日期：2026-03-31
本期聚焦：重点关注AI coding、AI SRE、AI辅助生活产品与工作流。
-
Popular AI gateway startup LiteLLM ditches controversial startup Delve（TechCrunch AI）
中文摘要：AI网关初创公司LiteLLM宣布终止与争议性安全合规服务商Delve的合作。此前，LiteLLM通过Delve获得了两项安全合规认证，但上周却遭遇了严重的凭证窃取恶意软件攻击。此次事件暴露了通过第三方服务商获取合规认证可能带来的安全隐患，也促使LiteLLM重新评估其安全供应链策略。对于依赖AI基础设施的企业而言，这一案例凸显了在追求合规认证的同时，必须对服务商的安全实践进行更严格的审查。
English Summary: AI gateway startup LiteLLM has cut ties with controversial security compliance provider Delve after suffering a severe credential-stealing malware attack last week. LiteLLM had obtained two security compliance certifications through Delve, highlighting the security risks associated with third-party compliance services and prompting a reassessment of supply chain security strategies.
-
15% of Americans say they’d be willing to work for an AI boss（TechCrunch AI）
中文摘要：根据Quinnipiac大学的一项最新民调，15%的美国受访者表示愿意接受AI作为自己的直属上司。这一现象反映了企业管理领域正在发生的深刻变革——越来越多的组织开始利用AI技术精简管理层级，业内有人将这一趋势称为"大扁平化"（The Great Flattening）。随着AI在任务分配、绩效评估和决策支持等方面的能力不断提升，传统的人机协作模式正在被重新定义，人机共管的新型组织架构可能成为未来职场的新常态。
English Summary: A Quinnipiac poll reveals that 15% of Americans would be willing to work for an AI boss. This reflects a profound shift in organizational management as companies increasingly use AI to eliminate management layers in what some call "The Great Flattening," redefining traditional human-AI collaboration models.
-
As more Americans adopt AI tools, fewer say they can trust the results（TechCrunch AI）
中文摘要：Quinnipiac大学的最新民调显示，尽管美国民众对AI工具的采用率持续上升，但对AI结果的信任度却在下降。大多数受访者对AI技术的透明度、监管框架以及其对社会产生的广泛影响表示担忧。这一矛盾现象表明，随着公众对AI技术的了解加深，人们对其潜在风险的认识也在提高。对于AI开发者和企业而言，建立可解释性机制、完善监管合规以及主动承担社会责任，将成为赢得用户信任的关键要素。
English Summary: A new Quinnipiac poll shows that while AI adoption is rising in the U.S., trust in AI results is declining. Most Americans express concerns about transparency, regulation, and AI's broader societal impact, indicating that as public understanding of AI deepens, awareness of its potential risks is also growing.
-
Mantis Biotech is making ‘digital twins’ of humans to help solve medicine’s data availability problem（TechCrunch AI）
中文摘要：Mantis Biotech正在开发人类"数字孪生"技术，旨在解决医学研究领域长期存在的数据可用性难题。该公司通过整合多源异构数据，构建合成数据集，从而创建能够模拟人体解剖结构、生理机能和行为模式的数字孪生模型。这一技术有望突破真实患者数据获取受限的瓶颈，为药物研发、临床试验设计和个性化医疗提供更安全、更高效的虚拟测试环境，加速医学创新进程。
English Summary: Mantis Biotech is creating 'digital twins' of humans to address medicine's data availability problem. The company synthesizes disparate data sources to build synthetic datasets representing human anatomy, physiology, and behavior, potentially overcoming limitations in accessing real patient data for drug development and clinical trials.
-
ScaleOps raises $130M to improve computing efficiency amid AI demand（TechCrunch AI）
中文摘要：云原生基础设施优化平台ScaleOps宣布完成1.3亿美元C轮融资，以应对当前AI算力需求激增带来的GPU短缺和云成本飙升问题。ScaleOps的核心能力在于实时自动化基础设施管理，通过智能调度算法动态优化Kubernetes集群中的GPU资源分配，帮助企业提升计算效率、降低运营成本。此轮融资反映了市场对AI基础设施优化工具的强劲需求，也预示着云资源智能化管理将成为AI产业的关键支撑环节。
English Summary: ScaleOps has raised $130M in Series C funding to tackle GPU shortages and soaring AI cloud costs through real-time infrastructure automation. The platform uses intelligent scheduling algorithms to dynamically optimize GPU resource allocation in Kubernetes clusters, improving computing efficiency and reducing operational expenses.
