早报｜苹果或为iPad Pro加入均热板/ChatGPT与Codex将共享额度/奇瑞回应智界独立：希望团队更像新势力
📱曝苹果明年上半年将更新三款 iPad，Air 拟换 OLED 屏
🚀我国首次载人登月乘组计划由 3 人组成，2 人着陆月球
💬ChatGPT 与 Codex 将从 12 月起共享额度，Instant 聊天不扣额度
🚗多家车企公布 10 月成绩单
🧠奥特曼与三星约 50 名高管座谈，讨论 AI 未来十年的商业应用
🏭TCL 华星 t8 产线设备进场，计划明年底量产印刷 OLED
🤖AI 控制机械臂完成活鼠脑内注射，兽医负责开颅
🎓西湖大学成立世界模型研究中心，计划三年内构建千亿参数模型
🎮Claude Code 限制 Opus 5.5 复刻现有游戏，原创游戏仍可制作
🧠Claude Ultra 取消已验证账户的 5 小时限额，周额度仍保留
🏢黑石下调评级预期，预计 Anthropic 上市后或仍处「垃圾级」
💬奇瑞回应智界独立：希望团队更像新势力，明年将导入新产品
👤Grok API 将默认加入隐形文本水印，团队可按模型关闭
🚗小鹏 MONA 明年将在部分城市开独立门店
📹TBC 把神经元研究用进视频模型，称生成速度至少提高 5 倍
📈长鑫称已与头部客户协商长协，预计四季度营收继续增长
💡DeepSeek Harness 团队：多任务 Agent 的瓶颈在人的注意力
🧑💻HomeView 代码出现安防摄像头模式，可按在家、外出和睡眠设置录制
📱ChatGPT 网页和手机端支持双账号切换，聊天与记忆各自独立
💻分析师预测《GTA6》PC 版还需约 6 至 7 个月
💾高通更新骁龙 X 显卡驱动，分代改进游戏稳定性
🎵Spotify 拟明年初推出 AI 混音，让歌迷改编已授权歌曲
🤖Meta 与傅利叶造出人形机器人 Max，已着手开发第二版
🔓开源安卓助手 Mika 可在后台操作 App，主屏仍能继续用
🚗小米澎程 N70 现车新增 3 年 0 息，定制新车继续提供低息方案
☕泰国高山咖啡 Akha Ama 在东京开出第三家店
🏛️教室照明新国标实施，新增频闪与采光均匀度要求
🎬《Hidari》导演谈基努·里维斯加盟：看完木偶短片，一周后约了视频会
👀《Godzilla Minus Zero》将于 11 月 3 日在日本上映，故事接续前作两年后
📷《银魂》导演谈重制吉原炎上篇：为新版重修 200 多个镜头
曝苹果明年上半年将更新三款 iPad，Air 拟换 OLED 屏
彭博社记者 Mark Gurman 在最新一期 Power On 中透露，苹果计划在 2027 年上半年更新入门款 iPad、iPad Air 和 iPad Pro。
其中，iPad Air 将转向 OLED 屏幕，与刚更新的 iPad mini 看齐；iPad Pro 将提升性能，高端机型的重要变化可能是加入均热板，改善游戏和高强度 AI 任务中的散热。入门款 iPad 也将迎来升级。
Gurman 表示，按他目前获得的信息，若计划没有推迟，明年下半年预计不会再有重大的 iPad 更新。
我国首次载人登月乘组计划由 3 人组成，2 人着陆月球
10 月 31 日，中国载人航天工程副总设计师兼月球探测工程总设计师王翔在第七届载人航天学术大会上介绍，我国首次载人月球探测任务乘组计划由 3 人组成，其中 2 人着陆月球表面。
据《环球时报》记者现场报道，我国计划在 2030 年前实现中国人登陆月球，采用「两次发射，一次登月」的环月轨道对接方案：先用一枚长征十号运载火箭将着陆器送往月球轨道，再发射载人飞船，与着陆器交会对接。航天员进入着陆器后，与载人飞船分离并着陆月面，开展科学探测。
月面任务完成后，航天员将乘着陆器上升，在月球轨道与载人飞船再次交会对接，随后乘载人飞船返回地球。
ChatGPT 与 Codex 将从 12 月起共享额度，Instant 聊天不扣额度
OpenAI 更新帮助中心说明，计划从 12 月起，将 Plus、Pro 和 Business 套餐在 ChatGPT 与 Codex 中的使用额度合并计算。这项调整与 ChatGPT Work 的智能体能力并入 ChatGPT 有关。
两项服务将按 token 统一衡量用量，消耗取决于所选模型和任务复杂度：更强的模型、复杂任务通常会消耗更多额度，Instant 即时聊天不占用这部分额度。
合并后的剩余额度与重置时间将显示在「设置 → 使用量」中。用完后，用户可等待重置、升级套餐，或购买积分继续使用符合条件的功能。
🔗 相关阅读：ChatGPT要和Codex共用额度，免费重置时代结束了
多家车企公布 10 月成绩单
多家车企公布 10 月销量或交付数据。零跑连续四个月交付超过 10 万辆，蔚来三品牌合计交付 34,246 辆；上汽集团整车销量同比下降，新能源汽车与海外业务继续增长。
- 零跑汽车：全球交付 106,127 辆，连续四个月月交付超过 10 万辆，全球累计交付突破 190 万辆。公司称，全新 Lafa5 将在 11 月广州车展亮相。
- 蔚来公司：交付 34,246 辆，其中蔚来品牌 18,125 辆、乐道 8,383 辆、萤火虫 7,738 辆。今年前十个月累计交付 334,547 辆，同比增长 38.5%。
- 理想汽车：交付 29,025 辆，其中新款 MEGA 超过 3000 辆。理想公布，11 月将开始推送 OTA 9.2，升级 MACH VLA 2.5 与座舱 Agent 功能。
- 上汽集团：据中国基金报、格隆汇转述的产销公告，10 月整车销量为 43.00 万辆，同比下降 5.28%；其中新能源汽车销量 25.86 万辆，同比增长 25.12%，出口及海外基地销量 16.25 万辆，同比增长 67.28%。
- 比亚迪：新能源汽车销量约 49.02 万辆，同比增长 10.97%；其中纯电动车约 30.57 万辆，插混约 17.74 万辆，出口新能源汽车约 17.02 万辆。
- 吉利汽车：总销量 310,183 辆，同比增长约 1%；纯电动车 140,758 辆，插混 74,238 辆，出口 107,153 辆。
- 长城汽车：销量 120,374 辆，同比下降 15.87%；其中新能源汽车 49,499 辆，海外销量 62,862 辆。
- 小米汽车：交付超过 50,000 辆。
- 鸿蒙智行：交付 33,537 辆，累计交付超过 160 万辆。
- 极氪：品牌交付 32,483 辆，同比增长 51.6%。
- 深蓝汽车：全球销量 32,602 辆，环比增长 8.01%。
- 长安启源：全球销量 63,648 辆，环比增长 22.24%；启源 Q06 上市首个完整月销量 10,152 辆。
- 昊铂埃安 BU：销量 42,357 辆，同比增长 56.79%。
- 岚图汽车：交付 13,030 辆，今年前十个月累计交付同比增长 13%。
- 北汽新能源：销量 32,502 辆，同比增长 6.42%。
- 赛力斯集团：汽车销量 27,037 辆，同比下降 50.29%；其中新能源汽车销量 24,921 辆。
