Tag
DoorDash revealed in its Q3 2026 earnings call that it has only received 'several hundred' orders from third-party AI agents like Muse, Instinct, and Fo, dwarfed by its 3 million daily deliveries, as CEO Tony Xu urged caution about the hype around agentic commerce.
A breakdown of what an 'agent harness' is — the code layer around a model that turns model decisions into real work — detailing 8 common components: the loop, tools/sandbox, subagents, context management, memory, guardrails, verification, and observability.
A report examining the intersection of video generation and agent harnesses, covering agentic video generation, agentic world modeling, and recursive self-improvement for autonomous multi-shot video production.
Addy Osmani, a core member of Anthropic's Claude Code team, argues that developers are systematically overprompting: trimming Claude Code's system prompt by 80% caused no drop in benchmark scores. With each new model generation, the winning move is subtraction — give only the goal and a way to verify it, avoid blindly piling on multiple agents, let the model interview you in reverse to clarify requirements, and use a separate agent to review diffs.
EditAgent 提出一种新的智能体图像编辑框架，通过将编辑步骤路由到多样化的专用编辑器并利用验证记录进行列表式评判，同时扩展了智能体的动作空间与观察空间，显著提升了复杂图像编辑任务的可靠性与效率。
微软宣布将源自 GitHub 的 Copilot harness 通过 Microsoft Foundry 向开发者和企业客户开放，提供托管式的多模型智能体编排能力，支持 20+ 前沿模型、MCP 与 agent skills 等开放标准。
Kai-Fu Lee 转发其在《金融时报》的最新评论文章，并宣传新书《AI Native》，探讨AI加速转型如何重塑公司层级结构，中层管理者与白领若无法有效与AI及人协作将面临最大被取代风险。
展示了 GPT-6 Astra 在一次 Codex 对话中自主完成从设计、模拟到驱动真实机械臂绘画、录制、配乐与剪辑的端到端创意流程，突出其在物理世界中的自主规划与执行能力。
GitHub announced a major overhaul of Copilot at Universe, introducing Copilot Fusion, a unified Copilot harness in Foundry, and Copilot Projects to drive an agentic future for the SDLC.
GitHub 记录到 2026 年 9 月有 73.8 亿次提交（同比增长逾 5 倍），为吸收海量智能体生成代码大幅扩容基础设施，并在 Universe 大会上发布 Copilot Fusion 等多项更新。
Satya Nadella announces major GitHub updates at Universe, including GitHub Copilot Fusion for faster, cheaper frontier-level quality, a unified Copilot harness in Foundry for long-running agents, and Copilot Projects for multi-agent orchestration across repos.
LangChain explains how nearly all proactive, always-on agents are built on cron-style scheduled runs—covering memory "dreaming," proactivity, triage, follow-ups, and orchestration—and shows how to build them with Managed Deep Agents on LangSmith, where over 70% of runs are already scheduled rather than user-initiated.
A new arXiv study finds that single-author scientific papers rose sharply after late-2025 agentic AI releases—up 70% in mathematics, 71% in condensed matter physics, and 44% in quantitative biology in early 2026—suggesting agentic AI is enabling individual researchers to work without collaborators. The shift is notably absent in experimental particle physics, where large teams remain essential.
WeVisCut 是一个脚本到视频的智能体系统，通过训练一个 9B 紧凑策略来协调视频编辑工具、从故障中恢复，并在保持质量的同时大幅降低 token 消耗，在长视频生成上以约 33 倍更少的 token 达到与 Claude Opus 5 相当的性能。
作者分享了数周以来使用 AI agent 7x24 小时监控生产系统的经验，指出需要 agent 理解系统、看到实际事件、访问代码并可验证证据，才能实现可信的自动化监控。
This preprint book presents a unified governance framework for generative and agentic AI, integrating technology, law, economics, and security lenses across eleven chapters with a proposed governance control plane design for 2030, anchored in Indian law with comparative cross-jurisdictional perspectives.
TechCrunch 报道 OpenAI 在开发者日推出的常驻 AI 个人助理 dots，其企业产品负责人 Alexander Embiricos 在 TechCrunch Disrupt 上讲述了该产品如何从 Codex 等编码工具演变为面向大众的主动式助手。
GEPA-Curator is a framework that improves agentic prompt optimization by filtering noisy failure signals and retrieving prior successful trajectories, achieving gains of up to 20.7% over GEPA and 21.2% over ACE on benchmarks like GoBrowse, ALFWorld, AppWorld, and τ-bench.
TAO 提出一个工具感知的联合调度框架，用于基于 LLM 的智能体服务，通过动态规划在容量约束下选择和放置会话，相比最优基线最高降低 26.8% 延迟、36.3% TTFT 并提升 29.5% 解码吞吐。
GPT-6.1 Terra 发布，OpenAI 称其在快速、能干的计算机使用（computer use）方面实现新突破，能理解屏幕内容、跨应用执行操作，并以更低价格提供更高效的工作能力。