2026-07-17 · AI 资讯日报
今日新收录 210 条公开资讯，按模型 / 产品 / 行业 / 论文 / 观点 自动归类汇编（非 AI 生成，点击可溯源原文）。
本日报由系统根据当天新收录的公开资讯自动汇编、按方向归类，非 AI 生成；点击任意条目可溯源原文。
今日精选6 条
- 1.
autonomous red teaming platform; multi-agent offensive-security meta-harness
- 2.
The open-source AI workbench for scientific research
- 3.
把 Markdown 一键排成可直接粘进公众号编辑器的精致 HTML —— 6 套精选主题 + 主题生成器 + 双关卡校验。An AI-agent skill that turns Markdown into paste-ready WeChat article HTML.
- 4.
Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT.
- 5.
The Fable Workflow: how Claude Fable 5 worked, distilled into skills any model can run, with the eval that keeps it honest. Think / act / prove.
- 6.
A practical, open-source guide to mastering WorkBuddy through real-world workflows.开源的 WorkBuddy 实战蓝皮书：教程、真实工作流、Skills、MCP、自动化与多智能体实践。
模型发布/更新6 条
- 1.
AI 导读 · 自建网关掌控Claude，企业级AI部署更灵活，降低依赖风险。
- 2.
AI 导读 · 极致软硬协同设计降低每token成本，让后训练阶段的智能性价比达到新高度，是智能体AI落地的关键指标
- 3.A scorecard for the AI age— OpenAI
Sarah Friar, CFO of OpenAI, introduces a practical AI scorecard to measure ROI through useful work, cost per successful task, dependability, and return on compute.
- 4.
AI 导读 · 揭示AI记忆机制的安全隐患，提醒用户重视隐私保护。
- 5.
AI 导读 · 智谱ARR半年飙升15倍达10亿美元，印证AI商业化进入爆发期，行业格局加速重塑。
- 6.
AI 导读 · Kimi K3模型与支付宝教育优惠升级，显示AI与支付场景的深度融合趋势。
产品发布/更新6 条
- 7.elder-plinius/T3MP3ST— GitHub
autonomous red teaming platform; multi-agent offensive-security meta-harness
- 8.synthetic-sciences/openscience— GitHub
The open-source AI workbench for scientific research
- 9.isjiamu/gzh-design-skill— GitHub
把 Markdown 一键排成可直接粘进公众号编辑器的精致 HTML —— 6 套精选主题 + 主题生成器 + 双关卡校验。An AI-agent skill that turns Markdown into paste-ready WeChat article HTML.
- 10.
Let Claude (or any LLM) actually watch a video — scene-aware, deduplicated frames + transcript, from a URL or local file. Runs locally, MIT.
- 11.Sahir619/fable-method— GitHub
The Fable Workflow: how Claude Fable 5 worked, distilled into skills any model can run, with the eval that keeps it honest. Think / act / prove.
- 12.AlephAITech/WorkBuddyGuide— GitHub
A practical, open-source guide to mastering WorkBuddy through real-world workflows.开源的 WorkBuddy 实战蓝皮书：教程、真实工作流、Skills、MCP、自动化与多智能体实践。
行业动态6 条
- 13.
AI 导读 · AI监控加剧护士工作压力，暴露科技应用与人文关怀的冲突。
- 14.The state of open source AI— Hacker News
- 15.
- 16.
I receive an AWS Budgets alert that my budget is exceeding the alert threshold. Threshold is 5$. Forecasted amount is listed as $3,005,575,870.47. (Yepp, right, that’s 3 billion do…
- 17.
- 18.
论文研究6 条
- 19.LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget— HuggingFace Papers
AI 导读 · 突破固定GPU预算下长上下文强化学习训练瓶颈，实现200万token以上规模，显著提升模型处理超长文
- 20.
AI 导读 · 开源视频理解模型新突破，效率与通用性兼得，为视频AI应用铺平道路。
- 21.SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning— HuggingFace Papers
Large language models are increasingly trained as interactive agents for long-horizon tasks involving multi-turn interaction, tool use, and environment feedback. Outcome-based reinforcement learning (…
- 22.SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration— HuggingFace Papers
Recent advances in Tool-Integrated Large Language Models have made web search a core capability of information-seeking agents. However, as interaction histories grow, agents increasingly struggle to t…
- 23.BadWAM: When World-Action Models Dream Right but Act Wrong— HuggingFace Papers
World-action models (WAMs) are emerging as a promising foundation for embodied control: rather than predicting actions alone, they learn representations that couple action generation with future world…
- 24.
Video generation increasingly relies on keyframe-based workflows, where creators specify a sequence of reference images to guide generation. Although recent models support multi-keyframe conditioning,…
技巧与观点6 条
- 25.
AI 导读 · 亚马逊Quick将AI销售助手覆盖全流程，从筛选客户到沟通签约，真正实现销售智能化升级。
- 26.
AI 导读 · 移动端优化让商业决策者随时查看数据，消除桌面端局限，提升实时响应效率。
- 27.
AI 导读 · Smartsheet在AWS上搭建远程MCP服务器，展示了企业级AI基础设施的安全与扩展性实践。
- 28.
AI 导读 · NVIDIA与Hugging Face联手，实现视频图像模型大规模微调，降低AI应用门槛。
- 29.Quoting Kimi K3— Simon Willison
Is there something I can actually help you with today? — Kimi K3 , after refusing to leak its system prompt Tags: kimi , ai-personality , generative-ai , ai , llms
- 30.LLM cliché highlighter— Simon Willison
Tool: LLM cliché highlighter I got frustrated reading yet another article that was crammed with the clichés of LLM-generated writing - "no fluff, no filler, no jargon" type stuff -…
快讯
- ·消息称马斯克旗下 SpaceX 正与美国五角大楼洽谈 AI 算力供应，拟通过低价抢市场— IT之家 · 7/18 07:12
- ·谷歌 DeepMind CEO 哈萨比斯：AI 不会令科学、技术、工程、数学专业失去价值— IT之家 · 7/18 07:11
- ·Vertu wants executives to pay $6,880 for an AI agent — here’s how it actually performs— TechCrunch · 7/18 06:55
- ·Kaiser nurses say AI, surveillance are making their jobs and patient care worse— Hacker News · 7/18 06:26
- ·Databricks hits $188B valuation, extending its run as AI’s favorite second act— TechCrunch · 7/18 06:12
- ·AI从“能对话”走向“能办事”，蚂蚁集团WAIC展示三层布局— InfoQ · 7/18 06:08