奥特曼与三星约 50 名高管座谈，讨论 AI 未来十年的商业应用
10 月 31 日，OpenAI CEO 萨姆·奥特曼受邀来到韩国龙仁的三星人力开发院，与三星电子、金融、生物医药等业务约 50 名高管举行座谈。三星向韩国媒体介绍，本次主题为「AI 将改变的未来十年：企业的选择」，讨论技术演进、产业应用和新市场机会。
据《每日经济新闻》报道，座谈前，高管还接受了生成式 AI 实操培训，学习整理分散的经营资料、建立个人知识库，并将其用于决策。三星自 6 月起推进覆盖研发、生产、营销和支持部门的 AI 转型，计划将高管培训扩展至普通员工，于 2027 年 2 月完成全员 AI 培训。
双方此前已在 AI 基础设施上展开合作：三星电子与 OpenAI 去年 10 月签署意向书，推进 Stargate 项目的存储芯片供应；三星 SDS 参与 AI 数据中心及企业 AI 服务合作，三星物产和三星重工则推进海上浮式数据中心的共同开发。
TCL 华星 t8 产线设备进场，计划明年底量产印刷 OLED
TCL 华星 10 月 31 日在广州举行 t8 项目设备搬入仪式，首台核心工艺设备进场，产线由厂房建设转入设备安装与调试阶段。这条 G8.6 印刷 OLED 产线计划于 2027 年第四季度量产。
据 TCL 华星项目新闻稿，t8 总投资约 295 亿元，设计月加工玻璃基板能力约 2.25 万片，主要面向平板、笔记本电脑和显示器，也可延伸至柔性、折叠显示屏。接下来将开展产线安装、动力系统联调和工艺验证。
印刷 OLED 通过喷墨打印，将有机发光材料按需沉积到玻璃基板上。TCL 华星称，相较真空蒸镀，这一路线能提高材料利用率；公司此前已在武汉 t12 的 G5.5 代线上实现印刷 OLED 量产。
AI 控制机械臂完成活鼠脑内注射，兽医负责开颅
生物技术公司 Exin Therapeutics 将 AI 智能体接入实验设备，让它选择基因疗法并控制机械臂，为一只癫痫模型小鼠完成脑内注射。创始人 Gabriel Ocana Santero 向 WIRED 介绍，兽医先完成开颅，AI 再从表格中的 19 个选项选择疗法、脑区和注射坐标，并控制注射量与速度。
团队最初测试 Claude Opus 5.5，它完成了实验规划，但拒绝将针头伸入小鼠大脑；换用 Claude Sonnet 5 后才继续操作。团队要求注射方案先由人类批准，并设置可切断设备电机的紧急按钮，同时禁止智能体改动设备底层代码。
过程中，设备坐标系统出现故障。智能体识别出注射位置计算有误，却因上述限制无法修改代码，团队另用一个 Opus 5 智能体修复。整场实验耗时略超一小时，常规同类操作约需 45 分钟。Santero 称小鼠恢复良好，团队正等待基因疗法起效，再评估对癫痫发作的影响。
🔗 相关阅读：Claude 给活鼠做了场脑部手术，Opus 5.5 却不肯动手
西湖大学成立世界模型研究中心，计划三年内构建千亿参数模型
10 月 31 日，西湖世界模型研究中心在西湖大学成立，由工学院助理教授、博士生导师于开丞担任主任。这个校级交叉科研平台将以构建通用世界模型为使命，于开丞在接受潮新闻专访时表示，规划在三年内做到 1000 亿参数规模。
他用拿杯子解释世界模型的思路：机器在伸手之前，先预演从正面还是侧面拿、会不会碰到其他物体，再选择把握更大的路径。模型预测的是世界的下一个状态，以便在行动前发现问题；机器人操作、自动驾驶等是他举出的应用场景。
于开丞同时强调，参数量不等于能力，训练数据、架构、训练方法与任务聚焦度同样关键，要「先做对，才能做大」。中心计划联通西湖大学各学院开展跨学科合作，并面向全球招引人才。
Claude Code 限制 Opus 5.5 复刻现有游戏，原创游戏仍可制作
Anthropic 开发者 Addy Osmani 确认，团队在 Claude Code 使用 Opus 5.5 时的提示词中增加了两句话，引导模型避开复刻他人的某款特定游戏。其依据是，使用 Claude 侵犯知识产权违反公司的使用政策。
Osmani 表示，原创游戏、某一类型的游戏，以及采用用户自有设计和素材制作的游戏，都应继续正常工作。他同时称团队会持续迭代规则，若模型拒绝处理用户自己的内容，可以把案例发给他。
这次说明回应了用户对 Claude Code 新规则的质疑。相关用户展示的规则写明，即使把现有游戏改名，或声称作品是原创，也不应直接复刻。
Claude Ultra 取消已验证账户的 5 小时限额，周额度仍保留
Anthropic 调整 Claude 订阅套餐，将原 Max 20x 更名为 Ultra，并取消已验证 Ultra 账户的 5 小时用量限制。原有订阅者自动转入新套餐，网页订阅价格仍为每月 200 美元。
据官方帮助文档，Ultra 每周用量为 Pro 的 10 倍，所有模型共用周额度，Fable 另有独立的每周限制。用户可以在周额度内自行安排使用节奏。
账户验证由系统依据账户和付款历史自动完成，尚未验证的账户暂时保留限制，验证后自动解除；结账时的手机号验证与这项审核分开。在「设置 — 用量」中看不到当前会话进度条，即表示账户已通过验证。
黑石下调评级预期，预计 Anthropic 上市后或仍处「垃圾级」
据《金融时报》报道，黑石向投资者表示，预计 Anthropic 在 IPO 后将获得双 B 区间的信用评级，处于投机级、俗称「垃圾级」的较高档位。黑石此前曾预计，这家公司会在 2027 年之前获得投资级评级。
TechCrunch 向 Anthropic 与黑石求证，双方均拒绝置评。黑石是 Anthropic 芯片融资交易的主要投资方，其判断变化涉及这家 AI 公司后续的融资条件。
若评级停留在投机级，Anthropic 为持续扩大的算力支出筹资将更困难，成本也可能更高。
奇瑞回应智界独立：希望团队更像新势力，明年将导入新产品
在 10 月 31 日的三季度媒体沟通交流会上，奇瑞汽车执行副总裁戚士龙回应智界独立运营时表示，希望智界团队更像新势力企业，避免做成「大一号」的传统车企。
他表示，奇瑞将从组织规模化、产品高端化及品牌逐步优化等方面，提升智界独立运营能力。据界面新闻报道，智界明年还会导入新产品，海外业务仍由奇瑞集团独立运营。
Grok API 将默认加入隐形文本水印，团队可按模型关闭
Grok 开发文档公布，11 月 7 日起，Grok 4.7、4.6、4.5、4.3 及三款 Grok 4.20 模型的 API 文本输出将默认加入隐形水印，覆盖全球及区域端点。团队管理员可在控制台按模型关闭水印。
水印在生成过程中微调 Token 的选择，不向回复添加隐藏字符或标记，普通文本、结构化输出和工具调用参数都在覆盖范围内。开发者无需修改请求，可通过响应中的 x-watermark 标头查看该次输出是否带水印。
官方称，此举是为遵守欧盟《人工智能法案》。检测采用统计方法，文本越长，信号越强；检测 API 需另行申请使用权限。
小鹏 MONA 明年将在部分城市开独立门店
