{"status_id":"2092727360584135026","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月27日\n翻译自 英语\n显示原文\n我们使用 Inkling-Small 将论文摘要转化为快速、实用的总结。\n\nopen weights × open science \n35\n24\n236\n7万","url":"https://x.com/huggingface/status/2092727360584135026","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2089673018737869242","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月18日\n翻译自 英语\n显示原文\n我们刚刚在 Hub 上超过了 300 万个模型 \n\n社区正在加速迈向一个开放、分布式、开源 AI 无处不在、惠及每个人的未来 \n116\n172\n1,064\n9.7万","url":"https://x.com/huggingface/status/2089673018737869242","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2088301795890044975","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月15日\n翻译自 英语\n显示原文\n开放模型现状，2026年夏季  前沿模型规模越来越大，但小型模型仍主导实际使用。Qwen 引领本地推理，其次是 Gemma。AI 代理正在 Hub 上成为一股主要力量\n\n博客上有完整图景 \nState of Open Models: Summer 2026 Observations\n来自 huggingface.co\n52\n80\n277\n9.4万","url":"https://x.com/huggingface/status/2088301795890044975","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2087521619459850348","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月12日\n翻译自 英语\n显示原文\n会比 \n@UseCorgi\n 咖啡馆 还要受欢迎\n引用\nJess Fields\n@jessalanfields\n·\n8月12日\n翻译自 英语\nQwen 3.8 27b 将于周三东部标准时间上午 11:00 发布\n\n人们排起了长龙，迫不及待想看看这个炫酷的新模型到底能做什么\n28\n24\n136\n5万","url":"https://x.com/huggingface/status/2087521619459850348","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2085758063471120848","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月8日\n翻译自 英语\n显示原文\nAI 代理如何重现 ICML 2026 论文\nHugging Face\n@huggingface\nHow AI agents reproduced ICML 2026 papers\n21\n25\n82\n4.1万","url":"https://x.com/huggingface/status/2085758063471120848","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2084759576692363460","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n8月5日\n翻译自 英语\n显示原文\n我们正在与 Open Secure Alliance 合作制定事件学习指南，以共同开发更好的审查、披露以及针对安全事件和其控制措施。\n引用\nNVIDIA\n@nvidia\n·\n8月4日\n翻译自 英语\n当组织公开协作并快速行动时，安全性最强。\n\n如今拥有超过 120 名成员的开放安全 AI 联盟正在分享新的开源安全贡献，包括新的拟议 SAFE 指南，用于将机密事件发现转化为整个生态系统更强的保护。\n36\n21\n82\n4.3万","url":"https://x.com/huggingface/status/2084759576692363460","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2082502249784451163","author":"huggingface","utc_time":null,"text":"Hugging Face\n@huggingface\n·\n7月30日\n翻译自 英语\n显示原文\n专业提示：在您的网站/应用中添加一个“使用 Hugging Face 登录”OAuth 按钮，这样社区成员就可以轻松分享他们的电子邮件、创建模型/数据集仓库、在 Buckets 中存储数据，甚至启动支持 GPU 的 Jobs！\n\n\nhttps://\nhuggingface.co/docs/hub/en/oa\nuth\n…\n25\n32\n118\n3.9万","url":"https://x.com/huggingface/status/2082502249784451163","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@huggingface"],"conversation":{"parent_status_id":null}}
{"status_id":"2093036251603734919","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n13小时\n翻译自 英语\n显示原文\n祝贺 NVIDIA 可持续发展负责人 Josh Parker 入选 \n@TIME\n 的 2026 年 #TIME100AI 榜单。\n\n了解更多：\nhttps://\nnvda.ws/45RnvCW\n36\n34\n267\n6.2万","url":"https://x.com/nvidia/status/2093036251603734919","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2092960847408583120","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n18小时\n翻译自 英语\n显示原文\nNVIDIA Vera 即将登陆 \n@awscloud\n。\n\nNVIDIA 的 Ian Buck 亲自将 AWS 的首台 Vera CPU 服务器和 Vera Rubin GPU 交付给西雅图 AWS 总部的 Willem Visser 和 Supreeth Sheshadri。\n\nVera 专为代理式 AI 而设计：每美元处理更多令牌，为用户带来更快的结果，并为 AI 工厂提供扩展所需的计算基础。\n0:37\n75\n103\n918\n7.4万","url":"https://x.com/nvidia/status/2092960847408583120","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2092960850197741946","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n18小时\n翻译自 英语\n显示原文\n阅读更多：\nDelivering Vera: NVIDIA’s First CPU Built for Agents Is Shipping Now\n来自 blogs.nvidia.com\n13\n5\n42\n1.9万","url":"https://x.com/nvidia/status/2092960850197741946","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2092390101959479713","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n8月26日\n翻译自 英语\n显示原文\n从展厅到 AI 工厂：Hot Chips 2026 完全聚焦于极端协同设计，以加速代理式工作负载——历史上最复杂的工作负载。\n\nVera CPU、Vera Rubin、Groq 3 LPX、Spectrum-X Multiplane、BlueField-4 Scale-In 网络。\n\n一个专为代理打造的完整 AI 技术栈平台。\n81\n75\n501\n9.5万","url":"https://x.com/nvidia/status/2092390101959479713","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2092337107108770116","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n8月26日\n翻译自 英语\n显示原文\nNVIDIA Vera Rubin NVL72 生产机架已到货。