-
AI chip startup Rebellions raises $400 million at $2.3B valuation in pre-IPO round（TechCrunch AI）
中文摘要：韩国AI芯片初创公司Rebellions在IPO前轮融资中筹集4亿美元，公司估值达到23亿美元。Rebellions专注于AI推理芯片的设计与研发，是挑战英伟达市场主导地位的新兴力量之一。该公司计划于今年晚些时候正式上市，此轮融资将为其扩大生产规模、加速技术研发提供充足资金支持。随着AI推理需求持续增长，专用推理芯片市场的竞争格局正在快速演变。
English Summary: AI chip startup Rebellions has raised $400 million at a $2.3 billion valuation in a pre-IPO round. The company designs chips specifically for AI inference and plans to go public later this year, positioning itself as another challenger to Nvidia's dominance in the growing AI inference market.
-
Article: Optimization in Automated Driving: from Complexity to Real-Time Engineering（InfoQ AI/ML）
中文摘要：本文深入探讨了自动驾驶系统的技术架构与实时工程优化方法。作者Avraam Tolmidis重点介绍了上下文感知传感器融合技术和模型预测控制（MPC）求解器等关键优化手段，这些技术能够将原始传感器数据高效转化为安全可靠的车辆控制指令。文章从系统复杂性的角度分析了自动驾驶面临的工程挑战，为从事AI系统可靠性设计和实时控制优化的工程师提供了有价值的参考框架。
English Summary: This article explores the technical architecture of autonomous vehicles and real-time engineering optimization methods. Author Avraam Tolmidis focuses on context-aware sensor fusion and Model Predictive Control (MPC) solvers that efficiently transform raw sensor data into safe vehicle control commands, offering valuable insights for engineers working on AI system reliability and real-time control optimization.
-
Google Unveils AppFunctions to Connect AI Agents and Android Apps（InfoQ AI/ML）
中文摘要：谷歌发布AppFunctions新功能，旨在将Android系统转型为"智能体优先"的操作系统。这一早期测试版功能支持以任务为中心的应用交互模式，允许AI智能体或助手调用应用提供的功能模块来完成用户目标。这标志着移动操作系统架构的重大演进——从传统的应用启动模式转向意图驱动的智能体协作模式，为AI辅助生活产品和工作流的深度集成奠定了基础，有望重塑用户与移动设备的交互方式。
English Summary: Google has unveiled AppFunctions to transform Android into an 'agent-first' OS. The new early beta features support a task-centric model where AI agents or assistants leverage functional building blocks provided by apps to fulfill user goals, marking a significant evolution from traditional app-launching to intent-driven agent collaboration.
-
Microsoft Launches Azure Copilot Migration Agent to Accelerate Cloud Migration Planning（InfoQ AI/ML）
中文摘要：微软正式推出Azure Copilot迁移智能体，这是一款集成在Azure门户中的AI助手，旨在加速企业云迁移规划流程。该工具支持无代理VMware环境发现、自动化迁移规划和着陆区创建等功能。尽管被标记为正式发布，但目前该智能体仍处于公开预览阶段，尚无法直接执行实际迁移操作，数据复制和切换割接等关键步骤仍需通过Azure Migrate手动完成。这一产品体现了AI在IT运维和云迁移场景中的应用潜力。
English Summary: Microsoft has launched the Azure Copilot Migration Agent, an AI assistant built into the Azure portal that automates migration planning, agentless VMware discovery, and landing zone creation. While billed as generally available, the agent is in public preview and cannot execute migrations, with replication and cutover remaining manual tasks in Azure Migrate.
-
ProxySQL Introduces Multi-Tier Release Strategy With Stable, Innovative, and AI Tracks（InfoQ AI/ML）
中文摘要：开源数据库中间件ProxySQL发布3.0.6版本，并推出全新的多层级发布策略。该策略包含三个轨道：稳定版轨道专注于生产环境的可靠性保障；创新版轨道提前引入新功能；AI/MCP轨道则探索包括AI集成在内的未来能力。这一发布模式反映了数据库基础设施软件在AI时代的演进方向——在保持核心稳定性的同时，为AI驱动的智能化数据库运维（AI SRE）预留创新空间，为企业的不同场景需求提供灵活选择。
English Summary: ProxySQL 3.0.6 introduces a multi-tier release strategy with three tracks: Stable for production reliability, Innovative for early feature access, and AI/MCP for exploring future capabilities including AI integrations. This approach reflects the evolution of database infrastructure software in the AI era, balancing core stability with innovation space for AI-driven intelligent database operations.