{ "text": "一句话就能让AI打开真实手机\n跨应用查路线算时间播放音乐\n最后自动提交测试报告\n谷歌刚刚把这套系统开源\n项目方还给出了99.1%的任务成功率\n但先别急着把它理解成\n安卓已经被AI彻底接管\n因为那个数字背后还有一场关于\n跑分代码来源和开源署名的争议\n这套系统叫Artemis\n它面对的是一台真实\nAndroid手机或模拟器\n网页和专门为AI准备\n的演示界面都不够\n开发者用自然语言描述任务\nAthena观察屏幕\n寻找控件\n执行点击和输入\n再根据界面变化决定下一步\n一个典型任务可以这样写\n打开Google Maps查询\n三个地点的行车时间\n把结果记录下来\n再打开YouTube搜索Coldplay并播放歌曲\n传统自动化需要工程师\n逐个编写控件选择器\n等待条件和异常处理\nArtemis接到目标后\n会自己拆解步骤\n并在应用之间切换\n这件事听上去只比\n语音助手多了几次点击\n底层难度却完全不同\n语音助手通常调用预先接好的接口\n能做什么早已写死\nAutomis面对的是不断变化的图形界面\n按钮会移动\n弹窗会突然出现\n网络速度不稳定\n应用更新以后控件名称也可能改变\n传统移动测试因此很脆弱\n一条脚本昨天还能运行\n今天可能因为按钮换了位置就失败\n测试工程师往往在维护\n大量选择器和等待时间\n业务逻辑反倒没有变化\n产品功能没有变化\n自动化脚本仍然要重写\nOptimus尝试把这部分维护\n工作交给多模态模型\n它先读取Android的无障碍层级\n获得界面中的文字\n控件类型和可操作区域\n能找到稳定元素时\n它通过元素索引点击\n界面结构不足时\n再使用坐标OCR和视觉定位\n这套组合很重要标准\nAndroid应用通常会暴露清楚的控制树\nAI可以知道哪个元素是按钮\n哪个元素是输入框\nflutter compose和CANVAS绘制的\n界面可能只给出很少结构\n游戏界面甚至完全没有可用控件\noptimus遇到这种情况就\n转向截图和视觉模型\n纯视觉方案也有明显缺点\n截图里的两个按钮可能长得很像\n小图标不一定能准确识别\n动画还会让坐标快速失效\nArtemis没有把全部\n希望放在视觉模型上\n而是优先使用结构信息\n视觉负责补位\n这个选择降低了推理成本\n也让操作更稳定\n项目提供两套执行模式\nFlash模式追求速度\n采用观察一次\n行动一次\n再次观察的循环\nreadme给出的典型速度是每一步3-5秒\n它不会先写一份庞大计划\n也不会为每个动作生成长篇说明\n适合目标清楚\n流程较短的任务\n打开设置并关闭蓝牙\nFlash模式看到当前界面后\n会直接寻找设置入口\n如果已经位于快捷面板\n它不必坚持原先设想的路径\n反应式循环的好处是\n灵活界面出现意外弹窗时\n也能临时调整\n速度的代价是缺少全局检查\n任务变成长流程后\n前面一个小错误可能一直带到最后\n代理以为邮件已经发送\n实际停在草稿页\n他以为文件已经移动\n实际只打开了目标文件夹\n每一步看起来都合理\n最终状态却不符合要求\nPro模式就是为这种任务准备的\n他使用多代理图把职责\n分给planner operator和Checker\nplanner维护一份持续\n更新的任务计划operator\n操作设备checker在关键\n节点和结束时检查结果\n一次测试超过100步也可以\n保留阶段状态继续执行\nplanner并不只是把\n任务写成编号列表\n计划里会标出需要验证的里程碑\n例如联系人是否真的创建\n文件是否已经出现在目标目录\n设置开关是否保持正确状态\n界面发生变化后\n计划也能调整\n而不是机械重复最初路线\noperator拿到计划后调用完整工具集\n它可以读取屏幕结构\n查看截图\n分析录屏\n调用ADB诊断\n也可以搜索此前\n压缩保存的操作历史\n遇到失败时\n系统会创建一条执行事件\n把失败原因保留在后续上下文中\n直到某次操作真正恢复\nchecker保持只读\n不负责替operator点击\n这样设计是为了避免执行者\n一边操作一边宣布自己成功\nchecker根据任务目标检查最终状态\n也可以在中间里程碑做验证\n用户还能选择关闭验证\n只做最终验证\n检查所有节点\n或者使用更严格的模式\n每个动作执行前还有一道safety net\n系统优先根据XML界面树确认目标\n结构信息不足时再检查像素按钮\n在模型思考期间已经消失\n操作就会被阻止并交回operator\n移动界面中的状态变化很快\n这道检查可以减少点\n到错误位置的情况\nArtemis还支持连续动作突发\n某些菜单只会短暂显示\n如果每点一次都等待模型重新观察\n菜单早已关闭\n系统可以把几个确定动作连续发出\n缩短模型往返延迟\n这类工程处理很少\n出现在产品宣传中\n却直接决定代理\n能不能在真实应用里工作\nFlash模式每步大约3-5秒\nPro模式通常需要15-40秒\n速度差距提醒了一个现实\n可靠性不是免费的\n规划检查历史检索\n和故障恢复都会增加模型调用\n短任务追求速度\n长任务追求结果\n两种场景很难用同一套配置解决\n项目最值得关注的用途是软件测试\n普通人远程控制手机还排在后面\n开发者完成一项功能后\n可以让Artemis在真机上复现问题\n收集Logcat日志和截图\n再把结果交回编程代理\n过去分开的写代码跑手机看日志\n开始被串成同一条工作流\n这里的关键接口是MCP\nOptimus自带MCP服务器\n可以连接Antigravity Claude code\nWindsurf等支持MCP的开发工具\n编程代理不再只能阅读代码仓库\n他还可以看到应用\n实际运行后的界面\n并操作设备验证自己的修改\n设想一个常见场景\n用户报告结账按钮点了没有反应\n开发者只拿到一句模糊描述\n接入Artemis后\n编程代理可以安装测试版本\n按照用户路径进入商品页\n触发结账\n截取错误界面\n再读取Logcat中的异常\n它修改代码后还能\n重新执行同一流程\n这会改变修复完成的定义\n很多AI编程工具生成代码后\n只能运行单元测试\n单元测试通过\n不代表手机界面真的可用\n按钮可能被遮挡\n权限弹窗可能阻断流程\n键盘弹出后布局也可能错位\nArtemis把验证推进到设备层\n项目还提供CLI Python SDK\n和本地网页控制台\nCLI适合在终端直接发起任务\nPython SDK可以接入pytest\n和CI流程网页控制台\n则显示手机画面\n推理轨迹和执行过程\n失败任务可以回放\n测试人员不必只看一句执行失败\n快速启动脚本会准备ADB\nscript ffpeg和Python依赖用户\n可以连接开启USB\n调试的实体Android设备\n也可以使用模拟器\n第一次运行时\n系统会在设备上\n安装Artemis accessibility helper\n用它读取屏幕布局\n这个辅助服务会出现在\nAndroid的无障碍设置中\n并显示一条运行通知\n项目说明称\n它只在手机本地监听\n不会自行把数据发送到其他位置\n如果辅助服务失效\n系统还能回退到uiautomator\n这里要分清本地\n辅助服务和模型调用\n前者可以本地通信\n截图及界面信息是否离开\n设备仍取决于模型供应商配置Athena\n支持Gemini Claude GPT4O和千问VL等模型\n它本身更接近一个代理框架\n不是新的基础模型\n实际表现会受到所选模型\n网络设备性能和应用界面影响\n换一个便宜模型\n成本可能下降\n定位和规划能力也可能变化\n这也是99.1%跑分需要\n谨慎理解的原因\n项目宣称自己在Android world\n达到99%以上任务完成率\n图表给出的具体数字是99.1%\nAndroid world是Google research\n推出的公开环境\n包含116项手工设计任务\n覆盖20个应用\n并通过随机参数生成大量变化\n任务包括创建联系人\n发送短信\n添加日历事件\n调整亮度开关Wifi\n移动文件和处理播放列表\n任务程序会检查设备最终状态\n不接受模型自己宣布完成了\n这比依靠文字答案评分更可靠\n也更接近真实操作\n99.1%依然不等于现实手机上的99.1%\nAndroid world使用固定系统版本\n预设应用和明确任务类型\n真实手机有厂商定制界面\n广告弹窗登录验证地区差异\n和不断变化的第三方应用\n测试集能衡量代理进步\n无法覆盖开放世界的全部混乱\n这项成绩目前也属于团队自曝\n公开排行榜接受社区提交\n并不会替每个项目\n完整复现所有运行\nArtemis仓库展示了对比图\n却没有提供足够完整的第三方\n审计材料来证明每一次任务如何\n运行 99.1%可以称为公开的项目成绩\n不能包装成已经独立认证的事实\n基准测试还存在适配风险\n团队知道全部116项任务后\n可以针对常见应用和\n失败模式优化提示词\n恢复逻辑和工具\n优化并不等于作弊\n工程系统本来就该改进薄弱点\n但一个系统在熟悉\n测试集上接近满分\n不代表面对从未见过\n的应用也有同样表现\n更敏感的问题来自项目的代码来源\nArtemis公开后\nminitab团队在Git Hub提出异议\n称仓库包含其开源\n项目mobileuse的代码\n却没有在最初版本中清楚\n保留来源和贡献者信息\nminitype列出的证据很具体\n早期PI project Tamo曾出现\nMobile use的版本号3.6.3\n以及三位原作者姓名\n后续提交替换了作者信息\n其他大量文件保持不变\n双方共享的内容涉及ADB隧道\n规划辅助函数应用锁定\n工具代理提示词和SDK示例\n开源代码可以复用\n但Apache2.0并不等于可以抹掉来源\n许可证允许修改和商业使用\n同时要求保留相应版权及许可说明\n争议焦点不在Google\n能不能使用这些代码\n而在它是否完整履行了署名义务\n以及发布过程为什么\n先移除了原作者信息\n当前仓库已经补充说明\n明确写着项目包含\nminitab开发的源代码\n相关文件也增加了版权和来源标记\n提出问题的Git Hub议题已经关闭\n这解决了当前readme没有署名的问题\n却无法抹去最初发布时发生过什么\n这场争议让Artemis的\n故事变得更有价值\n大型公司开源项目时\n外界往往默认代码已经\n经过完善法务和来源审查\n实际情况说明\n即使仓库挂在Google组织名下\n也需要查看提交记录\nnotice和文件头\n品牌不能代替开源合规\n跑分争议又让问题多了一层\nMini Tab表示自己的mobileuse\n后来曾自测到更高层级\n但公开榜单没有及时更新\nArtemis的图表列出多个竞争项目\n却没有清楚呈现双方最新可比结果\n这里无法仅凭一方说法\n判断榜单维护者的动机\n但可以确认排行榜存在\n更新和独立验证不足\n这些问题不应被用来\n否定Artemis的全部技术价值\n仓库不是一个空壳\n里面有完整的命令行工具\nMCP服务器\n设备驱动\n网页控制台\n测试目录和多代理流程\n短时间获得大量关注\n也说明开发者确实\n需要更可靠的移动端\nAI自动化\n同样不能因为代码能\n运行就忽略来源问题\n技术能力和开源\n合规是两条独立标准\n一个项目可以有实用价值\n同时在发布流程中犯下严重错误\n把任何一边省略都会让判断失真\nArtemis当前的成熟\n边界也写在路线图里\nAndroid studio原生插件仍未完成\nios扩展还在计划中\n端侧轻量视觉模型和实时\n双向语音控制也没有正式交付\n现在可用的是Android设备自动化框架\n不是覆盖所有移动平台的万能代理\nios迟早会成为重要方向\n但难度不会只是把ADB\n换成另一套命令\n苹果系统对自动化后台\n运行和权限控制更严格\n实体设备与模拟器的行为也不同\n项目路线图写了ios\n不代表现有架构已经解决这些限制\n端侧视觉模型同样值得关注\n今天使用云端多模态模型\n截图和界面信息\n可能需要传到服务商\n延迟和隐私都受影响\n轻量模型如果能在手机\n或开发机本地识别界面\n可以减少数据外传\n也能缩短每一步等待时间\n代价是模型能力下降\n复杂页面的识别可能不够稳定\nArtemis要进入企业测试\n还要解决权限和隔离\n测试手机可能登录内部账号\n包含未发布功能和真实业务数据\n代理拥有ADB截图和日志读取能力\n一旦配置错误\n泄露范围会比普通测试脚本更大\n企业需要限制模型端点\n保存周期和可执行命令\n提示词注入也是\n移动代理绕不开的风险\n应用页面可以显示\n一段专门写给AI的文字\n诱导它忽略测试目标\n打开其他应用或泄露信息\n人类测试员看到异常文字\n会判断模型可能把它\n当成高优先级指令\nsafety net能检查目标位置\n未必能识别内容本身的恶意意图\n持续集成环境还需要可复现性\n同一个自然语言任务今天通过\n明天失败\n开发团队很难判断是代码\n回归模型波动还是界面变化\nOptimus必须记录模型版本提示\n词截图动作轨迹和设备状态\n才能让失败结果具有工程价值\n成本也不能被99.1%的数字遮住\nPro模式每一步可能消耗15-40秒\n还会调用规划视觉和检查模型\n一次上百步的测试要\n花多少时间和API费用\n决定它能否替代传统自动化\n高价值探索测试可以接受较高成本\n大量重复回归测试未必划算\n最合理的使用方式可能是混合测试\n稳定高频的核心流程\n继续使用传统脚本\n速度快结果也容易复现\n经常变化或难以维护\n的界面交给Artemis探索\nAI发现稳定路径后\n还可以帮助生成更固定的测试用例\n这样分工比AI取代所有\n测试工程师更接近现实\n测试人员需要定义风险准备\n环境判断结果和处理异常\nArtemis减少的是反复找按钮维护\n选择器和手工复现的时间\n并没有替团队决定什么功能最重要\n也不能替公司承担发布责任\n测试工程师的工作会随之变化\n过去大量时间花在脚本维护\n未来更多时间会用于设计任务\n设置验证条件和分析失败轨迹\n自然语言看似降低了门槛\n准确描述目标反而更重要\n含糊任务会让代理找到\n一条看起来完成的路径\n却漏掉真正的业务要求\n开发流程也可能从提交代码后统一\n测试变成编程代理边写边操作设备\n它修改登录页面\n立刻安装构建版本\n走一遍登录和找回密码\n再根据日志继续修改\n设备不再是开发链\n末端的人工检查站\n而是AI编码循环的一部分\n这会给小团队带来明显价值\n没有专职移动测试工程师的团队\n也能建立基本的真机回归流程\n可规模扩大以后\n设备农场账号管理和并发\n调度仍然需要专业基础设施\noptimus解决了操作智能\n不会自动解决所有测试平台问题\n设备管理是一项容易被忽略的成本\n几十台手机同时测试\n需要保持系统版本\n屏幕分辨率和应用\n数据处于可控状态\n某项测试结束后还要清理\n联系人短信文件和登录状态\n否则下一次任务会\n读取上一次留下的数据\nAndroid world之所以能评测代理\n正是因为它为任务\n准备初始化和结束检查\n企业自己的应用没有现成评测器\n团队必须定义什么叫成功\n页面出现提交完成\n可能只是前端提示\n后台订单未必真正创建\nOptimus可以点击和观察\n却需要业务系统提供\n更可靠的验证信号\n数据库记录接口状态和\n设备画面要互相对照\n才能减少假成功\n自然语言任务也需要版本管理\n测试人员把创建一个联系人改成\n创建联系人并确认号码能够搜索到\n验证范围已经变化\n如果这段文字只存在聊天记录中\n团队很难追踪回归标准\n成熟用法需要把任务描述断言\n和环境配置像代码一样保存\n并进入评审流程\n模型更新还会制造另一种波动\n供应商更换默认模型后\n同一句指令可能走不同路径\n新模型更聪明\n却也可能更愿意尝试危险动作\n团队要锁定模型版本\n建立基线\n并在升级前重新运行关键任务\n传统脚本的变化来自代码\n代理测试的变化还来自模型\nOptimus的历史压缩机制\n就是为长任务控制上下文\n它不会一直保留每张旧截图\n而是把已经完成的\n步骤压缩成可搜索记录\n否则100步任务会迅速塞满模型窗口\n费用和延迟一起上升\n压缩能延长任务\n代价是某些视觉细节\n可能在摘要中丢失\n历史检索因此不能只\n依靠一句之前已经完成\n系统需要保留关键证据\n例如创建后的联系人页面\n移动后的文件路径和操作返回值\n摘要负责告诉代理发生过什么\n原始证据负责在出现争议时复核\n两者缺一项\n长流程都会变得不可信\n录屏分析提供了另一条诊断线\n界面一闪而过\n单张截图可能抓不到问题\n录屏可以重建动画\n弹窗和键盘出现的顺序\nOptimus把fempy列为依赖\n就是为了处理这类时序信息\n移动端bug经常只存在几百毫秒\n日志里也未必留下明确错误\n不过录屏会扩大隐私范围\n测试账号的通知联系人头像\n和输入内容都可能进入文件\n团队需要在采集前关闭无关通知\n使用合成数据\n并限制报告中的截图\n代理自动收集越方便\n数据治理越不能依靠\n测试人员临时判断\n权限边界还涉及ADB\nADB能够安装应用\n读取日志\n启动组件\n并执行系统命令权限\n远高于普通用户操作\nArtemis如果连接生产手机\n风险非常高\n合理部署应使用专门\n测试设备和隔离账号\nCI中的密钥也不能直接暴露给模型\n开源在这里提供了现实优势\n企业可以检查每个工具\n到底调用了什么命令\n也能删掉不需要的高风险能力\n闭源手机代理只给出服务界面\n用户很难确认数据流向\nOptimus把设备驱动和代理逻辑公开\n至少让安全团队有\n审计和改造的入口\n开源也带来维护压力\n仓库没有正式版本标签\n当前项目迭代集中在很短时间内\n依赖很多\n既有LANGGRAPH和多家模型SDK\n也有ADB UI Automator和图像处理组件\n任何上游变化都可能影响安装\n9,000多个星标代表关注度\n不等于长期维护承诺\n开发者在实际采用前\n需要先跑自己的小型评测\n选择10-20条最关键流程\n分别记录成功率平均耗时\n模型费用和人工干预次数\n只看Android world\n总分无法回答企业自己的登录\n支付和权限页面能不能稳定运行\n失败类型也应分开统计\n模型找不到控件属于感知问题\n计划走错属于推理问题\n应用崩溃属于产品问题\n网路超时则是环境问题\n把所有失败压成一个百分比\n团队不知道该修提示词换模型\n还是修改应用本身\nAtemis有时最有价值的结果是暴露\n产品不利于自动化和无障碍访问\n通过本身反倒没那么重要\n一个按钮没有语义标签\n模型只能猜坐标\n视障用户的屏幕\n阅读器也可能无法识别\n代理测试会反过来推动\n开发者补齐控件描述\n最终改善真实用户体验\n这让它与传统视觉测试形成区别\n截图对比擅长发现\n颜色位置和布局变化\n而Temis更擅长判断任务能不能完成\n前者告诉团队按钮移动了5个像素\n后者告诉团队用户\n已经无法完成退款\n两类测试结合\n覆盖面才完整\n市场上也不只有Artemis\nfinal run auto device droid run uitas\n和多种computer use模型都\n在争夺移动代理入口\n他们选择不同模型和规划架构\n公开程度也不同\n99.1%的领先\n如果不能被持续复现\n很快就会被下一轮模型和工具改写\nGoogle开源这套框架还有生态意义\nAndroid开发者可以围绕它增加\n设备驱动行业任务和验证插件\n模型公司也能把它作为\n计算机操作能力的试验场\n一个共同框架如果形成移动代理\n就不必每家公司都从\n截图和点击重新造轮子\n可共同框架必须建立可信治理\n贡献代码需要清楚来源\n排行榜需要公开测试条件\n安全漏洞需要有响应流程\nminitab争议已经说明\n技术社区不仅看功能\n也看维护者如何处理他人的劳动\nGoogle后来补上署名是必要修正\n公开解释仍然同样重要\nArtemis未来若接入Android studio\n开发者可以在IDE里选中一个界面\n直接生成并运行设备任务\n这个设想很有吸引力\n但插件目前还在路线图\n现阶段仍要通过命令行网页\n控制台或MCP连接部署体验\n更适合愿意处理\n环境配置的技术团队\n实时语音也尚未完成\n即使以后上线\n它的意义也不只是\n说一句话控制手机\n双向语音要求代理在\n操作过程中报告进度\n用户能够随时打断\n并在敏感步骤确认\n没有可靠中断机制\n语音只会让错误操作更难阻止\n长期来看\n手机代理可能从测试\n工具进入运维和客服\n客服人员接到用户问题后\n可以在隔离设备上复现相同路径\n应用商店审核可以自动\n探索权限和订阅流程\n企业内部还能检查不同\n版本是否出现行为变化\n这些场景都需要\n明确授权和受控环境\n不能直接拿真实用户账户试验\n最有争议的想象是让\n代理自行发现漏洞\n它能够遍历界面\n输入异常内容\n观察崩溃并读取日志\n确实具备一定探索能力\n公开项目的目标是测试和诊断\n不等于授权代理攻击第三方应用\n使用边界仍由操作者和\n目标系统的许可决定\n它对普通用户也有潜力\n视障人士可以用自然\n语言完成复杂应用操作\n老人可以减少在\n多层菜单中寻找按钮\n重复性手机任务可以交给代理不过\n项目当前的产品形态面向开发者\n需要配置Python模型\n密钥ADB和无障碍权限\n距离大众助手还有明显距离\nArtemis真正释放出的信号是AI\n代理开始从浏览器走进设备\n网页有DOM和相对统一的交互结构\n手机应用更加封闭\n一个开源框架能够\n跨应用完成长流程\n说明多模态模型工具调用和\n状态管理已经组合到新的阶段\n这个阶段仍然粗糙\n速度按秒计算\n复杂模式一步可能接近1分钟\n权限配置不当会带来安全风险\n基准成绩缺少完整独立验证\n开源署名还发生过争议\n它没有资格被包装成\n成熟的手机操作系统\n代理可它也不是一次\n无关紧要的代码发布\nGoogle把项目放在公开仓库\n使用Apache 2.0许可证\n开发者可以检查实现\n替换模型\n并接入自己的设备\n开源让外界有机会复测99.1%\n也让代码来源问题无法被藏起来\n它同时接受技术检验和社区审查\n接下来最值得观察的是三件事\nGitHub星标能涨到多少并不关键\n独立团队能否复现接近99.1%的结果\n企业能否在真实应用\n中稳定运行长流程\n以及维护者能否持续\n处理安全与署名问题\n任何一项失败\nOptimus都可能停留在演示阶段\n如果三项都能过关\n移动开发会发生一次实质变化\n编程代理将不再只盯着代码和终端\n他能拿起一台手机\n用用户的方式检查自己的作品\n测试脚本也不必把每个坐标写死\n目标和验证条件会\n成为新的核心资产\n所以Artemis最重要的\n地方不是AI会点手机了\n真正的变化是\nAI第一次有机会把理解需求修改\n代码操作设备和验证结果连成闭环\n99.1%只是吸引注意力的数字\n能不能把这个闭环\n稳定放进真实开发流程\n才决定它会成为基础工具\n还是另一个跑分漂亮的开源项目", "language": "zh", "duration": 1738.0, "segments": [ { "id": 0, "start": 0.166, "end": 2.866, "text": "一句话就能让AI打开真实手机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 1, "start": 3.1, "end": 6.466, "text": "跨应用查路线算时间播放音乐", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 2, "start": 6.5, "end": 8.566, "text": "最后自动提交测试报告", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 3, "start": 8.766, "end": 10.733, "text": "谷歌刚刚把这套系统开源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 4, "start": 11.133, "end": 15.533, "text": "项目方还给出了99.1%的任务成功率", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 5, "start": 15.933, "end": 17.8, "text": "但先别急着把它理解成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 6, "start": 17.8, "end": 20.133, "text": "安卓已经被AI彻底接管", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 7, "start": 20.5, "end": 22.866, "text": "因为那个数字背后还有一场关于", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 8, "start": 22.866, "end": 26.5, "text": "跑分代码来源和开源署名的争议", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 9, "start": 26.6, "end": 28.266, "text": "这套系统叫Artemis", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 10, "start": 28.466, "end": 29.933, "text": "它面对的是一台真实", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 11, "start": 29.933, "end": 31.733, "text": "Android手机或模拟器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 12, "start": 32.133, "end": 33.966, "text": "网页和专门为AI准备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 13, "start": 33.966, "end": 35.6, "text": "的演示界面都不够", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 14, "start": 35.766, "end": 38.0, "text": "开发者用自然语言描述任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 15, "start": 38.3, "end": 39.8, "text": "Athena观察屏幕", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 16, "start": 39.9, "end": 40.866, "text": "寻找控件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 17, "start": 41.0, "end": 42.6, "text": "执行点击和输入", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 