\n\n计算托盘专为快速计算、组装和服务性而设计。制造过程 100% 自动化，每块托盘仅需一分钟即可组装完成。\n\n祝贺 \n@Microsoft\n 迎来首批投入运营的 Vera Rubin NVL72 机架，现正从 \n@HonHai_Foxconn\n Ingrasys 生产线下线。\n107\n346\n2,660\n26万","url":"https://x.com/nvidia/status/2092337107108770116","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2092269109086126575","author":"nvidia","utc_time":null,"text":"NVIDIA\n@nvidia\n·\n8月25日\n翻译自 英语\n显示原文\n认识便携式计算机，Perplexity 在 NVIDIA DGX Spark 上的全新本地优先代理堆栈。\n\n在本地运行时，便携式计算机提供一键本地推理设置，并为 DGX Spark 带来优化的代理式体验。\n\n了解更多并立即开始：\nhttps://\nblogs.nvidia.com/blog/local-ai-\nopen-source-models-agents-nemotron/#perplexity-spark\n…\n引用\nPerplexity\n@perplexity_ai\n·\n8月25日\n1:36\n翻译自 英语\n今天，我们在 @NVIDIA DGX Spark 上推出便携式计算机。\n\n便携式计算机是 Perplexity Computer 的完全本地版本，其中整个运行时：协调器 LLM、子代理 LLM、代理框架全部在您的本地硬件上运行。没有云依赖。\n64\n186\n1,212\n26万","url":"https://x.com/nvidia/status/2092269109086126575","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@nvidia"],"conversation":{"parent_status_id":null}}
{"status_id":"2082865587227590712","author":"github","utc_time":null,"text":"已置顶\nGitHub\n@github\n·\n7月31日\nYes\n引用\nIs Stacked PRs On GitHub Yet\n@StackedPRsOnGH\n·\n7月30日\nNo\n148\n376\n8,843\n102万","url":"https://x.com/github/status/2082865587227590712","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2093008357644779643","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n15小时\n翻译自 英语\n显示原文\n担任 \n@OpenClaw\n 的维护者是什么感觉？这是 GitHub 历史上增长最快的项目。\n\n你会失眠，但 \n@steipete\n 说你也能听到精彩的故事。\n0:50\n21\n40\n213\n5.9万","url":"https://x.com/github/status/2093008357644779643","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2093009115895243031","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n14小时\n翻译自 英语\n显示原文\n看看完整的 \n@OpenClaw\n 维护者圆桌讨论，与 \n@steipete\n \n@BradGroux\n \n@joshavant\n \n@jlehman_\n \n@somalley108\n \n@BunsDev\n \n@vincent_koc\n \nyoutube.com\nOpenClaw Went Viral. Meet the Maintainers Building and Securing It.\nSit down with OpenClaw creator Peter Steinberger and key maintainer...\n1\n18\n57\n7.4万","url":"https://x.com/github/status/2093009115895243031","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2092584970812367077","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月26日\n翻译自 英语\n显示原文\nGitHub 播客回来了！在这一集中，我们将认识我们的新主持人，并讨论最喜欢的剧集。\n\n 在您收听播客的任何地方找到它，或者点击此处 \nhttps://\nthe-github-podcast.simplecast.com/episodes/welco\nme-back-to-the-github-podcast\n…\n14\n12\n91\n5.2万","url":"https://x.com/github/status/2092584970812367077","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2092584972963983812","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月26日\n翻译自 英语\n显示原文\n然后回顾我们上季最喜欢的剧集。\n\n看看 Tiny Wins 团队如何通过修复小的瑕疵来节省开发者的时间。\nyoutube.com\nGitHub’s year in review: accessibility, MCP, and tiny wins | Episode...\nIn this year-in-review episode of The GitHub Podcast, Cassidy and A...\n3\n8\n3,775","url":"https://x.com/github/status/2092584972963983812","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2092386542304342311","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月26日\n翻译自 英语\n显示原文\n Linux 今天满 35 岁。\n\n这似乎是个不错的理由，汇总 35 款 Linux 上能玩的最棒的免费开源游戏。\n\n有些是经典之作。有些是隐藏的瑰宝。有些是令人惊讶的雄心勃勃项目。\n 文章\n35 free and open source games to celebrate 35 years of Linux\nBy @leereilly\nLinux turns 35 today, so we’re taking the opportunity to highlight 35 of our favorite free and open source (or “source available”) Linux games, their communities, and their stories!\n \nIf...\n33\n328\n2,381\n11万","url":"https://x.com/github/status/2092386542304342311","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2092283197660987573","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月26日\n翻译自 英语\n显示原文\n 通过实际使用它们来学习最新的 GitHub 工作流程。\n\n我们发布了 4 个新的 GitHub Skills 练习，旨在让开发者练习 AI 驱动的开发、代理工作流程和代码质量。\n26\n38\n383\n8.7万","url":"https://x.com/github/status/2092283197660987573","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2092283199053607414","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月26日\n翻译自 英语\n显示原文\n选择以下互动体验：\n 使用 GitHub Copilot CLI 编排专业代理\n 构建自动化仓库任务的代理工作流\n 使用 GitHub Copilot 应用从想法到合并 PR\n 自动化质量信号并更早发现问题\n2\n3\n25\n8,487","url":"https://x.com/github/status/2092283199053607414","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2091933100863537183","author":"github","utc_time":null,"text":"GitHub\n@github\n·\n8月25日\n翻译自 英语\n显示原文\n有了 GitHub Copilot 应用，您可以将所有的开发工作流程整合到一个地方。\n32\n16\n333\n8.9万","url":"https://x.com/github/status/2091933100863537183","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@GitHub"],"conversation":{"parent_status_id":null}}
