https://jiangsee.comJiangSee 是一本关注 AI、基础设施与企业科技变化的中文独立科技刊物。zh-cnSat, 19 Sep 2026 18:16:10 GMThttps://jiangsee.com/article/robots-learn-from-failed-actionshttps://jiangsee.com/article/robots-learn-from-failed-actions失败轨迹不再只是被丢弃的训练垃圾，它正在成为机器人理解边界、恢复动作和现实摩擦的重要数据。Tue, 15 Sep 2026 00:00:00 GMT林雅Roboticshttps://jiangsee.com/article/cloud-platforms-build-decision-layerhttps://jiangsee.com/article/cloud-platforms-build-decision-layer资源越来越多之后，真正稀缺的不再是服务数量，而是把成本、性能和合规约束翻译成可执行选择的能力。Mon, 14 Sep 2026 00:00:00 GMT林雅Cloudhttps://jiangsee.com/article/admin-local-agent-testhttps://jiangsee.com/article/admin-local-agent-test这是一篇用于验证 JiangSee Admin 本地写作、预览、发布链路的测试文章。Sun, 13 Sep 2026 00:00:00 GMT林雅AI Agentshttps://jiangsee.com/article/advanced-packaging-test-bottleneckhttps://jiangsee.com/article/advanced-packaging-test-bottleneck模块数量增加后，如何在封装前识别坏芯片、在组合后定位问题，开始决定整套系统的交付效率。Sun, 13 Sep 2026 00:00:00 GMT林雅Chipshttps://jiangsee.com/article/ai-agent-from-chat-to-workflowhttps://jiangsee.com/article/ai-agent-from-chat-to-workflowAgent 的真正变化不在于它能说什么，而在于它能否稳定地完成一段有边界、有反馈、可追踪的工作。Sat, 12 Sep 2026 00:00:00 GMT林雅AI Agentshttps://jiangsee.com/article/inference-cost-and-product-complexityhttps://jiangsee.com/article/inference-cost-and-product-complexity当每一次调用不再昂贵，产品经理会把更多决策交给模型，系统也会因此需要更细的边界。Fri, 11 Sep 2026 00:00:00 GMT林雅AI Infrastructurehttps://jiangsee.com/article/cloud-platform-control-planehttps://jiangsee.com/article/cloud-platform-control-plane当基础设施越来越复杂，企业真正需要的是一个能解释变化、约束风险并保持秩序的控制面。Thu, 10 Sep 2026 00:00:00 GMT林雅Cloudhttps://jiangsee.com/article/robotics-changes-the-floorhttps://jiangsee.com/article/robotics-changes-the-floor机器人落地的难点不在于让单台设备完成动作，而在于让它成为一个可协作、可维护的生产节点。Wed, 09 Sep 2026 00:00:00 GMT林雅Roboticshttps://jiangsee.com/article/chiplet-and-the-new-chip-supply-chainhttps://jiangsee.com/article/chiplet-and-the-new-chip-supply-chain先进封装带来的不只是更高的集成度，也把芯片设计、验证和制造的协作方式重新拆开。Tue, 08 Sep 2026 00:00:00 GMT林雅Chipshttps://jiangsee.com/article/enterprise-agent-context-and-permissionshttps://jiangsee.com/article/enterprise-agent-context-and-permissions企业不会只购买一个更聪明的模型，它们需要的是一个知道什么能看、什么能做、什么必须留下记录的系统。Mon, 07 Sep 2026 00:00:00 GMT林雅AI Agentshttps://jiangsee.com/article/ai-data-center-energy-and-schedulinghttps://jiangsee.com/article/ai-data-center-energy-and-scheduling算力需求的增长，把电力、散热和任务调度从运营问题推到了 AI 基础设施的核心。Sun, 06 Sep 2026 00:00:00 GMT林雅AI Infrastructurehttps://jiangsee.com/article/robotics-vision-and-uncertaintyhttps://jiangsee.com/article/robotics-vision-and-uncertainty真实环境里最有价值的视觉系统，不是永远给出答案，而是知道什么时候应该停下来问人。Sat, 05 Sep 2026 00:00:00 GMT林雅Robotics