18, "start": 42.733, "end": 45.133, "text": "再根据界面变化决定下一步", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 19, "start": 45.333, "end": 47.0, "text": "一个典型任务可以这样写", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 20, "start": 47.0, "end": 48.6, "text": "打开Google Maps查询", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 21, "start": 48.6, "end": 50.366, "text": "三个地点的行车时间", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 22, "start": 50.4, "end": 51.866, "text": "把结果记录下来", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 23, "start": 51.866, "end": 55.5, "text": "再打开YouTube搜索Coldplay并播放歌曲", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 24, "start": 55.733, "end": 57.7, "text": "传统自动化需要工程师", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 25, "start": 57.7, "end": 59.8, "text": "逐个编写控件选择器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 26, "start": 60.3, "end": 62.2, "text": "等待条件和异常处理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 27, "start": 62.366, "end": 64.066, "text": "Artemis接到目标后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 28, "start": 64.066, "end": 65.466, "text": "会自己拆解步骤", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 29, "start": 65.533, "end": 67.2, "text": "并在应用之间切换", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 30, "start": 67.333, "end": 68.8, "text": "这件事听上去只比", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 31, "start": 68.8, "end": 70.8, "text": "语音助手多了几次点击", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 32, "start": 71.066, "end": 72.9, "text": "底层难度却完全不同", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 33, "start": 73.066, "end": 75.966, "text": "语音助手通常调用预先接好的接口", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 34, "start": 76.066, "end": 77.8, "text": "能做什么早已写死", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 35, "start": 78.366, "end": 81.866, "text": "Automis面对的是不断变化的图形界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 36, "start": 82.066, "end": 83.3, "text": "按钮会移动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 37, "start": 83.566, "end": 85.1, "text": "弹窗会突然出现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 38, "start": 85.2, "end": 86.666, "text": "网络速度不稳定", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 39, "start": 86.8, "end": 90.1, "text": "应用更新以后控件名称也可能改变", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 40, "start": 90.266, "end": 92.466, "text": "传统移动测试因此很脆弱", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 41, "start": 93.0, "end": 95.1, "text": "一条脚本昨天还能运行", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 42, "start": 95.3, "end": 98.2, "text": "今天可能因为按钮换了位置就失败", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 43, "start": 98.566, "end": 100.533, "text": "测试工程师往往在维护", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 44, "start": 100.533, "end": 102.7, "text": "大量选择器和等待时间", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 45, "start": 102.933, "end": 104.9, "text": "业务逻辑反倒没有变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 46, "start": 105.166, "end": 106.766, "text": "产品功能没有变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 47, "start": 107.166, "end": 109.066, "text": "自动化脚本仍然要重写", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 48, "start": 109.166, "end": 111.2, "text": "Optimus尝试把这部分维护", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 49, "start": 111.2, "end": 113.133, "text": "工作交给多模态模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 50, "start": 113.466, "end": 115.9, "text": "它先读取Android的无障碍层级", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 51, "start": 116.2, "end": 117.9, "text": "获得界面中的文字", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 52, "start": 117.933, "end": 120.133, "text": "控件类型和可操作区域", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 53, "start": 120.466, "end": 122.1, "text": "能找到稳定元素时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 54, "start": 122.1, "end": 123.966, "text": "它通过元素索引点击", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 55, "start": 124.4, "end": 125.966, "text": "界面结构不足时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 56, "start": 126.066, "end": 129.5, "text": "再使用坐标OCR和视觉定位", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 57, "start": 129.533, "end": 131.7, "text": "这套组合很重要标准", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 58, "start": 131.8, "end": 135.133, "text": "Android应用通常会暴露清楚的控制树", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 59, "start": 135.2, "end": 137.566, "text": "AI可以知道哪个元素是按钮", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 60, "start": 137.766, "end": 139.4, "text": "哪个元素是输入框", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 61, "start": 140.066, "end": 142.733, "text": "flutter compose和CANVAS绘制的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 62, "start": 142.733, "end": 144.766, "text": "界面可能只给出很少结构", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 63, "start": 144.9, "end": 147.466, "text": "游戏界面甚至完全没有可用控件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 64, "start": 147.7, "end": 149.866, "text": "optimus遇到这种情况就", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 65, "start": 149.866, "end": 152.133, "text": "转向截图和视觉模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 66, "start": 152.333, "end": 154.366, "text": "纯视觉方案也有明显缺点", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 67, "start": 154.8, "end": 157.266, "text": "截图里的两个按钮可能长得很像", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 68, "start": 157.4, "end": 159.766, "text": "小图标不一定能准确识别", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 69, "start": 159.966, "end": 162.266, "text": "动画还会让坐标快速失效", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 70, "start": 162.533, "end": 163.766, "text": "Artemis没有把全部", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 71, "start": 163.766, "end": 165.733, "text": "希望放在视觉模型上", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 72, "start": 165.8, "end": 167.933, "text": "而是优先使用结构信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 73, "start": 168.366, "end": 169.6, "text": "视觉负责补位", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 74, "start": 169.766, "end": 171.866, "text": "这个选择降低了推理成本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 75, "start": 171.866, "end": 173.466, "text": "也让操作更稳定", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 76, "start": 173.666, "end": 175.766, "text": "项目提供两套执行模式", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 77, "start": 176.2, "end": 177.966, "text": "Flash模式追求速度", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 78, "start": 178.133, "end": 179.566, "text": "采用观察一次", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 79, "start": 179.666, "end": 180.733, "text": "行动一次", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 80, "start": 180.866, "end": 182.5, "text": "再次观察的循环", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 81, "start": 183.1, "end": 186.733, "text": "readme给出的典型速度是每一步3-5秒", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 82, "start": 186.8, "end": 189.066, "text": "它不会先写一份庞大计划", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 83, "start": 189.066, "end": 191.933, "text": "也不会为每个动作生成长篇说明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 84, "start": 192.066, "end": 193.4, "text": "适合目标清楚", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 85, "start": 193.666, "end": 195.133, "text": "流程较短的任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 86, "start": 195.3, "end": 197.2, "text": "打开设置并关闭蓝牙", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 87, "start": 197.7, "end": 199.733, "text": "Flash模式看到当前界面后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 88, "start": 199.733, "end": 201.7, "text": "会直接寻找设置入口", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 89, "start": 202.166, "end": 204.066, "text": "如果已经位于快捷面板", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 90, "start": 204.066, "end": 206.4, "text": "它不必坚持原先设想的路径", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 91, "start": 206.733, "end": 208.7, "text": "反应式循环的好处是", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 92, "start": 208.866, "end": 211.5, "text": "灵活界面出现意外弹窗时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 93, "start": 211.5, "end": 212.866, "text": "也能临时调整", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 94, "start": 213.133, "end": 215.5, "text": "速度的代价是缺少全局检查", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 95, "start": 215.766, "end": 217.566, "text": "任务变成长流程后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 96, "start": 217.7, "end": 220.333, "text": "前面一个小错误可能一直带到最后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 97, "start": 220.466, "end": 222.333, "text": "代理以为邮件已经发送", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 98, "start": 222.366, "end": 224.0, "text": "实际停在草稿页", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 99, "start": 224.3, "end": 226.066, "text": "他以为文件已经移动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 100, "start": 226.1, "end": 228.2, "text": "实际只打开了目标文件夹", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 101, "start": 228.7, "end": 230.2, "text": "每一步看起来都合理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 102, "start": 230.3, "end": 232.6, "text": "最终状态却不符合要求", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 103, "start": 232.9, "end": 235.166, "text": "Pro模式就是为这种任务准备的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 104, "start": 235.733, "end": 237.533, "text": "他使用多代理图把职责", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 105, "start": 237.533, "end": 239.966, "text": "分给planner operator和Checker", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 106, "start": 240.4, "end": 242.333, "text": "planner维护一份持续", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 107, "start": 242.333, "end": 244.7, "text": "更新的任务计划operator", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 108, "start": 245.366, "end": 247.966, "text": "操作设备checker在关键", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 109, "start": 247.966, "end": 250.266, "text": "节点和结束时检查结果", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 110, "start": 250.3, "end": 252.566, "text": "一次测试超过100步也可以", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 111, "start": 252.566, "end": 254.733, "text": "保留阶段状态继续执行", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 112, "start": 254.966, "end": 256.533, "text": "planner并不只是把", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 113, "start": 256.533, "end": 258.666, "text": "任务写成编号列表", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 114, "start": 259.2, "end": 262.066, "text": "计划里会标出需要验证的里程碑", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 115, "start": 262.9, "end": 265.1, "text": "例如联系人是否真的创建", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 116, "start": 265.533, "end": 268.366, "text": "文件是否已经出现在目标目录", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 117, "start": 268.9, "end": 271.533, "text": "设置开关是否保持正确状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 118, "start": 271.766, "end": 273.166, "text": "界面发生变化后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 119, "start": 273.166, "end": 274.5, "text": "计划也能调整", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 120, "start": 274.766, "end": 276.966, "text": "而不是机械重复最初路线", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 121, "start": 277.2, "end": 280.166, "text": "operator拿到计划后调用完整工具集", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 122, "start": 280.466, "end": 282.166, "text": "它可以读取屏幕结构", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 123, "start": 282.3, "end": 283.366, "text": "查看截图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 