[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-google-managed-agents-api-one-call":3,"topics-all":33,"news-related-a4f1f3a9-2900-4ed0-9dff-016535a6f707":52},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":20,"news_slug":26,"published_at":27,"created_at":28,"modified_at":29,"is_published":30,"publish_type":31,"image_url":13,"view_count":32},"a4f1f3a9-2900-4ed0-9dff-016535a6f707","Google I\u002FO 推出 Managed Agents API：一条调用完成Agent部署，代价是放弃执行层控制","Google在I\u002FO大会上发布了Managed Agents in Gemini API，这个服务的核心卖点很简单——把数周的Agent工程工作压缩成一次API调用。配合新推出的Antigravity CLI，Google显然希望在Agent执行层也实现端到端控制。\n\n这背后反映的是一场正在发生的架构分歧：Agent管理应该放在模型层（Anthropic的方式），还是基础设施层（Google的方式）？\n\nAnthropic的Managed Agents将编排能力嵌入模型层，优点是企业保有执行控制权，模型负责推理和规划。而Google则更进一步，把模型、harness、沙箱三层当作一个整体来优化，全部跑在Google托管的安全环境里。Ramp的René Sultan评价很直接：有了这套东西，开发者可以专注打磨Agent的领域行为，迭代速度完全不一样。\n\n从技术上看，这种做法确实能解决部署Agent的最大痛点——前期那些\"无聊的工作\"：搭执行环境、配沙箱、接工具调用基础设施。有客户说自己\"以前需要两周的工作，现在一次调用就搞定了\"。\n\n但把执行层交给平台意味着把控制权也交了出去。XYO创始人Arie Trouw提醒了一个风险：开发者会把确定性服务换成概率性服务，一旦出问题，数据损坏或服务降级会比传统方案更难追踪和修复。\n\n我的观点：Google这套方案对于需要快速原型验证的企业很有吸引力，但生产级使用时，平台锁定和可观测性不足的风险不可忽视。选这条路的企业，最好提前想清楚自己在监控和故障恢复上能接受多大程度的\"黑箱化\"。","https:\u002F\u002Fventurebeat.com\u002Forchestration\u002Fgoogles-managed-agents-api-promises-one-call-deployment-at-the-cost-of-execution-layer-control","17ff6400-4413-4b16-86fb-99951dbbd08d",[10,14,17],{"id":11,"name":12,"slug":12,"description":13,"color":13},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"8cf7490f-2449-4ba7-be19-61befa0d92b4","google",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0a93ec8e-ea39-4693-81de-563ca8c173f7","inference",[21],{"id":22,"lang":23,"title":24,"summary":25,"content":13},"4e1572be-cb30-4d05-acb2-f736406d30b9","en","Google's Managed Agents API: one call to deploy, less control","At I\u002FO, Google launched Managed Agents in Gemini API. The core pitch is brutally simple: collapse weeks of agent engineering into a single API call. Paired with the new Antigravity CLI, Google is clearly gunning for end-to-end control at the agent execution layer.\n\nThis reflects an emerging architectural split: should agent management live in the model layer (Anthropic's way) or in the infrastructure layer (Google's way)?\n\nAnthropic's Managed Agents embeds orchestration capability in the model layer — enterprises retain execution control while the model handles reasoning and planning. Google goes further, treating the model, harness, and sandbox as one integrated whole, all running inside Google's managed secure environment. Ramp's René Sultan put it directly: with this stack, developers can focus on polishing the domain behavior of their agent, and the iteration speed is in a different league.\n\nTechnically, this approach does address the biggest deployment pain point — the upfront \"boring work\" of standing up execution environments, configuring sandboxes, and wiring tool-call infrastructure. One customer said they \"used to need two weeks of work, now it's done in a single call.\"\n\nBut handing the execution layer to the platform means handing over control as well. XYO founder Arie Trouw flags a real risk: developers end up swapping deterministic services for probabilistic ones, and when things break, data corruption or service degradation becomes much harder to trace and fix than with traditional stacks.\n\nMy take: Google's approach is highly attractive for enterprises that need to prototype fast, but for production use, the platform-lock-in and observability gaps can't be ignored. Teams choosing this path should think ahead about how much \"black-boxing\" they can tolerate in monitoring and incident response.","google-managed-agents-api-one-call","2026-05-25T04:10:00Z","2026-05-25T04:10:55.644805Z","2026-08-19T02:08:40.142862Z",true,"agent",156,[34,43],{"slug":35,"tag_slug":35,"title_zh":36,"title_en":37,"intro_zh":38,"intro_en":39,"id":40,"is_active":30,"created_at":41,"modified_at":42},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":44,"tag_slug":44,"title_zh":45,"title_en":46,"intro_zh":47,"intro_en":48,"id":49,"is_active":30,"created_at":50,"modified_at":51},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":53},[54,59,64,69,74,79],{"id":55,"title":56,"news_slug":57,"published_at":58},"a3e9cf60-e2dc-4ac6-a1a6-1089ee721bc9","Google Antigravity CLI 全面开放：子Agent并行编排进入终端开发时代","google-antigravity-cli-sub-agent-parallel","2026-05-31T10:05:00+00:00",{"id":60,"title":61,"news_slug":62,"published_at":63},"fcd676ca-d64a-4321-b453-45763c803a67","Google I\u002FO 2026 重新定义AI开发：Antigravity 2.0 将多智能体编排带入产品级工程","google-io-2026-antigravity-2-0-orchestration","2026-05-28T05:00:00+00:00",{"id":65,"title":66,"news_slug":67,"published_at":68},"af34b075-9fd7-4f0f-8f87-eb947fd81f4e","Dream-RSI:Google 让智能体在历史里做梦,发现调用省 162 倍","dream-rsi-replay-simulator","2026-09-15T23:20:00+00:00",{"id":70,"title":71,"news_slug":72,"published_at":73},"20568e5d-3b66-495d-8b7b-0a702f3c7877","模型在进化,训练环境却是死的:Google 开源 EnvHarness,给环境也套一层 harness","google-envharness-agent-environments","2026-08-20T10:42:06+00:00",{"id":75,"title":76,"news_slug":77,"published_at":78},"36055e5f-136f-497d-8763-3ed6609f59ff","Meta Muse Glimmer 30B 本地落地:Apache 2.0 的开源智能体,把 Agent 装进 24GB 显存","meta-muse-glimmer-30b-local-agent-apache2-r2","2026-08-19T03:00:00+00:00",{"id":80,"title":81,"news_slug":82,"published_at":83},"7d371b09-9792-465d-b73a-3d0af4735129","InferenceBench：15 个前沿 Agent 自主做 LLM 推理优化","inferencebench-open-ended-llm-optimization","2026-08-16T12:00:00+00:00"]