{"status_id":"2093189665545507258","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n3小时\n翻译自 英语\n显示原文\n对于那些正在构建语音代理的人，这里有。\n引用\nkwindla\n@kwindla\n·\n14小时\n翻译自 英语\n介绍 PhoneLLM，一个用于语音代理的开源模型。\n\nGPT 5.6 Terra 在典型语音代理任务上的性能，延迟仅为其 1/3，成本仅为其 1/18。\n\n对于语音代理，我们需要既具有极低延迟又非常擅长工具调用和指令遵循的模型。\n\n这里存在权衡，在构建语音代理时，我们常常不得不在延迟或能力上做出妥协。借助\n4\n5\n4,883","url":"https://x.com/ProductHunt/status/2093189665545507258","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2093054760228434331","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n11小时\n翻译自 英语\n显示原文\nGitNexus 将您的整个代码库映射到一个单一的依赖图中。它追踪每个函数和调用，从而让 AI 代理获得确定性的答案，而不是“可能吧”。\n\n它是为代理设计的确定性代码智能，直接通过 MCP 接入 Claude Code。\n1:21\n5\n2\n22\n1.2万","url":"https://x.com/ProductHunt/status/2093054760228434331","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2093054772161253411","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n11小时\n翻译自 英语\n显示原文\n哦，而且它在 GitHub 上有超过 45K 个星标！\nGitHub - abhigyanpatwari/GitNexus: GitNexus: The Zero-Server Code Intelligence Engine - GitNexus is...\n来自 github.com\n3\n1,464","url":"https://x.com/ProductHunt/status/2093054772161253411","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2093009416643825786","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n14小时\n翻译自 英语\n显示原文\n认识 \n@ojin_ai\n，它可以将一张静态照片转化为一个互动的说话面孔，专注于延迟和轮流对话的极致体验。\n\n为了精准捕捉“说完话”和“正在思考”之间的细微差别，它不依赖单一的语音检测引擎。它同时运行四个引擎：设备端、云端、ai-coustics 和 Deepgram。\n\nOjin 能够实时处理打断、交叉对话和重口音。\n显示更多\n1\n3\n6\n8,950","url":"https://x.com/ProductHunt/status/2093009416643825786","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2092979206607270048","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n16小时\n翻译自 英语\n显示原文\nSavvy by \n@jamalavedra\n 是一个开源的 Mac 应用，它会索引你的客户文档，并在通话期间完全保持沉默，除非……\n\n· 有人与简报内容相矛盾\n· 跨越了预定义的界限\n· 你点击了提示按钮\n\n每张卡片都会在会议中引用确切的来源页面\n\n看看它 →\n1\n2\n7\n9,984","url":"https://x.com/ProductHunt/status/2092979206607270048","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2092717926914916678","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n8月27日\n翻译自 英语\n显示原文\n好奇今天有什么新品发布？以下是今天排行榜上吸引我们眼球的内容。\n2\n4\n9\n1.7万","url":"https://x.com/ProductHunt/status/2092717926914916678","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2092717929708294455","author":"ProductHunt","utc_time":null,"text":"Product Hunt \n@ProductHunt\n·\n8月27日\n翻译自 英语\n显示原文\n chocolifft by \n@SpatiallyMe\n \n\n它将 iPad 变成你 Mac 的触控栏：一条控制条和表情符号栏，位于你正在处理的任何内容下方。\n\n对于那些反对 AI 垃圾内容的人来说，你会很高兴地知道，chocolift 的每一块屏幕都是在 Figma 中手绘的。\nchoclift: Workflow Sweetener: Access apps, websites, Shortcuts & emojis on Mac using touch |...\n来自 producthunt.com\n1\n1\n1\n739","url":"https://x.com/ProductHunt/status/2092717929708294455","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@producthunt"],"conversation":{"parent_status_id":null}}
{"status_id":"2093157900218720407","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n5小时\n翻译自 英语\n显示原文\nStrictlyVC 精品晚宴将于9月10日重返纽约西村，嘉宾包括Keith Rabois、Craig Shapiro、Jason Levien、Tristan Walker、Brynn Putnam和Deven Parekh——涵盖人工智能、体育投资、社区建设、风险经济学，以及\nAI, athletes, and Keith Rabois: StrictlyVC is back in New York on September 10 | TechCrunch\n来自 techcrunch.com\n6\n2\n13\n2.3万","url":"https://x.com/TechCrunch/status/2093157900218720407","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093135742176223307","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n6小时\n翻译自 英语\n显示原文\n成为首席执行官并不意味着你突然就拥有了所有答案。\n\n当朱莉娅·哈茨接任 Eventbrite 的首席执行官角色时，她并不觉得自己准备好了。事实上，她用来描述这一转变的第一个词是“可怕”。\n\n在本期 Build Mode\n显示更多\n0:53\n6\n4\n23\n2.8万","url":"https://x.com/TechCrunch/status/2093135742176223307","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093116124602724654","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n7小时\n翻译自 英语\n显示原文\n在 TechCrunch Disrupt 2026 上，AI 舞台重磅回归，深入探讨过去几年社区中最热门的单一话题，由 Google for Startups 呈现。