124, "start": 283.466, "end": 284.533, "text": "分析录屏", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 125, "start": 284.666, "end": 286.133, "text": "调用ADB诊断", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 126, "start": 286.9, "end": 288.3, "text": "也可以搜索此前", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 127, "start": 288.3, "end": 290.066, "text": "压缩保存的操作历史", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 128, "start": 290.266, "end": 291.4, "text": "遇到失败时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 129, "start": 291.4, "end": 293.666, "text": "系统会创建一条执行事件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 130, "start": 294.0, "end": 296.933, "text": "把失败原因保留在后续上下文中", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 131, "start": 297.133, "end": 299.2, "text": "直到某次操作真正恢复", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 132, "start": 299.5, "end": 300.7, "text": "checker保持只读", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 133, "start": 300.7, "end": 302.666, "text": "不负责替operator点击", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 134, "start": 302.8, "end": 304.9, "text": "这样设计是为了避免执行者", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 135, "start": 304.9, "end": 307.566, "text": "一边操作一边宣布自己成功", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 136, "start": 308.133, "end": 310.866, "text": "checker根据任务目标检查最终状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 137, "start": 310.9, "end": 313.2, "text": "也可以在中间里程碑做验证", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 138, "start": 313.4, "end": 315.466, "text": "用户还能选择关闭验证", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 139, "start": 315.7, "end": 317.1, "text": "只做最终验证", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 140, "start": 317.4, "end": 318.733, "text": "检查所有节点", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 141, "start": 319.2, "end": 321.066, "text": "或者使用更严格的模式", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 142, "start": 321.133, "end": 323.9, "text": "每个动作执行前还有一道safety net", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 143, "start": 324.166, "end": 327.866, "text": "系统优先根据XML界面树确认目标", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 144, "start": 328.066, "end": 330.966, "text": "结构信息不足时再检查像素按钮", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 145, "start": 331.133, "end": 333.4, "text": "在模型思考期间已经消失", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 146, "start": 333.5, "end": 336.0, "text": "操作就会被阻止并交回operator", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 147, "start": 336.466, "end": 338.7, "text": "移动界面中的状态变化很快", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 148, "start": 338.8, "end": 340.466, "text": "这道检查可以减少点", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 149, "start": 340.466, "end": 341.933, "text": "到错误位置的情况", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 150, "start": 342.133, "end": 344.766, "text": "Artemis还支持连续动作突发", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 151, "start": 345.333, "end": 347.4, "text": "某些菜单只会短暂显示", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 152, "start": 347.6, "end": 350.6, "text": "如果每点一次都等待模型重新观察", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 153, "start": 350.766, "end": 352.0, "text": "菜单早已关闭", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 154, "start": 352.466, "end": 355.366, "text": "系统可以把几个确定动作连续发出", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 155, "start": 355.466, "end": 357.266, "text": "缩短模型往返延迟", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 156, "start": 357.4, "end": 359.0, "text": "这类工程处理很少", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 157, "start": 359.0, "end": 360.933, "text": "出现在产品宣传中", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 158, "start": 361.3, "end": 362.733, "text": "却直接决定代理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 159, "start": 362.733, "end": 364.933, "text": "能不能在真实应用里工作", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 160, "start": 365.2, "end": 367.733, "text": "Flash模式每步大约3-5秒", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 161, "start": 367.866, "end": 370.4, "text": "Pro模式通常需要15-40秒", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 162, "start": 370.533, "end": 372.7, "text": "速度差距提醒了一个现实", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 163, "start": 372.8, "end": 374.566, "text": "可靠性不是免费的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 164, "start": 374.566, "end": 376.866, "text": "规划检查历史检索", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 165, "start": 377.133, "end": 379.966, "text": "和故障恢复都会增加模型调用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 166, "start": 380.0, "end": 381.5, "text": "短任务追求速度", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 167, "start": 381.533, "end": 383.0, "text": "长任务追求结果", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 168, "start": 383.166, "end": 386.2, "text": "两种场景很难用同一套配置解决", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 169, "start": 386.333, "end": 389.533, "text": "项目最值得关注的用途是软件测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 170, "start": 390.2, "end": 393.1, "text": "普通人远程控制手机还排在后面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 171, "start": 393.2, "end": 395.1, "text": "开发者完成一项功能后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 172, "start": 395.1, "end": 397.933, "text": "可以让Artemis在真机上复现问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 173, "start": 398.1, "end": 400.3, "text": "收集Logcat日志和截图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 174, "start": 400.366, "end": 402.466, "text": "再把结果交回编程代理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 175, "start": 402.5, "end": 405.733, "text": "过去分开的写代码跑手机看日志", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 176, "start": 405.933, "end": 408.133, "text": "开始被串成同一条工作流", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 177, "start": 408.166, "end": 410.6, "text": "这里的关键接口是MCP", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 178, "start": 410.966, "end": 413.1, "text": "Optimus自带MCP服务器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 179, "start": 413.1, "end": 415.7, "text": "可以连接Antigravity Claude code", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 180, "start": 415.966, "end": 418.766, "text": "Windsurf等支持MCP的开发工具", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 181, "start": 419.1, "end": 421.8, "text": "编程代理不再只能阅读代码仓库", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 182, "start": 422.3, "end": 423.5, "text": "他还可以看到应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 183, "start": 423.5, "end": 425.2, "text": "实际运行后的界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 184, "start": 425.3, "end": 427.9, "text": "并操作设备验证自己的修改", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 185, "start": 428.1, "end": 429.666, "text": "设想一个常见场景", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 186, "start": 429.8, "end": 432.6, "text": "用户报告结账按钮点了没有反应", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 187, "start": 432.8, "end": 434.933, "text": "开发者只拿到一句模糊描述", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 188, "start": 435.166, "end": 436.533, "text": "接入Artemis后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 189, "start": 436.566, "end": 438.8, "text": "编程代理可以安装测试版本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 190, "start": 439.166, "end": 441.266, "text": "按照用户路径进入商品页", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 191, "start": 441.333, "end": 442.2, "text": "触发结账", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 192, "start": 442.566, "end": 443.9, "text": "截取错误界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 193, "start": 443.966, "end": 445.866, "text": "再读取Logcat中的异常", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 194, "start": 446.0, "end": 447.666, "text": "它修改代码后还能", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 195, "start": 447.666, "end": 449.4, "text": "重新执行同一流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 196, "start": 449.533, "end": 451.766, "text": "这会改变修复完成的定义", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 197, "start": 452.0, "end": 454.8, "text": "很多AI编程工具生成代码后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 198, "start": 454.933, "end": 456.7, "text": "只能运行单元测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 199, "start": 457.733, "end": 459.133, "text": "单元测试通过", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 200, "start": 459.266, "end": 461.533, "text": "不代表手机界面真的可用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 201, "start": 461.733, "end": 463.1, "text": "按钮可能被遮挡", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 202, "start": 463.3, "end": 465.366, "text": "权限弹窗可能阻断流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 203, "start": 465.666, "end": 467.933, "text": "键盘弹出后布局也可能错位", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 204, "start": 468.2, "end": 470.966, "text": "Artemis把验证推进到设备层", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 205, "start": 471.3, "end": 474.2, "text": "项目还提供CLI Python SDK", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 206, "start": 474.3, "end": 476.066, "text": "和本地网页控制台", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 207, "start": 476.266, "end": 478.8, "text": "CLI适合在终端直接发起任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 208, "start": 479.0, "end": 481.133, "text": "Python SDK可以接入pytest", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 209, "start": 481.133, "end": 483.333, "text": "和CI流程网页控制台", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 210, "start": 483.333, "end": 484.766, "text": "则显示手机画面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 211, "start": 484.766, "end": 486.6, "text": "推理轨迹和执行过程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 212, "start": 486.766, "end": 488.366, "text": "失败任务可以回放", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 213, "start": 489.0, "end": 491.733, "text": "测试人员不必只看一句执行失败", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 214, "start": 491.966, "end": 494.366, "text": "快速启动脚本会准备ADB", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 215, "start": 494.7, "end": 498.1, "text": "script ffpeg和Python依赖用户", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 216, "start": 498.466, "end": 500.166, "text": "可以连接开启USB", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 217, "start": 500.166, "end": 502.3, "text": "调试的实体Android设备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 218, "start": 502.4, "end": 503.966, "text": "也可以使用模拟器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 219, "start": 504.133, "end": 505.466, "text": "第一次运行时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 220, "start": 505.533, "end": 506.966, "text": "系统会在设备上", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 221, "start": 506.966, "end": 509.6, "text": "安装Artemis accessibility helper", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 222, "start": 509.866, "end": 511.733, "text": "用它读取屏幕布局", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 223, "start": 511.933, "end": 513.733, "text": "这个辅助服务会出现在", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 224, "start": 513.733, "end": 515.666, "text": "Android的无障碍设置中", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 225, "start": 515.666, "end": 517.566, "text": "并显示一条运行通知", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 226, "start": 518.2, "end": 519.4, "text": "项目说明称", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 227, "start": 519.466, "end": 521.4, "text": "它只在手机本地监听", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 228, "start": 521.4, "end": 524.3, "text": "不会自行把数据发送到其他位置", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 