据界面新闻报道，小鹏集团总裁王凤英在巴黎车展期间表示，从 2027 年开始，部分城市将设立 MONA 独立门店，未来也「不排除 MONA 品牌独立的可能性」。
渠道端已经出现筹备动作。北京一家小鹏门店销售人员向界面新闻表示，小鹏和 MONA 预计会分开销售；新出行 CEO 贺磊则发现，广州海珠区一家商场出现待入驻的「小鹏 MONA 销售中心」。
产品规划也在调整。界面新闻获悉，小鹏近期把原有四条产品线合并为 G、D 两条，MONA 继续独享 D 产品线。贺磊称，明年仅一部分渠道会成为 MONA 独立渠道，现有小鹏门店仍会销售 MONA 车型。
TBC 把神经元研究用进视频模型，称生成速度至少提高 5 倍
The Biological Computing Co.（TBC）发布图生视频模型 Mischief 1。公司 CEO Alex Ksendzovsky 表示，团队把持续四年的活神经元研究转化成了可供外部使用的视频模型，生成速度至少达到基础模型的 5 倍。
他在回复中确认，Mischief 1 基于阿里开源模型 Wan 2.2，下一代模型也已在开发。研究中使用的神经元留在实验室，视频模型运行在普通芯片上。
Mischief 1 已在 fal 和 OpenRouter 提供服务，支持以图片生成短视频。Ksendzovsky 还预告，模型将进入 AWS Marketplace。
长鑫称已与头部客户协商长协，预计四季度营收继续增长
长鑫科技在 10 月 31 日披露的投资者关系活动记录中回应长期供货协议进展，称已与多家头部客户协商长协，以锁定中长期确定性需求。
据中新经纬对记录表的报道，公司认为 AI 应用正在推动结构性需求增长，服务器客户第四季度需求仍强劲，DRAM 供不应求的态势将延续，预计第四季度营收继续保持增长。
公司还解释，长鑫新桥与长鑫集电是第三季度营收和利润增长的主要驱动力，但集团在两家子公司的权益比例相对较低，因此归母层面的增长幅度受到影响。公司已在今年 8 月增加对长鑫集电的注资、提高权益比例。
💡 DeepSeek Harness 团队：多任务 Agent 的瓶颈在人的注意力
DeepSeek Harness 团队 10 月 31 日发表 Super-agent 架构探索长文，指出随着模型越来越能独立完成单个任务，多任务协作的瓶颈正在转向人的注意力。团队观察到，一个人通常只能跟进 3 至 5 个活跃会话，因此希望由系统接管会话管理，让用户专注于决策与验收。
团队的方案是将职责分成两层：常驻的 Router 负责跨任务协调与对外交互，上下文隔离的 Worker 并发执行具体任务。同一 GitHub PR 或群聊话题的后续消息可直接送达对应 Worker，减少反复路由；耗时工具在后台运行，前台继续接受新指令。
文章介绍，内部实例已能在 IM 与 GitHub 之间连续推进任务，并在群聊无人 @ 的情况下识别缺陷、尝试修复后提交 PR。为维持长期运行，团队还把计算、会话状态与执行环境分开，并用独立的 Watchdog 检查任务是否取得实质进展。
团队强调，这是一套工程架构探索，部分设计以 dsh-superagent-prototype 原型提供，尚非正式产品。其初步实验采用合成负载与简化环境；更大组织中的上下文隔离、跨 Agent 协作，以及提示词注入与凭据泄露风险，仍是团队列出的开放问题。
🔗 相关阅读：DeepSeek 重磅发布全新 Agent 架构：AI 开始指挥 AI 干活
HomeView 代码出现安防摄像头模式，可按在家、外出和睡眠设置录制
苹果 HomeView 智能家居中枢可能增加安防摄像头功能。The Verge 报道，Reddit 用户 pdfu 在 HomeView 27.2 代码中发现相关字符串，其中一个设置页面写明，HomeView 可通过 HomeKit Secure Video 用作安防摄像头。
代码描述的功能包括移动侦测、实时视频和录制。视频仅供家庭成员查看；有人查看画面或设备开始录制时，设备会亮起指示灯并播放声音，Secure Video 设置改变也会通知家庭成员。
用户还可根据在家、外出或睡眠状态，分别设置侦测、串流和录制内容。上述信息来自代码线索，苹果尚未宣布这项功能。
ChatGPT 网页和手机端支持双账号切换，聊天与记忆各自独立
OpenAI 已在 ChatGPT 网页端、iOS 和 Android 推出账号切换功能。用户可同时登录最多两个账号，从账号菜单或设置中切换，无需反复退出再登录；同一时间只有一个账号处于活跃状态。移动端功能正逐步开放。
OpenAI 的帮助文档说明，切换后显示的是对应账号的聊天、记忆、文件和设置，工作区也归属于各自账号。两个账号不会合并对话、订阅、账单或工作区，付费权益仍由各账号自己的订阅决定。
用户可以拥有两个以上的 ChatGPT 账号，但同一设备最多同时保持两个账号登录；已登录两个时，需先退出其中一个才能添加其他账号。移动端还可依照各账号设置和设备权限接收通知，点击另一账号的通知时会切换并打开对应内容。
分析师预测《GTA6》PC 版还需约 6 至 7 个月
Wedbush 分析师迈克尔·帕赫特在 BBC Business Today 节目中预测，《GTA6》大约还需 6 至 7 个月登陆 PC。相关采访片段于 10 月 28 日在社交平台流传。
Rockstar 为主机版公布的发售日是 2026 年 11 月 19 日，首发平台为 PS5 和 Xbox Series X|S。帕赫特给出的是自己的时间判断，Rockstar 尚未公布 PC 版发售日。
高通更新骁龙 X 显卡驱动，分代改进游戏稳定性
高通发布适用于骁龙 X 系列 Windows PC 的 Adreno 2026.10.1 驱动，更新引入同时支持初代骁龙 X 与骁龙 X2 的安装框架，两代平台各有独立的驱动版本和改进清单。
初代骁龙 X 的更新涉及《永劫无间》《荒野大镖客：救赎 2》和《机械战警：暴戾都市》的稳定性或画质；骁龙 X2 的改进则包括《极限竞速：地平线 6》《侠盗猎车手 5 传承版》《NBA 2K26》等。两代平台均列出 Luminar Neo 等应用的稳定性和性能改进。
官方说明还保留一项 Windows 11 已知问题：在 Build 26100.3915 及更高版本中，插拔外接显示器或切换显示模式、分辨率，可能导致《地铁：离去》《文明 VI》《求生之路 2》崩溃或卡死。
Spotify 拟明年初推出 AI 混音，让歌迷改编已授权歌曲
据《金融时报》报道，Spotify 正争取在 2027 年第一季度推出 AI 混音与翻唱服务，可能在 2 月格莱美奖前后亮相。付费订阅用户可用它改编已获授权的歌曲，例如把民谣改成舞曲，或制作歌手翻唱其他作品的版本。
Spotify 已与环球音乐、独立厂牌授权机构 Merlin 和 Believe 达成相关协议，仍在与索尼音乐、华纳音乐谈判。即使唱片公司签约，音乐人也可以自行决定是否参与。