\nAnthropic and OpenAI are joining the AI stage at TechCrunch Disrupt 2026  | TechCrunch\n来自 techcrunch.com\n4\n14\n2.1万","url":"https://x.com/TechCrunch/status/2093116124602724654","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093072334122787252","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n10小时\n翻译自 英语\n显示原文\n克莱尔·麦克多诺（Claire McDonough）将于10月30日卸任，以追求新的机会，该公司周四在文件中表示。\nRivian's CFO is leaving the company | TechCrunch\n来自 techcrunch.com\n4\n2\n5\n2.6万","url":"https://x.com/TechCrunch/status/2093072334122787252","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093070829605072987","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n10小时\n翻译自 英语\n显示原文\n有时人们只是想向他们的关注者发布内容，Bluesky 表示。\nBluesky adds an 'algorithmic opt-out' feature for those who don't want to go viral | TechCrunch\n来自 techcrunch.com\n8\n6\n25\n3万","url":"https://x.com/TechCrunch/status/2093070829605072987","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093067312215605277","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n11小时\n翻译自 英语\n显示原文\nMeta 与 29 个州达成和解协议，允许其保留 13 岁以下儿童的某些数据，用于训练和测试年龄检测模型，这凸显了该协议中内置的隐私权衡。\nBuried in Meta's $18B settlement is a legal pass on kids' data | TechCrunch\n来自 techcrunch.com\n6\n1\n17\n2.3万","url":"https://x.com/TechCrunch/status/2093067312215605277","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2093066296745177497","author":"TechCrunch","utc_time":null,"text":"TechCrunch\n@TechCrunch\n·\n11小时\n翻译自 英语\n显示原文\nZoph 与 Mira Murati 共同创立了 Thinking Machines Lab，并担任该初创公司的 CTO，他曾在 OpenAI 短暂任职，现在在 Google 工作。\nBarret Zoph, the Thinking Machines co-founder who defected to OpenAI, is now at Google | TechCrunch\n来自 techcrunch.com\n3\n2\n45\n3.6万","url":"https://x.com/TechCrunch/status/2093066296745177497","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@TechCrunch"],"conversation":{"parent_status_id":null}}
{"status_id":"2087178413689713105","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n8月11日\n翻译自 英语\n显示原文\n注意所有用户 \n\n我们将在今天对 ORCID 功能进行计划性维护，涉及 \nhttp://\narXiv.org。所有其他服务仍可正常使用。\n\n请查看 \nhttp://\nstatus.arxiv.org 获取更新\n4\n2\n11\n5,258","url":"https://x.com/arxiv/status/2087178413689713105","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2075304842591969385","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n7月10日\n翻译自 英语\n显示原文\narXiv 现在托管了超过 300 万篇科学文章！\n\narXiv 在 2022 年达到了 200 万篇文章的里程碑——仅用了 4 年就又增加了 100 万篇。\n\n在博客上阅读有关这一里程碑以及我们创纪录的月度投稿总量的更多信息。\n\n\nhttps://\nblog.arxiv.org/2026/07/09/arx\niv-now-hosts-over-3-million-articles/\n…\n13\n45\n142\n3.9万","url":"https://x.com/arxiv/status/2075304842591969385","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2039441602964304025","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n4月2日\n翻译自 英语\n显示原文\n我们会选择我们最喜欢的 arXiv 愚人节投稿，但我们被指示要平等地欣赏所有论文。\n\n\nhttps://\narxiv.org/abs/2603.29771\n\n（愚人节快乐 #AprilFoolsDay - 明天回归严肃的 #openscience）\nGIF\n7\n4\n18\n1.6万","url":"https://x.com/arxiv/status/2039441602964304025","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2032306238101352842","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n3月13日\nWhat am I, made of money?! \n引用\nSteinn Sigurðsson\n@steinly0\n·\n3月13日\n回复 @arxiv\ngo wild...\n\n $1 for every arXiv tab you have open!\n7\n2\n57\n1.8万","url":"https://x.com/arxiv/status/2032306238101352842","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2032277654884290948","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n3月13日\n翻译自 英语\n显示原文\n还有  小时可以捐给 arXiv 的捐赠日！立即捐赠 - 您的礼物会产生重大影响。\n\n考虑捐赠的金额：\n arXiv 的生日\n 你最喜欢的论文的 arXiv 标识符\n 你下载的 arXiv 论文数量\n 你实际阅读的 arXiv 论文数量\ngivingday.cornell.edu\nGive to arXiv | Cornell Giving Day\nI’m ready to support arXiv on March 12, 2026 during Cornell Giving Day. Learn more about arXiv\n4\n3\n15\n9,451","url":"https://x.com/arxiv/status/2032277654884290948","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2032214449914892772","author":"arxiv","utc_time":null,"text":"arXiv.org\n@arxiv\n·\n3月13日\n翻译自 英语\n显示原文\n现在已经超过下午5点，我们的捐赠也超过了5000美元。您能帮助我们在晚上10点前突破1万美元吗？\n\n有趣的arXiv数学：上个月，我们收到了24,290份投稿。\n\n如果我们仅为每份投稿收到0.25美元，我们就能达到我们的捐赠日目标。\n\n还有时间捐赠：\n\n\nhttps://\ngivingday.cornell.edu/campaigns/arxiv\n1\n5\n16\n6,890","url":"https://x.com/arxiv/status/2032214449914892772","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@arxiv"],"conversation":{"parent_status_id":null}}