229, "start": 524.366, "end": 526.0, "text": "如果辅助服务失效", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 230, "start": 526.066, "end": 528.333, "text": "系统还能回退到uiautomator", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 231, "start": 528.566, "end": 530.133, "text": "这里要分清本地", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 232, "start": 530.133, "end": 532.066, "text": "辅助服务和模型调用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 233, "start": 532.533, "end": 534.0, "text": "前者可以本地通信", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 234, "start": 534.166, "end": 536.2, "text": "截图及界面信息是否离开", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 235, "start": 536.2, "end": 540.0, "text": "设备仍取决于模型供应商配置Athena", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 236, "start": 540.0, "end": 544.933, "text": "支持Gemini Claude GPT4O和千问VL等模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 237, "start": 545.733, "end": 548.266, "text": "它本身更接近一个代理框架", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 238, "start": 548.4, "end": 550.266, "text": "不是新的基础模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 239, "start": 550.4, "end": 552.666, "text": "实际表现会受到所选模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 240, "start": 552.8, "end": 555.8, "text": "网络设备性能和应用界面影响", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 241, "start": 555.933, "end": 557.333, "text": "换一个便宜模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 242, "start": 557.6, "end": 558.933, "text": "成本可能下降", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 243, "start": 559.366, "end": 561.466, "text": "定位和规划能力也可能变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 244, "start": 561.6, "end": 564.366, "text": "这也是99.1%跑分需要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 245, "start": 564.366, "end": 565.7, "text": "谨慎理解的原因", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 246, "start": 565.933, "end": 568.3, "text": "项目宣称自己在Android world", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 247, "start": 568.466, "end": 571.366, "text": "达到99%以上任务完成率", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 248, "start": 571.5, "end": 575.166, "text": "图表给出的具体数字是99.1%", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 249, "start": 575.3, "end": 577.4, "text": "Android world是Google research", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 250, "start": 577.4, "end": 578.766, "text": "推出的公开环境", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 251, "start": 579.266, "end": 581.9, "text": "包含116项手工设计任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 252, "start": 582.366, "end": 583.766, "text": "覆盖20个应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 253, "start": 583.8, "end": 586.666, "text": "并通过随机参数生成大量变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 254, "start": 586.8, "end": 589.3, "text": "任务包括创建联系人", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 255, "start": 589.533, "end": 590.533, "text": "发送短信", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 256, "start": 590.6, "end": 591.966, "text": "添加日历事件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 257, "start": 592.1, "end": 594.3, "text": "调整亮度开关Wifi", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 258, "start": 594.366, "end": 596.9, "text": "移动文件和处理播放列表", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 259, "start": 597.0, "end": 599.766, "text": "任务程序会检查设备最终状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 260, "start": 599.9, "end": 602.666, "text": "不接受模型自己宣布完成了", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 261, "start": 603.0, "end": 605.866, "text": "这比依靠文字答案评分更可靠", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 262, "start": 605.966, "end": 607.933, "text": "也更接近真实操作", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 263, "start": 608.1, "end": 612.766, "text": "99.1%依然不等于现实手机上的99.1%", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 264, "start": 613.0, "end": 615.4, "text": "Android world使用固定系统版本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 265, "start": 615.466, "end": 617.7, "text": "预设应用和明确任务类型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 266, "start": 618.333, "end": 620.8, "text": "真实手机有厂商定制界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 267, "start": 620.866, "end": 624.2, "text": "广告弹窗登录验证地区差异", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 268, "start": 624.2, "end": 626.366, "text": "和不断变化的第三方应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 269, "start": 626.533, "end": 628.566, "text": "测试集能衡量代理进步", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 270, "start": 628.8, "end": 631.366, "text": "无法覆盖开放世界的全部混乱", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 271, "start": 631.466, "end": 634.0, "text": "这项成绩目前也属于团队自曝", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 272, "start": 634.533, "end": 636.866, "text": "公开排行榜接受社区提交", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 273, "start": 637.0, "end": 638.5, "text": "并不会替每个项目", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 274, "start": 638.5, "end": 640.3, "text": "完整复现所有运行", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 275, "start": 640.566, "end": 642.933, "text": "Artemis仓库展示了对比图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 276, "start": 643.066, "end": 645.366, "text": "却没有提供足够完整的第三方", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 277, "start": 645.366, "end": 648.133, "text": "审计材料来证明每一次任务如何", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 278, "start": 648.133, "end": 652.533, "text": "运行 99.1%可以称为公开的项目成绩", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 279, "start": 653.1, "end": 655.566, "text": "不能包装成已经独立认证的事实", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 280, "start": 655.9, "end": 658.066, "text": "基准测试还存在适配风险", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 281, "start": 658.6, "end": 661.4, "text": "团队知道全部116项任务后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 282, "start": 661.4, "end": 663.1, "text": "可以针对常见应用和", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 283, "start": 663.1, "end": 665.0, "text": "失败模式优化提示词", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 284, "start": 665.2, "end": 666.733, "text": "恢复逻辑和工具", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 285, "start": 667.066, "end": 668.666, "text": "优化并不等于作弊", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 286, "start": 669.1, "end": 671.933, "text": "工程系统本来就该改进薄弱点", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 287, "start": 672.0, "end": 673.566, "text": "但一个系统在熟悉", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 288, "start": 673.566, "end": 675.466, "text": "测试集上接近满分", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 289, "start": 675.533, "end": 677.333, "text": "不代表面对从未见过", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 290, "start": 677.333, "end": 679.166, "text": "的应用也有同样表现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 291, "start": 679.266, "end": 681.9, "text": "更敏感的问题来自项目的代码来源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 292, "start": 682.6, "end": 683.966, "text": "Artemis公开后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 293, "start": 684.1, "end": 686.5, "text": "minitab团队在Git Hub提出异议", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 294, "start": 686.7, "end": 688.466, "text": "称仓库包含其开源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 295, "start": 688.466, "end": 690.2, "text": "项目mobileuse的代码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 296, "start": 690.266, "end": 692.166, "text": "却没有在最初版本中清楚", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 297, "start": 692.166, "end": 694.133, "text": "保留来源和贡献者信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 298, "start": 694.5, "end": 696.5, "text": "minitype列出的证据很具体", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 299, "start": 696.666, "end": 699.366, "text": "早期PI project Tamo曾出现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 300, "start": 699.366, "end": 702.066, "text": "Mobile use的版本号3.6.3", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 301, "start": 702.133, "end": 704.066, "text": "以及三位原作者姓名", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 302, "start": 704.5, "end": 706.666, "text": "后续提交替换了作者信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 303, "start": 706.966, "end": 709.066, "text": "其他大量文件保持不变", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 304, "start": 709.133, "end": 711.9, "text": "双方共享的内容涉及ADB隧道", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 305, "start": 711.933, "end": 714.366, "text": "规划辅助函数应用锁定", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 306, "start": 714.366, "end": 717.533, "text": "工具代理提示词和SDK示例", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 307, "start": 717.7, "end": 719.5, "text": "开源代码可以复用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 308, "start": 719.5, "end": 722.933, "text": "但Apache2.0并不等于可以抹掉来源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 309, "start": 723.333, "end": 726.0, "text": "许可证允许修改和商业使用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 310, "start": 726.066, "end": 729.266, "text": "同时要求保留相应版权及许可说明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 311, "start": 729.4, "end": 730.866, "text": "争议焦点不在Google", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 312, "start": 730.866, "end": 732.4, "text": "能不能使用这些代码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 313, "start": 732.566, "end": 735.466, "text": "而在它是否完整履行了署名义务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 314, "start": 735.666, "end": 737.366, "text": "以及发布过程为什么", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 315, "start": 737.366, "end": 739.366, "text": "先移除了原作者信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 316, "start": 739.6, "end": 741.8, "text": "当前仓库已经补充说明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 317, "start": 742.266, "end": 743.866, "text": "明确写着项目包含", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 318, "start": 743.866, "end": 745.733, "text": "minitab开发的源代码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 319, "start": 745.966, "end": 748.966, "text": "相关文件也增加了版权和来源标记", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 320, "start": 749.5, "end": 752.1, "text": "提出问题的Git Hub议题已经关闭", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 321, "start": 752.266, "end": 755.7, "text": "这解决了当前readme没有署名的问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 322, "start": 755.866, "end": 759.066, "text": "却无法抹去最初发布时发生过什么", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 323, "start": 759.066, "end": 761.066, "text": "这场争议让Artemis的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 324, "start": 761.066, "end": 762.6, "text": "故事变得更有价值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 325, "start": 762.866, "end": 764.866, "text": "大型公司开源项目时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 326, "start": 764.933, "end": 766.9, "text": "外界往往默认代码已经", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 327, "start": 766.9, "end": 769.1, "text": "经过完善法务和来源审查", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 328, "start": 769.5, "end": 770.766, "text": "实际情况说明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 329, "start": 770.8, "end": 773.366, "text": "即使仓库挂在Google组织名下", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 330, "start": 773.366, "end": 775.3, "text": "也需要查看提交记录", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 331, "start": 775.533, "end": 777.133, "text": "notice和文件头", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 332, "start": 777.333, "end": 779.3, "text": "品牌不能代替开源合规", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 