按目前方案，粉丝混音产生的版税会流向原作音乐人。Spotify 还在开发配套机制，让这些作品进入播放列表，并把播放量计入原曲榜单成绩。消息人士称，上线时间和定价仍在讨论中。
Meta 与傅利叶造出人形机器人 Max，已着手开发第二版
据 The Information 报道，Meta 与上海机器人公司傅利叶用约三个月造出人形机器人原型 Max，并于 9 月开始开发第二版 Max 2。Meta 发言人回应称，与傅利叶的工作聚焦内部研究和实验，公司没有出售这些原型的计划。
Max 约高 1.5 米、重 48 公斤，搭载英伟达 Jetson Thor，胸部内置扬声器，面部屏幕可以模拟表情，五指手能抓起小瓶子和玩具。Meta 基于傅利叶前三代机器人的硬件与经验，调整了机身比例、头部、显示屏和摄像头，并增加双手等部件。
报道称，Meta 机器人产品负责人曾在傅利叶上海工厂驻留近两个月，原型于 8 月底在内部峰会上演示。第二版将重做躯干与双手，加入投影仪，以及帮助上下楼梯的传感器；Meta 内部将人形机器人视作其 Muse AI Agent 执行家务等任务的实体载体。
开源安卓助手 Mika 可在后台操作 App，主屏仍能继续用
独立开发者王赞开源了安卓 AI 助手 Mika。用户用自然语言交代任务后，它可以在独立的后台屏幕上打开 App、点击、输入和滚动，主屏幕仍留给用户使用。
项目展示了后台翻阅微信群聊并总结、向文件传输助手发消息，以及在购物 App 中筛选商品的用法。Mika 通过读取界面操作应用，不要求每个 App 额外提供接口；登录、验证码和付款等环节会交还用户。
跨 App 操作要求 Android 12 或更新系统，首次使用需完成无线调试配对。项目推荐 Android 14 及以上版本，虚拟屏幕可拥有独立输入焦点，减少后台任务对主屏输入的干扰。用户可以接入自己的模型 API，也可通过支持的模型服务账号登录；项目采用 Apache 2.0 许可。
小米澎程 N70 现车新增 3 年 0 息，定制新车继续提供低息方案
小米汽车公布 11 月购车权益，澎程 N70 的少量现车新增「3 年 0 息」金融方案，定制新车则提供「5 年低息」或「限时轻松还」。两类权益均适用于 11 月 1 日至 11 月 30 日 24 点前下定的订单。
- 澎程 N70 全系少量现车：首付 7.99 万元起，可在「3 年 0 息」「5 年分期 3 年 0 息」「6000 元保险补贴」中三选一。
- 澎程全系定制新车：低息方案首付 4.99 万元起，5 年月供 2932 元起，另有「限时轻松还」可选。
- SU7、YU7 与 SU7 Ultra：SU7 全系少量现车、YU7 指定版本少量现车可选「3 年 0 息」「5 年分期 3 年 0 息」或「6000 元保险补贴」；SU7 Ultra 提供限时「5 年 0 息」。
泰国高山咖啡 Akha Ama 在东京开出第三家店
来自泰国北部山区的 Akha Ama Coffee 在东京目黑区开出日本第三家咖啡店。据日经亚洲 10 月 31 日报道，品牌希望以高品质、可持续种植的泰国咖啡豆，进一步打开日本市场。
创始人 Lee Ayu Chuepa 来自阿卡族，「Akha Ama」在族语中意为「阿卡母亲」。品牌创办于 2010 年，从泰北农户处采购咖啡豆，其中包括创始人的母亲和姐姐，再加工、烘焙，供应清迈和东京的门店。农户在高海拔种植阿拉比卡咖啡，并与水果、坚果等作物混种，采用农林复合种植方式。
创始人表示，品牌重视品质和产地故事，正在通过减少化肥及化学品使用来提高品质。目前约九成咖啡豆在泰国销售，其余一成出口日本；品牌于 2020 年在东京神乐坂开出日本首店，2024 年再开第二家。
教室照明新国标实施，新增频闪与采光均匀度要求
11 月 1 日，强制性国家标准《中小学校教室采光和照明卫生标准》正式实施，教室采光和灯具照明的评价指标进一步细化。
据央视新闻报道，新标准增加采光均匀度、冬至日满窗日照时间、玻璃颜色透射指数和可见光透射比等要求，也将闪变指数、频闪效应可视度纳入指标。采光方向、采光系数、眩光控制、照明色温、显色指数及测量方法均有修订。
报道采访的首都医科大学附属北京同仁医院眼科医生田磊介绍，光源看起来稳定，视觉系统仍可能不断适应光线变化，长期处在这类频闪环境中容易出现眼睛疲劳。
《Hidari》导演谈基努·里维斯加盟：看完木偶短片，一周后约了视频会
日本定格动画长片《Hidari》的导演川村真司在东京国际电影节透露，基努·里维斯看过项目先导短片后，团队约一周便收到回复，安排双方视频通话。里维斯为主角配音的消息已于今年 5 月在戛纳公布；这次导演进一步讲述了邀请过程和长片筹备进展。
川村表示，作品受黑泽明电影影响，制作主角木偶时，他就希望将三船敏郎与里维斯的形象结合，再让角色显得更年长。《Hidari》采用木雕、定格动画与时代剧的组合，团队将它定位为动作片，并曾带着手工雕刻的木偶向海外发行商展示项目。
据 Variety 现场报道，团队约三年前自筹资金并借助众筹拍摄概念短片，上传 YouTube 后累计获得约 530 万次观看。长片目前仍在寻求补助和海外发行合作；制作人此前透露，一家美国公司曾提出承担全部预算并换取全部版权，相关谈判已陷入停滞。
《Godzilla Minus Zero》将于 11 月 3 日在日本上映，故事接续前作两年后
山崎贵执导的《哥斯拉 -1.0》续作《Godzilla Minus Zero》将于 11 月 3 日在日本上映。神木隆之介、滨边美波在东京国际电影节活动上透露，新片延续敷岛浩一家人的故事，时间设在前作约两年后，刚恢复日常生活的家庭再度面临哥斯拉的威胁。
两位主演继续饰演敷岛浩一与典子，永谷咲也回归饰演明子。神木隆之介表示，上一部中主要以哭泣或单词表达的明子，这次有了完整对白。
据 Variety 现场报道，主创刚从洛杉矶首映活动返回日本，观众曾在红毯高喊导演的名字。新片也是首部获得「Filmed for IMAX」认证的日本电影。
《银魂》导演谈重制吉原炎上篇：为新版重修 200 多个镜头
《银魂》动画电影《吉原炎上》在意大利卢卡漫画游戏展亮相。导演安藤尚也与制片人在 Variety 采访中透露，团队在今年 2 月日本上映的版本之后，借影碟发行的机会重新打磨了 200 多个镜头。
这部电影重访原作中的人气篇章。安藤尚也表示，为了让动作场面适合大银幕，团队采用 CinemaScope 超宽画幅，并利用 3D 技术增加摄影机运动与构图变化，也制作了大量新音乐。相比电视动画，电影还真正呈现了吉原燃烧的场面。
谈到《银魂》的幽默，他说团队安排了一位导演主要负责喜剧部分；他们希望让角色自然地互相开玩笑，让观众听见并感受到趣味。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
quota/Chiri responds to intellectual independence: I hope the team is more like a new force.