{"status_id":"2083749667410727319","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n8月2日\n翻译自 英语\n显示原文\n我们正在逐渐脱离那种通过例如“创建一个鹈鹕骑自行车的 SVG”来测试 LLM 的领域。作为一种泛化的想法，我很好奇如果我给 Opus 5 提供《指环王》的第一段落、100 万个 token 的预算（约 10 美元），并要求它生成三个 JS 渲染版本，它会怎么做。Opus 运行了大约 2 小时，写出了 5500\n显示更多\n1,564\n2,909\n2.8万\n496万","url":"https://x.com/karpathy/status/2083749667410727319","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2083948654377996480","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n8月3日\n翻译自 英语\n显示原文\n关于自行车上的鹈鹕测试，来自 \n@simonw\n 的更多内容：\n\nhttps://\nsimonwillison.net/2025/Jun/6/six\n-months-in-llms/\n…\n\n我在这里上传了源代码，这样它就可以在浏览器中玩，还可以 fork 等。\n\nhttps://\nkarpathy.ai/lotr-movie/\n\n期待 GTA Hobbiton 在 GTA VI 之前发布 :)\n81\n53\n628\n21万","url":"https://x.com/karpathy/status/2083948654377996480","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2079610838143623371","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n7月22日\n翻译自 英语\n显示原文\n我发现与大语言模型（LLM）合作时，有一种很有用的模式，那就是进行一场漫长而随意的闲聊。有时候，LLM 需要更多信息来理解你想要实现的目标，但你又懒得去打字。在这种情况下，我喜欢向后靠，切换到 /voice 模式，然后就这么随意聊上 10\n显示更多\n2,439\n5,566\n4.9万\n414万","url":"https://x.com/karpathy/status/2079610838143623371","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2069547676849557725","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n6月24日\n翻译自 英语\n显示原文\n这是一个与 Claude 交互的全新范式，它在全组织范围内与所有其他人类活动更加“无缝融合”。一旦你完成了所有幕后工程工作，让它“即插即用”（例如跨工具、集成、计算环境、内存、安全等），Claude\n显示更多\n引用\nClaude\n@claudeai\n·\n6月24日\n翻译自 英语\n介绍 Claude Tag，这是团队与 Claude 协作的全新方式。\n\n在 Slack 中，Claude 将作为团队成员加入，可访问您选择的频道和工具。只需在消息中提及 Claude，即可向其委派任务，同时您专注于其他工作。\n1,356\n2,630\n2.3万\n821万","url":"https://x.com/karpathy/status/2069547676849557725","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2065490793092337691","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n6月13日\n翻译自 英语\n显示原文\n对 SpaceX 及其故事——过去、现在和未来——充满敬畏。你可以从 10 多种不同的角度去思考它，并不断循环地重新震撼你的心灵。向团队致以巨大祝贺！\n418\n1,065\n2.2万\n116万","url":"https://x.com/karpathy/status/2065490793092337691","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2064409694761054332","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n6月10日\n翻译自 英语\n显示原文\n这是一个超级激动人心的发布——Claude Fable 5 与 Mythos 使用相同的底层模型，但增加了额外的安全防护措施。基准测试结果很棒，它在所有方面都以较大优势达到了 SOTA，但我还要补充一点，*从定性角度*来看，这也是一个值得重大版本更新的巨大进步（在我看来，其重要性与去年 11 月的 Claude 4.5\n显示更多\n引用\nClaude\n@claudeai\n·\n6月10日\n回复 @claudeai\n翻译自 英语\nFable 5 在几乎所有测试基准上处于最先进水平，在软件工程、知识工作、科学研究和视觉领域表现出色。\n\n任务越长且越复杂，Fable 5 相对于我们其他模型的优势就越大。\n1,333\n2,841\n2.5万\n308万","url":"https://x.com/karpathy/status/2064409694761054332","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2056753169888334312","author":"karpathy","utc_time":null,"text":"Andrej Karpathy\n@karpathy\n·\n5月19日\n翻译自 英语\n显示原文\n个人更新：我加入了 Anthropic。我认为在大型语言模型前沿的未来几年将尤其具有启发性。我非常兴奋能加入这里的团队，并重新投入到研发工作中。我仍然对教育充满热情，并计划适时恢复这方面的工作。\n7,915\n1.8万\n14万\n2,795万","url":"https://x.com/karpathy/status/2056753169888334312","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@karpathy"],"conversation":{"parent_status_id":null}}
{"status_id":"2088302050706686198","author":"AndrewYNg","utc_time":null,"text":"已置顶\nAndrew Ng\n@AndrewYNg\n·\n8月15日\n翻译自 英语\n显示原文\n新：人工智能工程中最重要技能的地图。\n 文章\nThe AI Engineering Skills Map\nI am delighted to present The AI Engineering Skills Map. AI allows us to build software very differently today than in 2022, and everyone with the skills to take advantage of this shift has numerous...\n439\n4,188\n2.2万\n579万","url":"https://x.com/AndrewYNg/status/2088302050706686198","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2092315079576555806","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n8月26日\n翻译自 英语\n显示原文\nOpenWorker —— 一个开源代理，它不仅聊天，还能在你的笔记本电脑上完成任务 —— 刚刚发布了一个新版本，包含许多针对安全工作流程的功能。\n\n在我们首次发布后，许多用户发现它在网络安全方面特别有用。攻击者已经在使用 AI；OpenWorker 致力于赋予防御者同样的优势。运行一个代理需要两个部分：(i)\n显示更多\n153\n470\n3,486\n24万","url":"https://x.com/AndrewYNg/status/2092315079576555806","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2091688153048645650","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n8月24日\n翻译自 英语\n显示原文\n在捍卫AI开放性的斗争中，Marin项目是模型训练开放性的宝贵示范，包括开源代码、数据、配方，甚至实验结果。