院」相关问题。当日当地成立联合调查组，开展线索核查、调查取证、溯源摸排等工作。调查组经调查后发布如下通报： 赖某明（化名如平），1968 年 3 月出生，户

看到柯洁表示大家放弃了自己的棋风，转而去背 AI 棋谱，认为围棋已死。但是为什么没有人提出背象棋 AI 棋谱，说「象棋已死」呢？

2026 年被称为「中国 FUV 元年」。继小米 YU7 完成品类启蒙后，岚图追光 S 、智界 RX 等产品接连入局。 FUV 概念最早由法拉利在 2022 年提出，本意是「法拉利多功能车」。把一台低

2026 年 7 月 1 日 -7 月 31 日，国家传染病智能监测预警前置软件统计的全国 31 个省（自治区、直辖市）及新疆生产建设兵团二级及以上医疗机构门急诊发热病例人次数波动于 41.8 万至

[图片]

韩国总统李在明周四在与幕僚的一次会议上表示，房地产是韩国面临的一个重要问题，深刻影响着社会的方方面面，而房地产泡沫已经达到不容忽视的程度。 「社会资源长期以来过度集中于房地产领域，加剧了社会两极分化和

加文·贝克透露，达里奥曾私下表示，Anthropic 公司未来可能会成为世界上唯一一家私营公司。贝克援引可靠消息来源称：「Anthropic 公司内部信心十足。我从多位信任的人那里得知，达里奥曾说过，

就几个小岛而已，真的有这么重要吗？ 如果因为北方四岛，从而让整个日本被拖入战火这真的值当吗？

物有所值，来好几次了】 8 月 7 日，胖东来创始人于东来在直播中称，许昌生活广场店将于 2026 年 12 月关闭。8 月 15 日，实拍胖东来许昌生活广场店，超市人声鼎沸人头攒动，1 至 28 号

我家的猫一到半夜就开始哇哇叫，听声音不像是发情了，因为已经绝育，很奇怪，每天下班我回去不怎么叫，就跟我屁股后面，很喜欢我陪玩，一到我睡觉后，家里的所有灯都关闭，它就开始喵喵叫了，真的很影响我的睡眠啊！

马云：你指定一个支付宝账号，获得 5 亿元花呗额度。 马化腾：你指定一个 qq 号，获得 1 亿 q 币。 你只能要其中一个，你会选择要哪个？

在无视线阻挡的情况下，人的肉眼可以看到一万米外的一栋楼吗？是否物体越高，越能在更远的地方被看到？地球是圆的对肉眼视距有影响吗（比如说我站在某个距离看楼可以看到整个楼，如果更远处还有栋相同的楼是不是一楼

别人犯蠢的时候会很烦躁。

为什么从「年少不听李宗盛，听懂已是不惑年」变成了「年少不听李宗盛，听懂已是老渣男」关于李宗盛的全变成了较差的评论

从前很小的时候就突然想到死亡，觉得人死了，意识就消失了，会陷入虚无，外界的所有都与你无关了，人生短短的几十年和宇宙比起来不值一提，就感觉十分恐慌，所以马上抽离这个问题，就没什么想法，现在我大二暑假待在

生活中，不少人都是「西红柿」的狂热爱好者，生西红柿爱吃、番茄酱爱吃、西红柿炒鸡蛋都爱吃…只要是西红柿做的都来者不拒。不得不说，这群人真的「赚」到了！ 日前，一项最新临床研究给出了一个令人意外的答案：每

[视频]