The iPad, Air to replace OLED screen will be updated in the first half of next year
The first flight plan for our country was made up of 3 people, 2 people landed on the moon.
ChatGPT and Cordex will share quota, Instant chat without quota
The October report card was released by many cars.
Otmann talks with 50 senior executives on the business applications of AI for the next 10 years.
🏭TCL China Star t8 Production Lines Entry, scheduled for mass printing OLED by the end of next year
Control the arm of the machine and the inside of the brain.
A World Modelling Research Centre was established at the University of Twissy Lake, and it is planned to build a multibillion-parameter model within three years
🎮Claude Code limits the Opus 5.5 to the existing game, which can still be created
🧠Claude Ultra removes the 5-hour limit of the certified account, and week quota remains
The black stone is expected to downgrade, and Anthropic is expected to be on the market or still in the trash class.
💬 奇 奇 奇 奇 奇 奇 回应 回应
👤Grok API will by default add invisible text watermarks, and the team can close by model
Peng, MONA will open an independent store in some cities next year.
📹TBC uses neuron research into video models to say production speed is at least 5 times higher
He claims to have consulted with the head client and expects to continue to grow in the fourth quarter.
DeepSeek Harness Team: Multitasking Agent's Bottlenecks are in human attention
🧑💻HomeView code displays a security camera mode that allows recording at home, outside and sleep settings
📱ChatGPT web page and cell phone support double account switching, chat and memory separate
The analyst predicts that the GTA6 version of PC will take about six to seven months.
X GPU drive to upgrade game stability in generations
🎵Spotify plans to launch an AI mix early next year, so that fans can adapt their authorized songs
Max, the second edition has been developed.
Mika, an assistant at the Open Source Andre, can operate App on the backstage and the main screen is still working.
The new car will continue to offer a low-interest programme.
Akha Ama opened a third shop in Tokyo.
New national sign implementation for classroom lighting with new frequency-smoke and light-simulation requirements
Hidari's director talked about Keanu Reves' union: after watching the puppet film, he had a video conference a week later.
"Godzilla Minus Zero" will be released in Japan on November 3rd.
The director of Silver Spirit talks about the re-programming of the Quibaritis: 200 more shot for the new edition
An iPad, Air to replace OLED screen will be updated in the first half of next year
In a recent issue of Power On, Bloomberg correspondent Mark Gurman revealed that Apple plans to update the introductions iPad, iPad Air and iPad Pro in the first half of 2027.
Of these, iPad Air will turn to the OLED screen, which is aligned with the newly updated iPad Mini; iPad Pro will enhance performance, and an important change in the high-end machine type may be to join the average board, improve the heat spread in the game and the high-intensity AI task. The introduction iPad will also be upgraded.
According to Gurman, there are no major iPad updates expected in the second half of next year, if the plan is not postponed, according to the information currently available to him.
Our first ride to the moon was made up of 3 people, 2 people landed on the moon.
On 31 October, Wang Shang, Deputy Chief Designer of Manned Space Engineering and General Designer of Moon Exploration Engineering of China, presented at the Seventh Human Space Academic Congress, the country ' s first manned Moon mission crew consisting of three people, two of whom landed on the surface of the Moon.
According to live reports from Globe Times journalists, China plans to land on the Moon by 2030, using the "Two Launches, One Moon" orbit docking programme: The lander was first delivered to the Moon orbit by a Long March 10 carrier rocket, then the manned spacecraft was launched to meet the lander. When the astronauts entered the lander, they separated from the manned spacecraft and landed on the moon surface for scientific detection.
Upon completion of the moon-front mission, the astronaut will rise by a lander, meet again in the moon orbit with the manned spacecraft and then return to Earth by the manned ship.
ChatGPT and Codex will share quota, Instant chat without quota
Open AI Update Help Centre Note that it is planned to combine the Plus, Pro and Business packages in ChatGPT with Codex using quota from December. This adjustment relates to the incorporation of ChatGPT Work 's AI agent capability into ChatGPT.
The two services will be measured as token and consumption will depend on the complexity of the selected model and task: a stronger model, a more complex task will typically consume more quota, and an instant instant chat will not occupy this part of quota.
The remaining quota and the replacement time after the merger will be shown in the Setup Usage. Upon completion, the user may wait for the replacement, upgrade or purchase of credits to continue to use the eligible functionality.
Read about it: ChatGPT is sharing quota with Codex. The free replacement era is over.
A number of car companies announced their October report card.
Many cars publish data on sales or deliveries in October. Zero runs delivered over 100,000 vehicles for four months in a row, bringing the total of three brands to 34,246; the total sales of the upper-air group dropped on a year-on-year basis, and new energy vehicles continued to grow in line with overseas operations.
- Zero-run vehicles: 106,127 global deliveries, over 100,000 for four consecutive months, and a cumulative global delivery of 1.9 million. According to the company, the brand-new Lafa5 will be on display in Guangzhou in November.
- Ulley Company: 34,246 delivered, of which 18,125 were branded, 8,383 were bands, and 7,738 were fireflies. The cumulative delivery in the first 10 months of the year was 334,547, an increase of 38.5 per cent over the previous year.
- Ideal cars: 29,025 delivered, of which the new MEGA exceeded 3000. Ideas release, November will begin to transfer OTA 9.2, upgrade MCH VLA 2.5 and cabin Agent.
- Upper Volkswagen Group: According to the China Foundation, the sale of 430 million cars as a whole in October, which is a decrease of 5.28 per cent over the same period, as reported in the Grunt Remittal, 2586 million new energy vehicles, an increase of 25.12 per cent over the same period, and 16.25 million exports and overseas bases, an increase of 67.28 per cent over the same period.
- Biadi: The number of new energy vehicles is approximately 49.02 million, an increase of 10.97 per cent over the same period; of these, some 30.57 million are pure electric; about 17.74 million are plugged and about 17.02 million are exported.
- Jiley vehicles: total sales 310,183, an increase of approximately 1 per cent over the same period; pure electric vehicles 140,758, mixed 74,238 and exported 107,153.
- Grand Wall vehicles: 120,374, a decrease of 15.87 per cent over the same period; of these, 49,499 are new energy vehicles and 62,862 are overseas.
- Mi cars: delivered over 50,000.
- Hong Mon: Delivery of 33,537 units, cumulative delivery of over 1.6 million.
- Polar krypton: brand delivery 32,483, an increase of 51.6 per cent over the previous year.
- Deep blue cars: global sales 32,602, with an increase of 8.01 per cent.
- Long Agent: 63,648 worldwide, 22.24 per cent ring growth; first full monthly sales of 10,152 at source Q06.
- 昊铂an BU: sales of 42,357 vehicles, 56.79 per cent over the same period.
- Cars: 13,030 were delivered, with a cumulative increase of 13 per cent over the previous 10 months.
- North gas new energy: sales 32,502 vehicles, an increase of 6.42 per cent over the same period.