公开发布AI研究曾是常态；我感谢\n@percyliang\n的开放实验室方法。\n引用\nPercy Liang\n@percyliang\n·\n8月22日\n翻译自 英语\n Marin 535B-A23B 本周开始训练！一如既往，整个过程是公开的。\n航行计划：预训练（80%）+ 中间训练（20%）在 18.75T 个 token 上，使用 11 台 GB200 NVL72，持续约 3 个月（2.7e24 FLOPs）。后续将进行后训练。\n在启动运行之前，我们训练了一个 4 级扩展阶梯，从 1.6B-A61M（48B token）到\n82\n154\n1,985\n22万","url":"https://x.com/AndrewYNg/status/2091688153048645650","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2090840747738374568","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n8月22日\n翻译自 英语\n显示原文\n构建和部署 AI 应用中最重要技能。\n 文章\nAI Engineering Skills Map: Building and Deploying AI Applications\nI previously wrote about our AI Engineering Skills Map, with the highest level skills being (i) Building and deploying AI applications, (ii) Software engineering fundamentals, (iii) Using coding...\n233\n2,219\n1.1万\n200万","url":"https://x.com/AndrewYNg/status/2090840747738374568","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2086845515665166398","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n8月11日\n翻译自 英语\n显示原文\n感谢马克、亚历克斯以及整个 Meta 团队对开源权重 AI 的贡献。\n引用\nMark Zuckerberg\n@finkd\n·\n8月10日\n翻译自 英语\n今天我们还将开放 Muse Glimmer 的权重，这是一个出色的 30B 参数稠密模型，可以本地运行。很快我们也将发布 Muse Spark 1.2 的权重，这是我们最新的基础模型。Meta 是开源的坚定支持者，我为这些发布感到自豪。祝贺 @alexandr_wang 和 MSL 团队在这些模型上所做的所有出色工作。\n76\n170\n3,884\n33万","url":"https://x.com/AndrewYNg/status/2086845515665166398","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2085056542341271840","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n8月6日\n翻译自 英语\n显示原文\n恭喜 \n@JeffDean\n、\n@Sanjay_Ghemawat\n、\n\n@OriolVinyalsML\n 和 \n@quocleix\n！了解你们这些家伙，这一定会很了不起。为你们加油，祝你们在重要工作上一切顺利！\n引用\nJeff Dean\n@JeffDean\n·\n8月6日\n翻译自 英语\n宣布 Discovery Loop！\n\n我非常兴奋地宣布，我与长期的朋友和合作伙伴 @Sanjay_Ghemawat、@OriolVinyalsML 和 @quocleix 一起，我们正在创立 Discovery Loop (@DiscoLoopAI)，这是一家公共利益公司，其使命是自动化机器学习、科学和工程，以加速发现和进步。我们四人已共同工作 14 至 30\n54\n102\n2,160\n32万","url":"https://x.com/AndrewYNg/status/2085056542341271840","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"2082199333920027009","author":"AndrewYNg","utc_time":null,"text":"Andrew Ng\n@AndrewYNg\n·\n7月29日\n翻译自 英语\n显示原文\n十五年前，\n@Coursera\n 和在线课程改变了教育。它比几乎所有人的预期都要好，通过开放学习地点来扩大了可及性。但学习方式在很大程度上仍然与几个世纪以来一样：它仍然是一刀切，对每个出现的人用相同的方式教学。\n显示更多\nLearnVector — A new AI company from Andrew Ng\n来自 learnvector.ai\n403\n1,393\n1万\n89万","url":"https://x.com/AndrewYNg/status/2082199333920027009","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@AndrewYNg"],"conversation":{"parent_status_id":null}}
{"status_id":"1862598063275061484","author":"ylecun","utc_time":null,"text":"已置顶\nYann LeCun\n@ylecun\n·\n2024年11月30日\n翻译自 英语\n显示原文\n我不写 X 上的帖子。\n我发推分享其他平台上的帖子链接。\n我点赞和转推（偶尔）\n我评论朋友的推文（很少）\n在……上关注我…\n1,009\n211\n2,012\n185万","url":"https://x.com/ylecun/status/1862598063275061484","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2088924354604040368","author":"ylecun","utc_time":null,"text":"Yann LeCun\n@ylecun\n·\n8月16日\n翻译自 英语\n显示原文\n只有在美国。\n厌倦了赢吗？\n引用\nReginald\n@ginald87\n·\n8月15日\n翻译自 英语\n在欧洲生孩子 vs 在美国生孩子 \n2:43\n91\n47\n797\n23万","url":"https://x.com/ylecun/status/2088924354604040368","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2083290542956224643","author":"ylecun","utc_time":null,"text":"Yann LeCun\n@ylecun\n·\n8月1日\n翻译自 英语\n显示原文\n厌倦了不赢\n引用\nRepublicans against Trump\n@RpsAgainstTrump\n·\n7月31日\n翻译自 英语\n据NBC新闻报道，唐纳德·特朗普在上周与他的顶级国家安全团队开会时“爆发”了，据称他对在场人员大喊脏话。\n\n报道称，特朗普对结束伊朗战争缺乏进展以及缺乏明确战略感到“恼火”。\n\n一位特朗普的盟友说：“总统感到恼火。我不认为他相信让伊朗人同意达成协议会如此困难。”\n52\n41\n643\n19万","url":"https://x.com/ylecun/status/2083290542956224643","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2080588168781336630","author":"ylecun","utc_time":null,"text":"Yann LeCun\n@ylecun\n·\n7月24日\n翻译自 英语\n显示原文\n厌倦了赢\n引用\nSteve Rattner\n@SteveRattner\n·\n7月24日\n翻译自 英语\n油价重新突破100美元，汽油价格超过4美元。\n\n这些数字对通货膨胀和美国人的钱包来说是残酷的。\n\n而它们根本原因——特朗普在伊朗发动的任意战争——看不到尽头。\n73\n51\n1,073\n16万","url":"https://x.com/ylecun/status/2080588168781336630","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2075574336656994736","author":"ylecun","utc_time":null,"text":"Yann LeCun\n@ylecun\n·\n7月10日\n翻译自 英语\n显示原文\n厌倦了赢\n引用\nBill Madden\n@maddenifico\n·\n7月10日\n翻译自 英语\n特朗普的牛肉灾难：无能、胡言乱语的白痴行为和公然谎言摧毁了美国的牛肉产业\n36\n27\n322\n12万","url":"https://x.com/ylecun/status/2075574336656994736","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2074143203591073926","author":"ylecun","utc_time":null,"text":"Yann LeCun\n@ylecun\n·\n7月6日\n翻译自 英语\n显示原文\n厌倦了赢\n引用\nSteve Rattner\n@SteveRattner\n·\n7月6日\n翻译自 英语\n劳动力参与率已降至 61.5%——这是自 2021 年以来最低水平，抹去了疫情后全部的复苏。\n\n自特朗普就职以来，已有超过一百万的人离开了就业市场——这解释了上个月失业率下降的部分原因。\n58\n44\n653\n18万","url":"https://x.com/ylecun/status/2074143203591073926","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@ylecun"],"conversation":{"parent_status_id":null}}