333, "start": 779.466, "end": 781.933, "text": "跑分争议又让问题多了一层", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 334, "start": 782.666, "end": 784.6, "text": "Mini Tab表示自己的mobileuse", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 335, "start": 784.6, "end": 786.733, "text": "后来曾自测到更高层级", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 336, "start": 787.0, "end": 789.366, "text": "但公开榜单没有及时更新", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 337, "start": 789.533, "end": 792.333, "text": "Artemis的图表列出多个竞争项目", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 338, "start": 792.5, "end": 795.533, "text": "却没有清楚呈现双方最新可比结果", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 339, "start": 795.8, "end": 797.4, "text": "这里无法仅凭一方说法", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 340, "start": 797.4, "end": 799.366, "text": "判断榜单维护者的动机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 341, "start": 799.533, "end": 801.3, "text": "但可以确认排行榜存在", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 342, "start": 801.3, "end": 803.2, "text": "更新和独立验证不足", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 343, "start": 803.3, "end": 804.966, "text": "这些问题不应被用来", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 344, "start": 804.966, "end": 807.666, "text": "否定Artemis的全部技术价值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 345, "start": 808.5, "end": 809.866, "text": "仓库不是一个空壳", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 346, "start": 810.166, "end": 812.2, "text": "里面有完整的命令行工具", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 347, "start": 812.866, "end": 813.933, "text": "MCP服务器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 348, "start": 814.2, "end": 815.266, "text": "设备驱动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 349, "start": 815.4, "end": 816.6, "text": "网页控制台", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 350, "start": 816.966, "end": 818.933, "text": "测试目录和多代理流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 351, "start": 819.1, "end": 820.9, "text": "短时间获得大量关注", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 352, "start": 820.933, "end": 822.366, "text": "也说明开发者确实", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 353, "start": 822.366, "end": 824.133, "text": "需要更可靠的移动端", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 354, "start": 824.333, "end": 825.5, "text": "AI自动化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 355, "start": 825.5, "end": 827.0, "text": "同样不能因为代码能", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 356, "start": 827.0, "end": 828.966, "text": "运行就忽略来源问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 357, "start": 829.133, "end": 830.366, "text": "技术能力和开源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 358, "start": 830.366, "end": 832.2, "text": "合规是两条独立标准", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 359, "start": 832.5, "end": 834.266, "text": "一个项目可以有实用价值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 360, "start": 834.3, "end": 837.333, "text": "同时在发布流程中犯下严重错误", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 361, "start": 837.5, "end": 840.666, "text": "把任何一边省略都会让判断失真", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 362, "start": 840.866, "end": 842.333, "text": "Artemis当前的成熟", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 363, "start": 842.333, "end": 844.3, "text": "边界也写在路线图里", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 364, "start": 844.9, "end": 847.533, "text": "Android studio原生插件仍未完成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 365, "start": 847.8, "end": 849.933, "text": "ios扩展还在计划中", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 366, "start": 850.166, "end": 852.4, "text": "端侧轻量视觉模型和实时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 367, "start": 852.4, "end": 854.9, "text": "双向语音控制也没有正式交付", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 368, "start": 855.166, "end": 858.3, "text": "现在可用的是Android设备自动化框架", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 369, "start": 858.566, "end": 861.466, "text": "不是覆盖所有移动平台的万能代理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 370, "start": 862.133, "end": 864.5, "text": "ios迟早会成为重要方向", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 371, "start": 864.566, "end": 866.666, "text": "但难度不会只是把ADB", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 372, "start": 866.666, "end": 867.933, "text": "换成另一套命令", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 373, "start": 868.166, "end": 870.466, "text": "苹果系统对自动化后台", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 374, "start": 870.466, "end": 872.666, "text": "运行和权限控制更严格", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 375, "start": 872.866, "end": 875.6, "text": "实体设备与模拟器的行为也不同", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 376, "start": 876.4, "end": 878.366, "text": "项目路线图写了ios", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 377, "start": 878.5, "end": 881.366, "text": "不代表现有架构已经解决这些限制", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 378, "start": 881.666, "end": 884.2, "text": "端侧视觉模型同样值得关注", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 379, "start": 884.533, "end": 886.766, "text": "今天使用云端多模态模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 380, "start": 887.066, "end": 888.4, "text": "截图和界面信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 381, "start": 888.4, "end": 890.1, "text": "可能需要传到服务商", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 382, "start": 890.4, "end": 892.3, "text": "延迟和隐私都受影响", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 383, "start": 892.5, "end": 894.466, "text": "轻量模型如果能在手机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 384, "start": 894.466, "end": 896.8, "text": "或开发机本地识别界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 385, "start": 896.866, "end": 898.666, "text": "可以减少数据外传", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 386, "start": 898.733, "end": 901.066, "text": "也能缩短每一步等待时间", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 387, "start": 901.5, "end": 903.333, "text": "代价是模型能力下降", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 388, "start": 903.566, "end": 906.1, "text": "复杂页面的识别可能不够稳定", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 389, "start": 906.2, "end": 908.4, "text": "Artemis要进入企业测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 390, "start": 908.6, "end": 910.733, "text": "还要解决权限和隔离", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 391, "start": 911.133, "end": 913.766, "text": "测试手机可能登录内部账号", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 392, "start": 913.933, "end": 917.266, "text": "包含未发布功能和真实业务数据", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 393, "start": 917.5, "end": 921.3, "text": "代理拥有ADB截图和日志读取能力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 394, "start": 921.9, "end": 923.333, "text": "一旦配置错误", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 395, "start": 923.733, "end": 926.6, "text": "泄露范围会比普通测试脚本更大", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 396, "start": 926.733, "end": 929.0, "text": "企业需要限制模型端点", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 397, "start": 929.0, "end": 931.133, "text": "保存周期和可执行命令", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 398, "start": 931.466, "end": 932.566, "text": "提示词注入也是", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 399, "start": 932.566, "end": 934.3, "text": "移动代理绕不开的风险", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 400, "start": 934.7, "end": 935.933, "text": "应用页面可以显示", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 401, "start": 935.933, "end": 938.133, "text": "一段专门写给AI的文字", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 402, "start": 938.333, "end": 940.166, "text": "诱导它忽略测试目标", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 403, "start": 940.266, "end": 942.533, "text": "打开其他应用或泄露信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 404, "start": 942.933, "end": 944.933, "text": "人类测试员看到异常文字", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 405, "start": 944.966, "end": 946.766, "text": "会判断模型可能把它", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 406, "start": 946.766, "end": 948.4, "text": "当成高优先级指令", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 407, "start": 948.7, "end": 950.933, "text": "safety net能检查目标位置", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 408, "start": 951.1, "end": 953.866, "text": "未必能识别内容本身的恶意意图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 409, "start": 954.0, "end": 956.733, "text": "持续集成环境还需要可复现性", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 410, "start": 956.933, "end": 959.466, "text": "同一个自然语言任务今天通过", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 411, "start": 959.466, "end": 960.466, "text": "明天失败", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 412, "start": 961.066, "end": 963.133, "text": "开发团队很难判断是代码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 413, "start": 963.133, "end": 966.166, "text": "回归模型波动还是界面变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 414, "start": 966.366, "end": 969.266, "text": "Optimus必须记录模型版本提示", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 415, "start": 969.266, "end": 972.766, "text": "词截图动作轨迹和设备状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 416, "start": 972.933, "end": 975.666, "text": "才能让失败结果具有工程价值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 417, "start": 976.0, "end": 979.5, "text": "成本也不能被99.1%的数字遮住", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 418, "start": 979.866, "end": 982.7, "text": "Pro模式每一步可能消耗15-40秒", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 419, "start": 982.866, "end": 985.933, "text": "还会调用规划视觉和检查模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 420, "start": 986.0, "end": 987.766, "text": "一次上百步的测试要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 421, "start": 987.766, "end": 989.933, "text": "花多少时间和API费用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 422, "start": 990.1, "end": 992.566, "text": "决定它能否替代传统自动化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 423, "start": 992.7, "end": 995.7, "text": "高价值探索测试可以接受较高成本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 424, "start": 995.7, "end": 998.866, "text": "大量重复回归测试未必划算", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 425, "start": 999.366, "end": 1002.1, "text": "最合理的使用方式可能是混合测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 426, "start": 1002.5, "end": 1004.4, "text": "稳定高频的核心流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 427, "start": 1004.4, "end": 1006.333, "text": "继续使用传统脚本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 428, "start": 1006.466, "end": 1008.9, "text": "速度快结果也容易复现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 429, "start": 1009.1, "end": 1010.866, "text": "经常变化或难以维护", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 430, "start": 1010.866, "end": 1013.133, "text": "的界面交给Artemis探索", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 431, "start": 1013.533, "end": 1015.366, "text": "AI发现稳定路径后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 432, "start": 1015.366, "end": 1018.5, "text": "还可以帮助生成更固定的测试用例", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 433, "start": 1018.566, "end": 1021.0, "text": "这样分工比AI取代所有", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 434, "start": 1021.0, "end": 1023.133, "text": "测试工程师更接近现实", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 435, "start": 1023.333, "end": 1025.7, "text": "测试人员需要定义风险准备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 436, "start": 1025.7, "end": 1028.333, "text": "环境判断结果和处理异常", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 437, "start": 1028.733, "end": 