- The Cyrus Group: the number of cars sold was 27,037 - a decrease of 50.29 per cent over the same period; of these, 24,921 were new energy vehicles.
Otman talks with about 50 senior executives about AI's business applications for the next 10 years.
On October 31, Open AI CEO Sam Otman was invited to the Korea Dragon-in-Sung Sam Manpower Development Institute to talk to approximately 50 executives in Samsung electronics, finance, biomedicine, etc. Samsung introduced the Korean media on the theme "AI's Ten Years to Change: Business Choice" to discuss technology evolution, industrial applications and new market opportunities.
According to the Daily Economic News, prior to the talks, executives were also trained in generating AI hands-on training to collate decentralized business information, build a personal knowledge base and use it for decision-making. Since June, Samsung has advanced the AI transition covering R & D, production, marketing and support sectors. It is planned to extend senior management training to general staff and complete full-time AI training in February 2027.
The two parties had previously cooperated on AI infrastructure: Samsung Electronics and Open AI signed a letter of intent last October to advance the supply of storage chips for the Stargate project; Samsung SDS was involved in AI data centre and enterprise AI service cooperation, while Samsung and Samsung heavy works advanced the co-development of the offshore floating data centre.
TCL Chinastar t8 Production Lines in progress, scheduled for mass printing at the end of next year
TCL Wah Xin held a t8 project equipment migration ceremony in Guangzhou on 31 October, with the first core process equipment entering the field and the production line moving from plant construction to installation and commissioning phase. This G8.6 Printing OLED Production Line is planned for production in the fourth quarter of 2027.
According to the TCL China Star project press release, t8 invested about $29.5 billion in the design of a monthly glass base plate capacity of approximately 225,000 pieces, which is mainly aimed at tablets, laptops and monitors and can be extended to flexible, folded displays. This will be followed by the installation of production lines, power system alignment and process validation.
OLED is printed in ink-jet-printed, and organic light materials are deposited on the glass base plate as required. TCL Wah Sing claims that this route can increase material utilization compared to vacuum varnish; the company has previously achieved OLED mass production in Wuhan t12 on G5.5.
AI, control the arm of the machine for internal injection of the live mouse. Veterinary's in charge of the skull.
The biotechnology company Exin Therapeutics will connect AI AI agent to the experimental equipment to allow it to select gene therapy and control the arm of the machine, and complete a brain injection for an epilepsy model mouse. The founder, Gabriel Ocana Santaro, introduced WIRED to WIRED, where the veterinarian first completed his skull, AI then selected the therapy, brain section and injection coordinates from the 19 options in the table, and controlled the injection and speed.
The team initially tested Claude Opus 5.5, which completed the experimental planning but refused to reach the needle into the mouse's brain; it did not continue until Claude Sonet 5 was replaced. The team requested that the injection programme be first approved by humans and that an emergency button be installed to cut off the equipment ' s electric power, while AI agent was prohibited from changing the equipment ' s bottom code.
In the process, the equipment coordinates were malfunctioning. AI agent identified an error in the calculation of the injection position, but was unable to change the code because of the above-mentioned restrictions, and the team repaired it with an additional Opus 5 AI agent. The entire experiment took just over an hour, and the normal type of operation took approximately 45 minutes. Santero claims that the mice are well recovered and the team is waiting for the gene therapy to be effective before assessing the impact on epilepsy.
🔗 Read about it: Claude had a brain surgery for the living mouse, Opus 5.5, but he wouldn't move. Hands.
The University of West Lake established the World Modelling Research Centre, which plans to build a trillion-billion-parameter model within three years
On 31 October, the West Lake World Modelling Research Centre was established at the University of West Lake and was headed by an Assistant Professor at the College of Engineering and a Ph.D.-student mentor at Kenzo. This school-level cross-scientific platform, with its mission of building a universal world model, stated during a wave news interview that it was planned to be 100 billion parameters over three years.
He used the idea of taking a cup to explain the world model: before reaching out, the machine previews whether to take it from the front or from the side, whether to touch other objects, and chooses to take a larger path. The model predicts the next state of the world in order to detect problems before action is taken; robotics, autopilots, etc., are the applications he has cited.
At the same time, it was stressed that parameter count was not equal to capacity, that training data, architecture, training methods were as crucial as the focus of the mission, and that it was “to do the right thing before doing it”. The Centre plans to establish interdisciplinary cooperation with the universities of the University of Lake West and to target the global recruitment of talent.
Claude Code limits Opus 5.5 to existing games, originals can still be made
Anthropic Developer Addy Osmani confirms that the team added two sentences to prompt when Claude Code uses Opus 5.5 to steer the model away from a particular game that regraves others. It is based on the fact that the use of Claude ' s infringement of intellectual property violates the company ' s use policy.
Osmani states that original games, some type of games, and games made using user-owned designs and asset should continue to work normally. He also stated that the team would continue the iterative rule and that if the model refused to process the user ' s own content, the case could be sent to him.
This note responds to user questions about Claude Code's new rules. The rules displayed by the users concerned state that even if the existing game is renamed or the work is claimed to be original, it should not be directly rewritten.
Claude Ultra removes the 5-hour limit of the certified account, and week quota remains
Anthropic adjusts Claude Subscriptions, renames Max 20x to Ultra and removes the 5-hour limit on the certified Ultra account. The original subscriber automatically moved to the new set, and the cost of web-page subscriptions remained at $200 per month.
According to the official help file, Ultra uses 10 times as much as Pro per week, all models share quota, Fable has separate weekly limits. Users can arrange their own rhythms within quota.
The verification of the accounts is done automatically by the system based on the history of the accounts and payments, while those accounts that have not yet been certified are temporarily retained and automatically released after validation; the cell phone number at the time of closure is separated from this audit. The current session progress bar cannot be seen in Settings - Usage, which means that the account is verified.
Blackstone downgrades are expected, and Anthropic is expected to be on the market or still in the trash class.
According to the Financial Times, black stones indicate to investors that Anthropic is expected to receive credit ratings between double B after IPO, at a higher level of speculation, commonly referred to as “waste”. Blackstone had previously predicted that the company would obtain an investment rating by 2027.
TechCrunch sought evidence from Anthropic and Blackstone, both of which refused to comment. Blackstone is the main investor in the Anthropic chip financing transaction, and the change of judgement relates to the subsequent financing conditions of AI.
If ratings remain at the speculative level, Anthropic will be more difficult and more costly to fund for continued expansion of computing expenditure.
Kirill responded to the intellectual independence: hope the team is more like a new force and will import new products next year.
On October 31, at a three-quarter media exchange, Zilong, Executive Vice President of Chirey Automobile, responded to the independent operation of the intellectual community by expressing the hope that the intellectual team would look more like a new power enterprise and avoid becoming a "Big One" traditional car.
He indicated that Chiri would enhance the ability of the intellectual community to operate independently in terms of the size of the organization, the high end of the product and the gradual optimization of the brand. According to the interface, the intellectuals will be importing new products next year and overseas operations will continue to be run independently by the Chire Group.
Grok API will by default add invisible text watermarks and the team can close by model
The API text output of the Grok 4.7, 4.6, 4.5, 4.3 and three Grok 4.20 models will by default add invisible watermarks to cover global and regional endpoints as of 7 November. Team managers can close watermarks by model on the console.