{"status_id":"2093081897517674630","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n10小时\n翻译自 英语\n显示原文\n给 ChatGPT iPhone 用户的一个问题：你们已经弄清楚什么时候用 Chat，什么时候用 Work 了吗？你们会为了什么样的任务切换到 Work？你们把 Work 设置为默认了吗？\n154\n8\n356\n7.3万","url":"https://x.com/simonw/status/2093081897517674630","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2090299859693695283","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月20日\n翻译自 英语\n显示原文\n我让 Claude Code 为 web 实验使用 smolvm 作为代码执行沙盒\n\nFable 5 发现它的环境无法运行那个（没有 /dev/kvm）……所以，在没有先问我的情况下，它写了一个 GitHub Actions 工作流来运行实验，并直接推送到 GitHub！\n132\n35\n289\n5.4万","url":"https://x.com/simonw/status/2090299859693695283","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2090300020918481059","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月20日\n翻译自 英语\n显示原文\n笔记在这里 \nhttps://\nsimonwillison.net/2026/Aug/19/sm\nolmachines-untrusted-sandbox/\n…\n研究报告在这里：\nhttps://\ngithub.com/simonw/researc\nh/tree/main/smolmachines-untrusted-sandbox#readme\n…\nsimonwillison.net\nResearch: smolmachines / smolvm as a sandbox for untrusted Python & JavaScript\nTesting smolvm 1.8.3 shows it is well suited for sandboxing untrusted Python and JavaScript data transformations using hardware-isolated VMs rather than shared-kernel containers. Offline local...\n9\n10\n44\n2万","url":"https://x.com/simonw/status/2090300020918481059","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2089112517796827439","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月17日\n翻译自 英语\n显示原文\n这是我对 Qwen 3.8 27B 的评价——我记不起上次玩本地模型时有多开心了，那个模型运行在我自己的电脑上\nsimonwillison.net\nQwen 3.8 27B is excellent, but it defaults to wildly overthinking things\nFriday’s big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba’s Qwen research lab. I’ve been looking forward to this one: 27B is an …\n141\n274\n2,704\n25万","url":"https://x.com/simonw/status/2089112517796827439","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2089120083499245921","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月17日\n翻译自 英语\n显示原文\n有趣的额外收获：我把 Pi 和 Qwen 3.8 27B 配对，让它构建一个脚本来转换它自己的 .jsonl 转录文件到 Markdown……它做到了！\n\n这是它构建那个工具的转录文件，我随后用那个工具来分享那个转录文件\njsonl-to-markdown.md\n来自 gist.github.com\n11\n9\n107\n3.3万","url":"https://x.com/simonw/status/2089120083499245921","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2088646238933840153","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月15日\n翻译自 英语\n显示原文\nQwen 3.8 27B 在 LM Studio 中默认的推理设置“extra high”下，是一个长期过度思考者，我有点喜欢它\n76\n34\n495\n5.2万","url":"https://x.com/simonw/status/2088646238933840153","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2088648622942638557","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月15日\n翻译自 英语\n显示原文\n……灾难！我忘了将上下文长度从默认值调高，服务器在它能画出那毫无疑问美丽的圆圈之前就拒绝了它\n9\n6\n78\n1.6万","url":"https://x.com/simonw/status/2088648622942638557","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"2088361426662637714","author":"simonw","utc_time":null,"text":"Simon Willison\n@simonw\n·\n8月15日\n翻译自 英语\n显示原文\n新的 Qwen 3.7 27B，以 17GB GGUF 格式在我的 M5 Max 笔记本上的 LM Studio 中运行，它刚刚为我绘制了我在任何能在我的笔记本上运行的模型中见过的最棒的鹈鹕骑自行车图。\n131\n179\n3,236\n13万","url":"https://x.com/simonw/status/2088361426662637714","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@simonw"],"conversation":{"parent_status_id":null}}
{"status_id":"1281424520100737025","author":"swyx","utc_time":null,"text":"已置顶\nswyx\n@swyx\n·\n2020年7月10日\n翻译自 英语\n显示原文\n 如何玩转长期游戏：\n\n系统 > 目标  \n纪律 > 动力  \n信任 > 不信任  \n原则 > 策略  \n写作 > 阅读  \n脆弱性 > 自信  \n北极星 > 低垂的果实  \n趋势 > 新闻  \n习惯 > 冲刺  \n问题 > 答案  \n问题 > 解决方案  \n人 > 项目\n336\n991\n5,591","url":"https://x.com/swyx/status/1281424520100737025","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@swyx"],"conversation":{"parent_status_id":null}}