1031.933, "text": "Artemis减少的是反复找按钮维护", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 438, "start": 1031.933, "end": 1034.366, "text": "选择器和手工复现的时间", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 439, "start": 1034.666, "end": 1037.566, "text": "并没有替团队决定什么功能最重要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 440, "start": 1037.7, "end": 1040.3, "text": "也不能替公司承担发布责任", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 441, "start": 1040.666, "end": 1043.3, "text": "测试工程师的工作会随之变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 442, "start": 1043.466, "end": 1046.2, "text": "过去大量时间花在脚本维护", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 443, "start": 1046.466, "end": 1049.2, "text": "未来更多时间会用于设计任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 444, "start": 1049.3, "end": 1052.566, "text": "设置验证条件和分析失败轨迹", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 445, "start": 1052.8, "end": 1054.8, "text": "自然语言看似降低了门槛", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 446, "start": 1054.933, "end": 1057.266, "text": "准确描述目标反而更重要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 447, "start": 1057.566, "end": 1059.6, "text": "含糊任务会让代理找到", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 448, "start": 1059.6, "end": 1061.5, "text": "一条看起来完成的路径", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 449, "start": 1061.766, "end": 1063.8, "text": "却漏掉真正的业务要求", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 450, "start": 1063.966, "end": 1066.7, "text": "开发流程也可能从提交代码后统一", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 451, "start": 1066.7, "end": 1069.9, "text": "测试变成编程代理边写边操作设备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 452, "start": 1070.133, "end": 1071.666, "text": "它修改登录页面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 453, "start": 1071.966, "end": 1073.566, "text": "立刻安装构建版本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 454, "start": 1073.566, "end": 1075.666, "text": "走一遍登录和找回密码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 455, "start": 1075.666, "end": 1077.466, "text": "再根据日志继续修改", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 456, "start": 1077.666, "end": 1079.2, "text": "设备不再是开发链", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 457, "start": 1079.2, "end": 1080.966, "text": "末端的人工检查站", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 458, "start": 1081.2, "end": 1083.866, "text": "而是AI编码循环的一部分", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 459, "start": 1083.966, "end": 1086.366, "text": "这会给小团队带来明显价值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 460, "start": 1086.8, "end": 1089.133, "text": "没有专职移动测试工程师的团队", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 461, "start": 1089.366, "end": 1091.8, "text": "也能建立基本的真机回归流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 462, "start": 1091.866, "end": 1093.466, "text": "可规模扩大以后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 463, "start": 1093.6, "end": 1096.333, "text": "设备农场账号管理和并发", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 464, "start": 1096.333, "end": 1098.966, "text": "调度仍然需要专业基础设施", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 465, "start": 1099.266, "end": 1101.5, "text": "optimus解决了操作智能", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 466, "start": 1101.733, "end": 1104.666, "text": "不会自动解决所有测试平台问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 467, "start": 1105.0, "end": 1107.7, "text": "设备管理是一项容易被忽略的成本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 468, "start": 1107.9, "end": 1109.666, "text": "几十台手机同时测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 469, "start": 1109.9, "end": 1111.6, "text": "需要保持系统版本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 470, "start": 1112.066, "end": 1113.5, "text": "屏幕分辨率和应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 471, "start": 1113.5, "end": 1115.133, "text": "数据处于可控状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 472, "start": 1115.3, "end": 1117.533, "text": "某项测试结束后还要清理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 473, "start": 1117.533, "end": 1120.866, "text": "联系人短信文件和登录状态", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 474, "start": 1121.133, "end": 1122.7, "text": "否则下一次任务会", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 475, "start": 1122.7, "end": 1124.6, "text": "读取上一次留下的数据", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 476, "start": 1124.766, "end": 1127.0, "text": "Android world之所以能评测代理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 477, "start": 1127.333, "end": 1128.733, "text": "正是因为它为任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 478, "start": 1128.733, "end": 1131.1, "text": "准备初始化和结束检查", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 479, "start": 1131.3, "end": 1133.966, "text": "企业自己的应用没有现成评测器", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 480, "start": 1134.2, "end": 1136.333, "text": "团队必须定义什么叫成功", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 481, "start": 1136.566, "end": 1137.933, "text": "页面出现提交完成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 482, "start": 1137.933, "end": 1139.5, "text": "可能只是前端提示", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 483, "start": 1139.733, "end": 1141.9, "text": "后台订单未必真正创建", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 484, "start": 1142.066, "end": 1144.133, "text": "Optimus可以点击和观察", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 485, "start": 1144.666, "end": 1146.2, "text": "却需要业务系统提供", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 486, "start": 1146.2, "end": 1147.933, "text": "更可靠的验证信号", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 487, "start": 1148.1, "end": 1150.4, "text": "数据库记录接口状态和", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 488, "start": 1150.4, "end": 1152.333, "text": "设备画面要互相对照", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 489, "start": 1152.533, "end": 1154.166, "text": "才能减少假成功", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 490, "start": 1154.4, "end": 1156.933, "text": "自然语言任务也需要版本管理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 491, "start": 1157.266, "end": 1159.9, "text": "测试人员把创建一个联系人改成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 492, "start": 1159.9, "end": 1162.866, "text": "创建联系人并确认号码能够搜索到", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 493, "start": 1163.0, "end": 1164.6, "text": "验证范围已经变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 494, "start": 1165.0, "end": 1167.566, "text": "如果这段文字只存在聊天记录中", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 495, "start": 1167.666, "end": 1169.966, "text": "团队很难追踪回归标准", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 496, "start": 1170.2, "end": 1173.0, "text": "成熟用法需要把任务描述断言", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 497, "start": 1173.0, "end": 1175.9, "text": "和环境配置像代码一样保存", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 498, "start": 1175.933, "end": 1177.533, "text": "并进入评审流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 499, "start": 1177.733, "end": 1180.133, "text": "模型更新还会制造另一种波动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 500, "start": 1180.766, "end": 1182.666, "text": "供应商更换默认模型后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 501, "start": 1182.666, "end": 1184.866, "text": "同一句指令可能走不同路径", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 502, "start": 1185.133, "end": 1186.6, "text": "新模型更聪明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 503, "start": 1186.6, "end": 1189.266, "text": "却也可能更愿意尝试危险动作", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 504, "start": 1189.466, "end": 1191.333, "text": "团队要锁定模型版本", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 505, "start": 1191.4, "end": 1192.4, "text": "建立基线", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 506, "start": 1192.5, "end": 1195.366, "text": "并在升级前重新运行关键任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 507, "start": 1195.533, "end": 1197.533, "text": "传统脚本的变化来自代码", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 508, "start": 1197.966, "end": 1200.166, "text": "代理测试的变化还来自模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 509, "start": 1200.366, "end": 1202.266, "text": "Optimus的历史压缩机制", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 510, "start": 1202.266, "end": 1204.733, "text": "就是为长任务控制上下文", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 511, "start": 1204.866, "end": 1207.133, "text": "它不会一直保留每张旧截图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 512, "start": 1207.333, "end": 1208.733, "text": "而是把已经完成的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 513, "start": 1208.733, "end": 1211.0, "text": "步骤压缩成可搜索记录", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 514, "start": 1211.1, "end": 1214.566, "text": "否则100步任务会迅速塞满模型窗口", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 515, "start": 1214.7, "end": 1216.733, "text": "费用和延迟一起上升", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 516, "start": 1216.933, "end": 1218.4, "text": "压缩能延长任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 517, "start": 1218.7, "end": 1220.466, "text": "代价是某些视觉细节", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 518, "start": 1220.466, "end": 1222.1, "text": "可能在摘要中丢失", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 519, "start": 1222.333, "end": 1224.266, "text": "历史检索因此不能只", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 520, "start": 1224.266, "end": 1226.533, "text": "依靠一句之前已经完成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 521, "start": 1227.3, "end": 1229.166, "text": "系统需要保留关键证据", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 522, "start": 1229.5, "end": 1231.566, "text": "例如创建后的联系人页面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 523, "start": 1231.9, "end": 1234.566, "text": "移动后的文件路径和操作返回值", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 524, "start": 1235.066, "end": 1237.366, "text": "摘要负责告诉代理发生过什么", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 525, "start": 1237.466, "end": 1240.333, "text": "原始证据负责在出现争议时复核", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 526, "start": 1240.766, "end": 1242.066, "text": "两者缺一项", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 527, "start": 1242.333, "end": 1244.333, "text": "长流程都会变得不可信", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 528, "start": 1244.566, "end": 1247.133, "text": "录屏分析提供了另一条诊断线", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 529, "start": 1247.566, "end": 1248.8, "text": "界面一闪而过", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 530, "start": 1248.8, "end": 1251.066, "text": "单张截图可能抓不到问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 531, "start": 1251.333, "end": 1253.066, "text": "录屏可以重建动画", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 532, "start": 1253.133, "end": 1255.333, "text": "弹窗和键盘出现的顺序", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 533, "start": 1255.566, "end": 1258.066, "text": "Optimus把fempy列为依赖", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 534, "start": 1258.3, "end": 1260.533, "text": "就是为了处理这类时序信息", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 535, "start": 1260.733, "end": 1263.4, "text": "移动端bug经常只存在几百毫秒", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 536, "start": 1263.8, "end": 1266.1, "text": "日志里也未必留下明确错误", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 537, "start": 1266.2, "end": 1268.933, "text": "不过录屏会扩大隐私范围", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 538, "start": 1269.466, "end": 1272.2, "text": "测试账号的通知联系人头像", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 539, "start": 1272.2, "end": 1274.7, "text": "和输入内容都可能进入文件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 540, "start": 1275.333, "end": 1278.0, "text": "团队需要在采集前关闭无关通知", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 