Watermarks fine-tune the selection of Token during their generation without adding hidden characters or tags to the response, with normal text, structured output and tool call parameters all covered. The developer does not need to modify the request and can check if the output is watermarked by the x-watermark header in the response.
Officially, this was done to comply with the EU Artificial Intelligence Act. The longer the text, the stronger the signal, the more statistically the test will be; the API will have to apply for a separate permission.
Peng, MONA will open an independent store in some cities next year.
According to the interface news, Wang Fengying, the president of Peng Group, said that, starting in 2027, some cities would set up MONA independent doors, and that the future "does not exclude the possibility of independence of the MONA brand".
A preparatory action has already taken place at the end of the channel. In Beijing, a Pengmen shop salesman reported to the interface that Peng and MONA are expected to be sold separately, while the new trip to CEO Hao found that there was a pending "Peng MONA Marketing Centre" in Haiju District, Guangzhou.
Product planning is also being adjusted. The interface news learned that Peng recently merged the four original product lines into G and D, and that MONA continues to have exclusive access to the D product line. He said that only part of the channel next year would be an independent MONA channel and that the existing Pengmen store would still sell the MONA model.
TBC uses neuron research into video models to say that the rate of generation is at least five times higher
The Biological Company Co. (TBC) publishZ13Z model Mischief 1. The company CEO Alex Ksendzovsky states that the team converted four years of active neuron research into a video model for external use, generating at least five times as fast as the underlying model.
In his response, he confirmed that Mischief 1 was based on the Ali Open Source Model Wan 2.2 and that the next generation model was also being developed. The neurons used in the study remain in the laboratory and the video model operates on the normal chip.
Mischief 1 has provided services in fal and OpenRouter to support the generation of short video in pictures. Ksendzovsky also predicted that the model would enter the AWS Marketplace.
Changxiang claims to have consulted with the head client and expects to continue to grow in the fourth quarter
In response to the progress of long-term supply agreements in the record of investor relations activities disclosed on 31 October, Changsung technology stated that consultations had been held with multiple head customers to target long- and medium-term certainty requirements.
According to a report on the log at Chinese and Chinese latitudes, the company believes that the AI application is driving structural demand growth, that the demand for server clients remains strong in the fourth quarter, and that the demand for DRAM will continue, and that the fourth quarter revenue collection is expected to continue to grow.
The company also explained that the long-stamp bridge and long-stamp collection were the main drivers of income and profit growth in the third quarter, but that the group ' s interest in the two subsidiaries was relatively low, thus affecting the increase at the maternal level. In August of this year, the company increased its funding and increased its equity share in the collection of long-term electricity.
💡DeepSeek Harness Team: Multitasking Agent's Bottleneck is in human focus
The DeepSeek Harness team published a long paper on Super-agent architecture exploration on October 31, noting that as models become more and more independent in carrying out individual tasks, bottlenecks in multi-tasking collaboration are turning to attention. The team observed that one person could usually follow only 3 to 5 active sessions, and therefore wanted the system to take over the session management and focus on decision-making and acceptance check.
The programme of the team is to divide responsibilities into two levels: a permanent Router, responsible for cross-mission coordination and external interaction, and a context-separated Worker with specific tasks. Follow-up information on the same GitHub PR or group talk can be delivered directly to the corresponding Worker to reduce the number of repeat paths; time-consuming tools are running backstage and the front desk continues to receive new instructions.
The article describes internal examples that have been able to advance between IM and GitHub on a continuous basis, and to identify deficiencies and try to repair them and submit PRs after the group is empty. In order to maintain long-term operations, the team also separates the computation, session status from the performance environment and uses an independent Watchdog to check whether substantive progress has been made.
The team emphasized that this was a set of engineering architecture explorations, partially designed to be available in the dsh-suberagent-prototype prototype, but not an informal product. The initial experiment, using synthetic loads and a simplified environment; the isolation of context in larger organizations, cross-Agent collaboration, and the risk of prompt infusion and evidence-based disclosure, remains a team-listed open issue.
Read about: DeepSeek Heavy publish New Age Structure: AI Start Command AI
HomeView code displays a secure camera mode and can be recorded at home, outside and sleep settings
Apple HomeView smarthouses may add security cameras. The Verge reports that Reddit user pdfu found a string in HomeView 27.2, one of which sets out that HomeView can be used as a security camera via HomeKit Security Video.
The functions described in the code include mobile detection, real-time video and recording. The video is available to family members only; when people view the video or the device start recording, the device will turn on the light and play the sound, and the family members will be informed of the changes to the Secure Video settings.
Users can also set up detection, streaming and recording content, depending on the state of home, out-of-home or sleep. The above information comes from a code trail that Apple has not yet declared.
ChatGPT webpages and mobile phones support double account switching, and chat and memory are separate.
Open AI has launched account switching on ChatGPT pageend, iOS and Android. Users can log in up to two accounts at the same time, switching from the account menu or settings without having to quit again and again; only one account is active at the same time. Mobile-end functionality is gradually open.
Open AI's Help Document describes the chats, memories, files and settings of the corresponding account, and the workspaces belong to their respective accounts. The two accounts do not combine dialogue, subscription, billing or workspace, and paid entitlements are still determined by the respective accounts ' own subscriptions.
Users can have more than two ChatGPT accounts, but the same device maintains a maximum of two accounts at the same time; when two entries are entered, one of them must be withdrawn before other accounts can be added. The mobile end can also set and receive notifications by account number and equipment permission, and click on the notification for another account to switch and open the corresponding contents.
Analyst predicts it'll take about six to seven months for GTA6 PC.
In BBC Business Today, Wedbush Analyst Michael Pachert predicted that the GTA6 would be about six to seven months to reach the PC. The interview footage was circulated on social platforms on October 28th.
Rockstar's launch date for the mainframe version was November 19, 2026, and the launch platform was PS5 and Xbox Series X|S. Pacht gave his own time judgement, and Rockstar had yet to publish the PC release date.
Hightower X GPU drive to improve game stability from generation to generation
The publish is applied to the Adreno 2026.10.1 drive of the X-System Windows PC to update the installation framework that has been introduced and also supports the initial X-System X2 with separate driver versions and improved lists for each of the two-generation platforms.
The update of the original Tyrone X related to the stability or the quality of The Unfinished Darts: Redemption 2 and Mechanical Combat Police: Riot City; the improvements of Kiryu X2 included the Extreme Race: Horizon 6 and the Runner 5 Legacy NBA 2K26. The stability and performance improvements of applications such as Luminar Neo are listed on both generation platforms.
The official note also retained a known problem with Windows 11: in Build 26100.3915 and beyond, the insertion of external monitors or the switch to display mode, resolution could lead to the collapse or death of Subway: Leave, Civilization VI, Path to Life II.
Spotify plans to launch an AI mix at the beginning of next year so that fans can adapt authorized songs
According to the Financial Times, Spotify is working on the launch of the AI Mixing and Singing Service in the first quarter of 2027, which is likely to be visible before and after the Grameen Prize in February. paid subscribers can use it to adapt mandated songs, for example, by turning folk songs into dances, or by producing singers who sing songs of other works.
Spotify has entered into relevant agreements with Globe Music, the Independent Authorised Institute Merlin and Believe, and is still negotiating with Sony and Warner. Even when a record company contracts, musicians can decide whether or not to participate.
