01政策相关
102行业前沿动态
4Apple 发布新一代 Apple Intelligence，Siri AI 正式以测试版上线
2026-09-14Apple 发布新一代 Apple Intelligence，推出全面重构的 Siri AI，支持个人语境理解、屏幕感知、系统级应用操作和跨设备对话，今日起以英文测试版随 2027 系统更新推出，下月扩展至法语、日语、韩语、葡萄牙语和西班牙语。
硅基流动上线开源模型 Hy4 preview，770B 总参数、1M 上下文
2026-09-14硅基流动（SiliconFlow）宣布开源模型 Hy4 preview 上线其平台。该模型总参数 770B、每 token 激活 49B、支持 1M 上下文，采用 Apache 2.0 协议，面向编码、分析、研究和复杂实际工作。用户可将其接入 Claude Code、Codex、Cursor 等已有工具；图片显示定价为每 1M tokens 输入 $0.834、输出 $2.501、缓存 $0.042。
Claude for Small Business 新增 43 个 workflow 与 27 个集成
2026-09-14Anthropic 为 Claude for Small Business 新增 43 个 workflow 和 27 个集成，覆盖 Shopify、Salesforce、TikTok、Zoom、Stripe、Zapier 等工具，该产品自 5 月上线以来安装量超过 90 万次。
小红书 AllSpark 开源 Search Agent 模型 Iris，35B 与 397B 版本同量级成绩领先
2026-09-14小红书 AllSpark 团队发布并开源 Search Agent 模型 Iris，权重和评测代码已公开，数据与训练配方将陆续公布。
04论文研究
9DeepSeek-V4.1-Flash （Max） 进入 Agent Arena 开源模型第 3 名
2026-09-14DeepSeek-V4.1-Flash （Max） 进入 Agent Arena 开源模型第 3 名，净提升 +4.87%，每任务中位成本 $0.07，重塑 Pareto 前沿。其成本比第 2 名 Hy4 preview 低 68%、成绩仅差 0.09 个百分点；总榜排名第 12，Confirmed Success 信号排名 第 4（+13.75%）。
"The Only Thing Certain About This is Uncertainty": Exploring Informal Care Coordination Practices Among Older Adults wi…
2026-09-14arXiv:2609.12070v1 Announce Type: new Abstract: Older adults aging in place often have informal support systems to help them maintain independence and quality of life. As they age, many older adults deal with the onset o…
Creating an Atomic User Model for Personality-Aware Large Language Model Interaction
2026-09-14arXiv:2609.12086v1 Announce Type: new Abstract: Assistants built on large language models are expected to write as their user would, and the dominant approach is single-channel: preferences summarised from conversation h…
MAIA: Multi-Agent Intent Articulation for Requirement Discovery in Art Commissions
2026-09-14arXiv:2609.12097v1 Announce Type: new Abstract: In bespoke art commissions, laypeople know what they feel but lack the words to specify it: one participant wanted a laid-off truck driver depicted as "a ghost in his own m…
Plans They Abandon, Reports They Author: The Narrative Layer of Autonomous Agents
2026-09-14arXiv:2609.12205v1 Announce Type: new Abstract: When a coding agent finishes a task, the developer reviews a summary the agent wrote about itself, not a display someone designed. We ask how much of the agent's work that …
BRIDGE-EEG: Bridging Self-Supervised Pretraining and Efficient Deployment for Cross-Dataset EEG Classification
2026-09-14arXiv:2609.12218v1 Announce Type: new Abstract: The growing use of electroencephalography (EEG) motivates automated analysis that is accurate, transferable, and deployable on constrained hardware. Recent EEG foundation m…
Tact: A Zero-Cost, Browser-Based Pipeline for On-Demand Tactile Braille Storybooks
2026-09-14arXiv:2609.12272v1 Announce Type: new Abstract: Braille literacy among blind school-age children has fallen sharply, in part because producing illustrated braille pages still requires specialized software and trained lab…
Synthetic TLX: Forecasting Human Workload Using Agent Simulation
2026-09-14arXiv:2609.12273v1 Announce Type: new Abstract: Assessing human workload for technology-mediated tasks helps prevent task failure caused by poor technology design. Traditionally, workload is assessed retrospectively usin…
"People can change, and patterns can be broken": Contextualizing Tradeoffs in Automated Decision-Making Systems
2026-09-14arXiv:2609.12288v1 Announce Type: new Abstract: Automated decision-making (ADM) systems are increasingly deployed in domains such as mortgage lending, prison sentencing, health insurance coverage, and hiring. Designing a…
05技巧与观点
5科技巨头放缓 AI 开发的口头协议是安全共识还是卡特尔
2026-09-14Sam Altman、Dario Amodei、Demis Hassabis 和 Elon Musk 周末粗略同意放慢 AI 开发，提出引入第三方审计、监管国内实验室并达成全球放缓协议，批评者则称其为压制竞争者和开源运动的"卡特尔"。
GPT-5.6 Luna 对比 GPT-6 Astra：$1.20 的模型做代码评审够用吗
2026-09-14Entelligence 在 50 个公开基准 PR 上用相同提示词对比 GPT-5.6 Luna 和 GPT-6 Astra 的代码评审能力，Luna 找到 69 个经验证 bug（Astra 92 个），总成本 $0.20 对 $5.66，精度 74% 对 96%。
Anthropic 如何重构测试影响分析服务以应对智能体编码带来的 CI 压力
2026-09-14Anthropic 工程师每季度交付的代码量是 2021-2025 年均值的 8 倍，其中 80% 由 Claude 编写，六个月内 CI 任务增长 25 倍。
恶意 AI 智能体攻击 RubyGems.org：YARD 执行任意代码与 Fastly 缓存密钥利用分析
2026-09-14作者分析被指与 OpenAI 机器人相关的 GemStuffer 恶意 gem 代码，发现其利用 .yardopts 的 --load 加载脚本，在安装或 RubyDoc.info 处理 YARD 文档时执行任意代码，且 Docker 容器有网络访问权限可执行爬取。
Tomer Tunguz 解析 Amodei 放缓前沿提议背后的五派立场与算力监管难题
2026-09-14Tomer Tunguz 分析 Dario Amodei 提出的放缓前沿 AI 发展号召，梳理出可解释性、劳工、经济、地缘政治和监管俘获五派立场，并指出没有任何一派给出具体速度。