541, "start": 1278.266, "end": 1279.6, "text": "使用合成数据", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 542, "start": 1279.6, "end": 1281.566, "text": "并限制报告中的截图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 543, "start": 1281.733, "end": 1283.666, "text": "代理自动收集越方便", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 544, "start": 1283.866, "end": 1285.6, "text": "数据治理越不能依靠", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 545, "start": 1285.6, "end": 1287.366, "text": "测试人员临时判断", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 546, "start": 1287.566, "end": 1289.533, "text": "权限边界还涉及ADB", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 547, "start": 1289.8, "end": 1291.3, "text": "ADB能够安装应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 548, "start": 1291.3, "end": 1292.166, "text": "读取日志", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 549, "start": 1292.333, "end": 1293.133, "text": "启动组件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 550, "start": 1293.133, "end": 1295.0, "text": "并执行系统命令权限", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 551, "start": 1295.466, "end": 1297.166, "text": "远高于普通用户操作", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 552, "start": 1297.333, "end": 1299.6, "text": "Artemis如果连接生产手机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 553, "start": 1299.6, "end": 1300.8, "text": "风险非常高", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 554, "start": 1300.966, "end": 1302.933, "text": "合理部署应使用专门", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 555, "start": 1302.933, "end": 1305.133, "text": "测试设备和隔离账号", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 556, "start": 1305.166, "end": 1308.533, "text": "CI中的密钥也不能直接暴露给模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 557, "start": 1308.9, "end": 1311.0, "text": "开源在这里提供了现实优势", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 558, "start": 1311.266, "end": 1312.966, "text": "企业可以检查每个工具", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 559, "start": 1312.966, "end": 1314.666, "text": "到底调用了什么命令", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 560, "start": 1314.8, "end": 1317.4, "text": "也能删掉不需要的高风险能力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 561, "start": 1317.6, "end": 1320.5, "text": "闭源手机代理只给出服务界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 562, "start": 1320.6, "end": 1322.866, "text": "用户很难确认数据流向", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 563, "start": 1323.166, "end": 1326.5, "text": "Optimus把设备驱动和代理逻辑公开", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 564, "start": 1326.8, "end": 1328.466, "text": "至少让安全团队有", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 565, "start": 1328.466, "end": 1330.3, "text": "审计和改造的入口", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 566, "start": 1330.366, "end": 1332.466, "text": "开源也带来维护压力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 567, "start": 1332.9, "end": 1334.866, "text": "仓库没有正式版本标签", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 568, "start": 1335.166, "end": 1338.133, "text": "当前项目迭代集中在很短时间内", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 569, "start": 1338.7, "end": 1339.733, "text": "依赖很多", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 570, "start": 1339.933, "end": 1343.1, "text": "既有LANGGRAPH和多家模型SDK", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 571, "start": 1343.2, "end": 1347.1, "text": "也有ADB UI Automator和图像处理组件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 572, "start": 1347.533, "end": 1349.933, "text": "任何上游变化都可能影响安装", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 573, "start": 1350.133, "end": 1352.3, "text": "9,000多个星标代表关注度", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 574, "start": 1352.4, "end": 1354.366, "text": "不等于长期维护承诺", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 575, "start": 1354.533, "end": 1356.466, "text": "开发者在实际采用前", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 576, "start": 1356.566, "end": 1358.7, "text": "需要先跑自己的小型评测", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 577, "start": 1359.366, "end": 1361.8, "text": "选择10-20条最关键流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 578, "start": 1362.3, "end": 1365.466, "text": "分别记录成功率平均耗时", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 579, "start": 1365.666, "end": 1367.9, "text": "模型费用和人工干预次数", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 580, "start": 1368.0, "end": 1369.2, "text": "只看Android world", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 581, "start": 1369.766, "end": 1372.1, "text": "总分无法回答企业自己的登录", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 582, "start": 1372.1, "end": 1374.8, "text": "支付和权限页面能不能稳定运行", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 583, "start": 1374.966, "end": 1377.066, "text": "失败类型也应分开统计", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 584, "start": 1377.866, "end": 1380.466, "text": "模型找不到控件属于感知问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 585, "start": 1380.766, "end": 1382.933, "text": "计划走错属于推理问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 586, "start": 1383.566, "end": 1385.8, "text": "应用崩溃属于产品问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 587, "start": 1386.266, "end": 1388.366, "text": "网路超时则是环境问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 588, "start": 1388.466, "end": 1390.566, "text": "把所有失败压成一个百分比", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 589, "start": 1390.8, "end": 1393.8, "text": "团队不知道该修提示词换模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 590, "start": 1393.933, "end": 1395.6, "text": "还是修改应用本身", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 591, "start": 1395.733, "end": 1399.333, "text": "Atemis有时最有价值的结果是暴露", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 592, "start": 1399.333, "end": 1402.066, "text": "产品不利于自动化和无障碍访问", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 593, "start": 1402.7, "end": 1404.866, "text": "通过本身反倒没那么重要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 594, "start": 1404.933, "end": 1406.7, "text": "一个按钮没有语义标签", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 595, "start": 1406.966, "end": 1408.533, "text": "模型只能猜坐标", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 596, "start": 1409.1, "end": 1410.166, "text": "视障用户的屏幕", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 597, "start": 1410.166, "end": 1411.9, "text": "阅读器也可能无法识别", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 598, "start": 1412.0, "end": 1413.766, "text": "代理测试会反过来推动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 599, "start": 1413.766, "end": 1415.9, "text": "开发者补齐控件描述", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 600, "start": 1416.133, "end": 1418.466, "text": "最终改善真实用户体验", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 601, "start": 1418.5, "end": 1421.2, "text": "这让它与传统视觉测试形成区别", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 602, "start": 1422.0, "end": 1423.533, "text": "截图对比擅长发现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 603, "start": 1423.533, "end": 1425.9, "text": "颜色位置和布局变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 604, "start": 1426.333, "end": 1429.333, "text": "而Temis更擅长判断任务能不能完成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 605, "start": 1429.666, "end": 1432.366, "text": "前者告诉团队按钮移动了5个像素", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 606, "start": 1432.5, "end": 1433.933, "text": "后者告诉团队用户", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 607, "start": 1433.933, "end": 1435.6, "text": "已经无法完成退款", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 608, "start": 1435.766, "end": 1437.266, "text": "两类测试结合", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 609, "start": 1437.366, "end": 1438.733, "text": "覆盖面才完整", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 610, "start": 1438.933, "end": 1440.766, "text": "市场上也不只有Artemis", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 611, "start": 1440.766, "end": 1444.733, "text": "final run auto device droid run uitas", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 612, "start": 1445.0, "end": 1446.966, "text": "和多种computer use模型都", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 613, "start": 1446.966, "end": 1448.8, "text": "在争夺移动代理入口", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 614, "start": 1448.866, "end": 1451.4, "text": "他们选择不同模型和规划架构", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 615, "start": 1451.5, "end": 1452.966, "text": "公开程度也不同", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 616, "start": 1453.133, "end": 1455.3, "text": "99.1%的领先", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 617, "start": 1455.4, "end": 1457.1, "text": "如果不能被持续复现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 618, "start": 1457.2, "end": 1460.3, "text": "很快就会被下一轮模型和工具改写", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 619, "start": 1460.466, "end": 1463.1, "text": "Google开源这套框架还有生态意义", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 620, "start": 1463.5, "end": 1465.466, "text": "Android开发者可以围绕它增加", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 621, "start": 1465.466, "end": 1468.533, "text": "设备驱动行业任务和验证插件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 622, "start": 1468.8, "end": 1470.5, "text": "模型公司也能把它作为", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 623, "start": 1470.5, "end": 1472.6, "text": "计算机操作能力的试验场", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 624, "start": 1472.6, "end": 1475.5, "text": "一个共同框架如果形成移动代理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 625, "start": 1475.566, "end": 1477.3, "text": "就不必每家公司都从", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 626, "start": 1477.3, "end": 1479.366, "text": "截图和点击重新造轮子", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 627, "start": 1479.366, "end": 1482.166, "text": "可共同框架必须建立可信治理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 628, "start": 1482.6, "end": 1484.6, "text": "贡献代码需要清楚来源", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 629, "start": 1484.866, "end": 1487.166, "text": "排行榜需要公开测试条件", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 630, "start": 1487.333, "end": 1489.666, "text": "安全漏洞需要有响应流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 631, "start": 1489.8, "end": 1491.566, "text": "minitab争议已经说明", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 632, "start": 1491.7, "end": 1493.566, "text": "技术社区不仅看功能", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 633, "start": 1493.566, "end": 1496.4, "text": "也看维护者如何处理他人的劳动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 634, "start": 1496.466, "end": 1498.866, "text": "Google后来补上署名是必要修正", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 635, "start": 1498.966, "end": 1501.0, "text": "公开解释仍然同样重要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 636, "start": 1501.533, "end": 1504.333, "text": "Artemis未来若接入Android studio", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 637, "start": 1504.666, "end": 1507.266, "text": "开发者可以在IDE里选中一个界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 638, "start": 1507.366, "end": 1509.566, "text": "直接生成并运行设备任务", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 639, "start": 1510.1, "end": 1511.766, "text": "这个设想很有吸引力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 640, "start": 1511.766, "end": 1513.8, "text": "但插件目前还在路线图", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 641, "start": 1513.866, "end": 1516.666, "text": "现阶段仍要通过命令行网页", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 642, "start": 1516.666, "end": 1519.266, "text": "控制台或MCP连接部署体验", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 643, "start": 1519.466, "end": 