Under the current scheme, royalties generated by fan mix will flow to original musicians. Spotify is also developing a mechanism to bring these works into the playlist and to include the play volume in the original list. According to the source, the timing and pricing are still under discussion.
Meta and Fourier built a human robot, Max.
According to The Information, Meta and the Shanghai Robotics company, Fourier, built the human robot prototype Max for about three months, and started developing the second version of Max 2 in September. In response, the spokesperson for Meta stated that the company had no plans to sell the prototypes, with the work of Fourier focusing on internal research and experiments.
Max is about 1.5 metres tall and 48 kg heavy, carrying Weeda Jetson Thor, with a speaker in the chest, a face screen that simulates expression, and a five-finger hand that grabs a small bottle and a toy. Meta, based on the hardware and experience of the first three generations of Fourier robots, adjusted the body, head, monitor and camera and added parts such as hands.
It was reported that the head of the Meta robotics product had been present at the Fuleya Shanghai plant for almost two months and that the prototype had been demonstrated at an internal summit at the end of August. The second edition will rework the body and hands, add projectors and sensors to help up and down the stairs; the human robots are considered within Meta as its physical carriers for tasks such as housework.
Open source Andre's assistant, Mika, can operate in the backstage. App, the main screen still works.
The independent developer Wang Zan opened up Andre AI assistant Mika. When a user delivers a task in a natural language, it opens the App, clicks, enters and scrolls on a stand-alone backscreen, which remains for users.
The project shows the use of backstage viewing and summing up WeChat conversations, texting to file transfer assistants and sifting commodities in shopping App. Mika does not require each App to provide an additional interface by reading an interface; login, authentication code and payment are returned to the user.
Crossing the App operational requirement Android 12 or updating the system, for the first use you need to complete a wireless debugging pair. Project recommends Android 14 and above that virtual screens have a stand-alone input focus and reduce the interference of background tasks with the main screen input. Users can access their own model API or log in through a supported model service account; the project uses the Apache 2.0 license.
The new car will continue to provide a low-interest programme.
In November, Mi cars announced their right to purchase cars, and a small “three-year zero-center” financial scheme was added to N70's N70, and customized new cars offered “five years of low interest” or “limited-time easy to pay off”. Both categories of interest apply to orders placed by 24 o'clock from 1 November to 30 November.
- The N70 is a small, cash car: a down payment of $799 million, with three out of three out of "three years at 0 interest" and "five years at 0 interest" and "6,000 insurance benefits".
- A new car is customised: a down payment of $49.9 million for a low-interest scheme, 2932 for five months of the year, and an alternative "limited-time easy to pay" option.
- SU7, YU7 and SU7 Ultra: SU7 is a small number of cash vehicles, and a small number of cash vehicles in a specified version of YU7 are available for "3 years at 0 interest" or "0 interest at 0 interest at 5 years" or "6000 yuan insurance subsidy"; SU7 Ultra provides limited-time "0 interest at 5 years".
Akha Ama opened a third store in Tokyo.
Akha Ama Coffee from the northern mountains of Thailand opened a third Japanese coffee shop in Tokyo. On October 31, Japan reported that brands wanted to further open the Japanese market with high-quality, sustainable Thai coffee beans.
The founder, Lee Ayu Chuepa, is from the Akha Ama ethnic group, which means "Aka mother". The brand was created in 2010 to purchase coffee beans from farmers in Taipei, including the founder ' s mother and sister, to reprocess, bake and supply Chiang Mai and Tokyo doors. Farmers grow Alabica coffee at high altitudes and combine it with crops such as fruit and nuts, using agroforestry as a combination.
The founders stated that brands valued quality and origin stories and were improving quality by reducing fertilizer and chemical use. About 90 percent of the coffee beans are currently sold in Thailand and the rest are exported to Japan; the brand opened its first shop in Tokyo in 2020 and a second in 2024.
Implementation of new national standards for classroom lighting, with new requirements for the synchronization of flash and light
On 1 November, the mandatory national standard, “Standards for the luminant and lighting of classrooms in primary and secondary schools”, was officially implemented, and evaluation indicators for lighting and lighting in classrooms were further refined.
According to press reports, the new standards also incorporate the flash index and the frequency-blind-effect vision into the indicators by adding requirements such as light-sizing evenness, full-window window time from winter to sun, glass colour perceiving index and visible light perceiving ratio. The light orientation, light coefficients, dizziness control, light colour temperature, colour index and measurement methods have been revised.
The ophthalmologist Tanaka, attached to Beijing ' s fellow hospital at the Capital Medical University, who was interviewed, explained that light sources appeared stable, that the visual system could still adapt to changes in light and that it was prone to eye fatigue in such a flashing environment for a long time.
Hidari director talks about Keanu Reavis' union: after watching a puppet short film, a week later, we had a video conference.
As revealed at the Tokyo International Film Festival by the director of the Japanese fixed-animation film Hidari, Kinu Reavis saw the project's first short film, the team received a response in about a week, arranging for video calls from both sides. In May of this year, Reavis was released in Cannes as the leading actor, voiceover; the director further described the process of the invitation and the preparations for the long film.
Kawamura stated that when the work was influenced by the Kurosawa Ming film, the leading puppet, he wanted to combine the image of the three ships with that of Reeves and to make the role look older. Hidari uses a combination of carvings, carvings and time operas, which the team positioned as action films and presented to overseas distributors with hand-carved puppets.
According to Variety live reports, about three years ago, the team was self-financing and using crowd-sourced concept clips, uploading YouTube and getting about 5.3 million views. Long films are still seeking subsidies and overseas distribution cooperation; the producers have previously revealed that a United States company has offered to cover the full budget in exchange for full copyright and that negotiations have stalled.
"Godzilla Minus Zero" will be released in Japan on November 3, two years after the story began.
Godzilla Minus Zero will be launched in Japan on 3 November. In the Tokyo International Film Festival, Mikisuke and Mimibo, the new film continues the story of the Katoshima family, who, after about two years of their previous existence, are again threatened by Godzilla.
The two leading actors continued to play the role of Koichi and Shigeshima, and Nagagi returned to play Mingko. Mikisuke said that the last one had a full pair of words, mainly in the form of tears or words.
According to live reports by Variety, the founders had just returned to Japan from the L.A. premiere, and the audience had shouted the name of the director in the red carpet. The new film is also the first Japanese film to be certified as "Fimmed for IMAX".
"Silver Spirit" director talks about the re-programming of Quihara's last article: Re-engineering 200 more shots for the new edition
"Silver Spirit" animated the movie "On Gita Inflammation" in Luca, Italy. In an interview with the filmmaker Variety, director Ando Sang also revealed that after the Japanese version released in February this year, the team took the opportunity to rework more than 200 shots on video distribution.
The film revisited the original people's page. Ando Sang also indicated that, in order to make the action scene suitable for the big screen, the team used CinemaScope super-wide banners and used 3D technology to increase camera motion and image changes and produced a large amount of new music. Compared to television animations, the film actually shows the burning of Yoshihara.
Turning to the humor of Silver Spirit, he said that the team had placed a director in charge of the comedy section; they wanted the characters to naturally joke with each other and the audience to hear and feel fun.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.