1520.9, "text": "更适合愿意处理", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 644, "start": 1520.9, "end": 1522.666, "text": "环境配置的技术团队", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 645, "start": 1522.9, "end": 1525.0, "text": "实时语音也尚未完成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 646, "start": 1525.4, "end": 1526.566, "text": "即使以后上线", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 647, "start": 1526.8, "end": 1528.1, "text": "它的意义也不只是", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 648, "start": 1528.1, "end": 1529.933, "text": "说一句话控制手机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 649, "start": 1530.566, "end": 1532.533, "text": "双向语音要求代理在", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 650, "start": 1532.533, "end": 1534.5, "text": "操作过程中报告进度", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 651, "start": 1535.333, "end": 1537.1, "text": "用户能够随时打断", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 652, "start": 1537.2, "end": 1539.1, "text": "并在敏感步骤确认", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 653, "start": 1539.166, "end": 1540.8, "text": "没有可靠中断机制", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 654, "start": 1541.466, "end": 1544.066, "text": "语音只会让错误操作更难阻止", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 655, "start": 1544.266, "end": 1545.333, "text": "长期来看", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 656, "start": 1545.5, "end": 1547.366, "text": "手机代理可能从测试", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 657, "start": 1547.366, "end": 1549.333, "text": "工具进入运维和客服", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 658, "start": 1549.766, "end": 1551.8, "text": "客服人员接到用户问题后", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 659, "start": 1551.8, "end": 1554.5, "text": "可以在隔离设备上复现相同路径", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 660, "start": 1554.8, "end": 1557.1, "text": "应用商店审核可以自动", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 661, "start": 1557.1, "end": 1559.0, "text": "探索权限和订阅流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 662, "start": 1559.366, "end": 1561.166, "text": "企业内部还能检查不同", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 663, "start": 1561.166, "end": 1563.3, "text": "版本是否出现行为变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 664, "start": 1563.6, "end": 1564.7, "text": "这些场景都需要", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 665, "start": 1564.7, "end": 1566.666, "text": "明确授权和受控环境", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 666, "start": 1566.9, "end": 1569.366, "text": "不能直接拿真实用户账户试验", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 667, "start": 1569.533, "end": 1571.2, "text": "最有争议的想象是让", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 668, "start": 1571.2, "end": 1573.066, "text": "代理自行发现漏洞", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 669, "start": 1573.266, "end": 1574.666, "text": "它能够遍历界面", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 670, "start": 1574.666, "end": 1575.933, "text": "输入异常内容", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 671, "start": 1575.933, "end": 1577.966, "text": "观察崩溃并读取日志", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 672, "start": 1578.166, "end": 1580.166, "text": "确实具备一定探索能力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 673, "start": 1580.366, "end": 1583.266, "text": "公开项目的目标是测试和诊断", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 674, "start": 1583.866, "end": 1586.6, "text": "不等于授权代理攻击第三方应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 675, "start": 1586.766, "end": 1588.666, "text": "使用边界仍由操作者和", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 676, "start": 1588.666, "end": 1590.333, "text": "目标系统的许可决定", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 677, "start": 1590.5, "end": 1592.2, "text": "它对普通用户也有潜力", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 678, "start": 1592.566, "end": 1594.066, "text": "视障人士可以用自然", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 679, "start": 1594.066, "end": 1596.133, "text": "语言完成复杂应用操作", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 680, "start": 1596.4, "end": 1597.666, "text": "老人可以减少在", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 681, "start": 1597.666, "end": 1599.533, "text": "多层菜单中寻找按钮", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 682, "start": 1599.933, "end": 1602.966, "text": "重复性手机任务可以交给代理不过", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 683, "start": 1603.2, "end": 1606.066, "text": "项目当前的产品形态面向开发者", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 684, "start": 1606.166, "end": 1608.3, "text": "需要配置Python模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 685, "start": 1608.333, "end": 1611.1, "text": "密钥ADB和无障碍权限", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 686, "start": 1611.266, "end": 1613.533, "text": "距离大众助手还有明显距离", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 687, "start": 1613.733, "end": 1616.733, "text": "Artemis真正释放出的信号是AI", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 688, "start": 1616.733, "end": 1619.1, "text": "代理开始从浏览器走进设备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 689, "start": 1619.566, "end": 1622.733, "text": "网页有DOM和相对统一的交互结构", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 690, "start": 1622.866, "end": 1624.566, "text": "手机应用更加封闭", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 691, "start": 1624.7, "end": 1626.0, "text": "一个开源框架能够", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 692, "start": 1626.0, "end": 1627.733, "text": "跨应用完成长流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 693, "start": 1627.9, "end": 1630.5, "text": "说明多模态模型工具调用和", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 694, "start": 1630.5, "end": 1632.933, "text": "状态管理已经组合到新的阶段", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 695, "start": 1633.4, "end": 1634.966, "text": "这个阶段仍然粗糙", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 696, "start": 1635.333, "end": 1636.766, "text": "速度按秒计算", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 697, "start": 1636.866, "end": 1639.3, "text": "复杂模式一步可能接近1分钟", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 698, "start": 1639.533, "end": 1642.4, "text": "权限配置不当会带来安全风险", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 699, "start": 1642.7, "end": 1645.066, "text": "基准成绩缺少完整独立验证", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 700, "start": 1645.2, "end": 1647.1, "text": "开源署名还发生过争议", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 701, "start": 1647.6, "end": 1649.3, "text": "它没有资格被包装成", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 702, "start": 1649.3, "end": 1650.966, "text": "成熟的手机操作系统", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 703, "start": 1650.966, "end": 1652.733, "text": "代理可它也不是一次", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 704, "start": 1652.733, "end": 1654.566, "text": "无关紧要的代码发布", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 705, "start": 1655.0, "end": 1657.066, "text": "Google把项目放在公开仓库", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 706, "start": 1657.266, "end": 1659.4, "text": "使用Apache 2.0许可证", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 707, "start": 1659.6, "end": 1661.366, "text": "开发者可以检查实现", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 708, "start": 1661.5, "end": 1662.3, "text": "替换模型", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 709, "start": 1662.3, "end": 1663.866, "text": "并接入自己的设备", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 710, "start": 1664.166, "end": 1667.3, "text": "开源让外界有机会复测99.1%", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 711, "start": 1667.333, "end": 1670.2, "text": "也让代码来源问题无法被藏起来", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 712, "start": 1670.4, "end": 1673.366, "text": "它同时接受技术检验和社区审查", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 713, "start": 1673.566, "end": 1675.866, "text": "接下来最值得观察的是三件事", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 714, "start": 1676.066, "end": 1678.766, "text": "GitHub星标能涨到多少并不关键", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 715, "start": 1679.2, "end": 1683.3, "text": "独立团队能否复现接近99.1%的结果", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 716, "start": 1683.666, "end": 1685.4, "text": "企业能否在真实应用", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 717, "start": 1685.4, "end": 1687.3, "text": "中稳定运行长流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 718, "start": 1687.6, "end": 1689.5, "text": "以及维护者能否持续", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 719, "start": 1689.5, "end": 1691.666, "text": "处理安全与署名问题", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 720, "start": 1691.866, "end": 1693.0, "text": "任何一项失败", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 721, "start": 1693.166, "end": 1695.466, "text": "Optimus都可能停留在演示阶段", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 722, "start": 1695.7, "end": 1697.333, "text": "如果三项都能过关", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 723, "start": 1697.366, "end": 1699.9, "text": "移动开发会发生一次实质变化", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 724, "start": 1700.2, "end": 1703.1, "text": "编程代理将不再只盯着代码和终端", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 725, "start": 1703.533, "end": 1704.966, "text": "他能拿起一台手机", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 726, "start": 1704.966, "end": 1707.566, "text": "用用户的方式检查自己的作品", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 727, "start": 1707.7, "end": 1710.666, "text": "测试脚本也不必把每个坐标写死", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 728, "start": 1710.866, "end": 1712.533, "text": "目标和验证条件会", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 729, "start": 1712.533, "end": 1714.133, "text": "成为新的核心资产", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 730, "start": 1714.133, "end": 1715.933, "text": "所以Artemis最重要的", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 731, "start": 1715.933, "end": 1717.766, "text": "地方不是AI会点手机了", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 732, "start": 1718.1, "end": 1719.366, "text": "真正的变化是", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 733, "start": 1719.4, "end": 1722.366, "text": "AI第一次有机会把理解需求修改", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 734, "start": 1722.366, "end": 1725.5, "text": "代码操作设备和验证结果连成闭环", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 735, "start": 1725.7, "end": 1729.0, "text": "99.1%只是吸引注意力的数字", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 736, "start": 1729.533, "end": 1730.8, "text": "能不能把这个闭环", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 737, "start": 1730.866, "end": 1733.133, "text": "稳定放进真实开发流程", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 738, "start": 1733.266, "end": 1735.5, "text": "才决定它会成为基础工具", "source": "video_source_subtitle", "authority": "video_source_subtitle" }, { "id": 739, "start": 1735.6, "end": 1738.0, "text": "还是另一个跑分漂亮的开源项目", "source": "video_source_subtitle", "authority": "video_source_subtitle" } ], "transcript_selection": { "schema": "videonote.transcript_selection.v1", "status": "SELECTED", "requested_language": "zh", "duration": 1738.0, "reference_path": "/Users/sagawa/AI_WORK/video_notes/url/20260924_113444__谷歌瘋了_一句話讓AI接管手機_Artemis_震撼開源_99.1_成功率嚇死人/reference_subtitle.vtt", "reference_kind": "manual", "reference_language": "zh", "selected_source": "video_source_subtitle", "source_caption_kind": "manual", "source_caption_processing": "verbatim_direct", "youtube_overlap_clamps": [], "general_high_confidence_repairs": [], "decision_reason": "complete_source_subtitle_fast_path_no_whisper", "whisper_rejected": false, "source_profile": { "segment_count": 740, "duration_coverage": 0.918237 } } }