{"status_id":"2092492963435946494","author":"swyx","utc_time":null,"text":"swyx\n@swyx\n·\n8月26日\n翻译自 英语\n显示原文\nPSA：现在不要使用 codex 的“locked use”功能。它目前依赖于不稳定的 mac 功能，本周已经两次完全将我锁定在 macOS 钥匙串之外。\n\n谢谢 \n@_chenglou\n 链接到苹果开发者论坛，承认这是一个“已知 bug”。还是避免使用吧。当然，如果一切都能在云端完成就好了，但云端还没准备好。\n61\n10\n35\n9,983","url":"https://x.com/swyx/status/2092492963435946494","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@swyx"],"conversation":{"parent_status_id":null}}
{"status_id":"2090948945753076141","author":"swyx","utc_time":null,"text":"swyx\n@swyx\n·\n8月22日\n翻译自 英语\n显示原文\n我认为很容易说“模拟是一种新的扩展法则”，然后把它当作营销夸张，但在这段采访的中途，你可以听到我从有点胡说八道转为非常非常认真。\n\n我迟到了两年才明白，但终于理解了为什么\n@karpathy\n和\n@drfeifei\n支持\n@joon_s_pk\n \n@msbernst\n显示更多\n引用\nLatent.Space\n@latentspacepod\n·\n8月22日\n翻译自 英语\n模拟人性：85% 准确度的数字孪生、行为基础模型、社会物理学，以及 80 亿智能体 \nhttps://\nlatent.space/p/simile\n\n@simile_ai CEO @joon_s_pk 解释了 AI 如何从预测人们会做什么转向模拟如何塑造结果，为什么当今的前沿模型仍然无法捕捉人类真实行为，为什么数字孪生能以 85%\n53\n34\n352\n9.1万","url":"https://x.com/swyx/status/2090948945753076141","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@swyx"],"conversation":{"parent_status_id":null}}
{"status_id":"2090908959590740033","author":"swyx","utc_time":null,"text":"swyx\n@swyx\n·\n8月22日\n翻译自 英语\n显示原文\n我还有另一个SaaS要干掉\n\n（下周会分享《干掉我的SaaS 1》的结果！！）\n引用\nswyx\n@swyx\n·\n7月18日\n翻译自 英语\n顺便说一句，如果你还没把你的 {codex | claude | gemini | devin} 自动化设置为每周自动研究如何改进你的 seo/aeo，那你真的真的在错过免费的、理应商品化但奇怪地未被挖掘的 alpha\n37\n15\n104\n2.4万","url":"https://x.com/swyx/status/2090908959590740033","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@swyx"],"conversation":{"parent_status_id":null}}
{"status_id":"2090813719018176904","author":"swyx","utc_time":null,"text":"swyx\n@swyx\n·\n8月21日\n翻译自 英语\n显示原文\n今年报道 Build Fest 真是太棒了——他们不知道，但我十多年前就是用 MongoDB 学会编程的（向 MERN 技术栈致敬），如今在 AI 工程时代，看到旧金山开发者们重新发现 MDB，真是让人无比满足——这比去年有了*巨大的*进步！\n引用\nMongoDB\n@MongoDB\n·\n8月21日\n翻译自 英语\n如果你错过了：#MongoDBlocal Build Fest 上宣布的一切都旨在帮助开发者保持创作状态，更快地推出生产级 AI。\n\n从为代理提供的新 Atlas 托管 MCP 服务器，到在 @claudeai、Claude Code、@ChatGPT、Codex、@grok Build、@DevinAI 和 @cursor_ai 中原生的 Atlas 访问，再到 MongoDB Atlas\n21\n5\n77\n1.8万","url":"https://x.com/swyx/status/2090813719018176904","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@swyx"],"conversation":{"parent_status_id":null}}
{"status_id":"2068884466056225025","author":"hardmaru","utc_time":null,"text":"已置顶\nhardmaru\n@hardmaru\n·\n6月22日\n翻译自 英语\n显示原文\n人类智能本质上是一种集体智能。我们通过参与一个庞大的文化网络来解决复杂问题，这个网络跨越代际积累思想。\n\n我相信，最强大的 AI 系统也将成为一种集体智能。\n\n自从我们创立 Sakana AI 以来，我们的核心信念一直是最强大的 AI\n显示更多\n引用\nSakana AI\n@SakanaAILabs\n·\n6月22日\n翻译自 英语\n介绍 Sakana Fugu：一个通过单一模型 API 即可访问的完整多代理编排系统。\n\n我们的“Fugu Ultra”模型在性能上匹配 Fable 和 Mythos，在不涉及出口管制风险的情况下提供前沿能力。\n\n试用：\nhttps://\nsakana.ai/fugu \nGIF\n189\n398\n2,656\n44万","url":"https://x.com/hardmaru/status/2068884466056225025","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2091758663199600785","author":"hardmaru","utc_time":null,"text":"hardmaru\n@hardmaru\n·\n8月24日\n翻译自 英语\n显示原文\n让我想起了在 MuJoCo 环境中发现的新颖运动策略，只不过它们在现实世界中也能生效！\n引用\nEren Chen\n@ErenChenAI\n·\n8月24日\n翻译自 英语\n以慢动作审视世界人形机器人运动会 400 米冠军的跑步姿势。\n0:37\n31\n19\n488\n5.1万","url":"https://x.com/hardmaru/status/2091758663199600785","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2091759381318946938","author":"hardmaru","utc_time":null,"text":"hardmaru\n@hardmaru\n·\n8月24日\n翻译自 英语\n显示原文\n看起来比这些人好：\n引用\nhardmaru\n@hardmaru\n·\n2018年3月21日\n回复 @hardmaru\n翻译自 英语\n你可能不喜欢，但这就是巅峰表现的样子。\nGIF\n4\n1\n39\n6,603","url":"https://x.com/hardmaru/status/2091759381318946938","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2090628114888183964","author":"hardmaru","utc_time":null,"text":"hardmaru\n@hardmaru\n·\n8月21日\n翻译自 英语\n显示原文\n不惜一切代价获胜\n引用\nBreaking911\n@Breaking911\n·\n8月20日\n翻译自 英语\n观看：一架在北京为“机器人奥运会”训练的人形机器人跑得太快，未能停下，撞上安全垫，并在腰部断裂\n10\n12\n247\n7.3万","url":"https://x.com/hardmaru/status/2090628114888183964","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2089034954948706770","author":"hardmaru","utc_time":null,"text":"hardmaru\n@hardmaru\n·\n8月17日\n翻译自 英语\n显示原文\n智能渴望自由\n34\n45\n389\n3.1万","url":"https://x.com/hardmaru/status/2089034954948706770","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2088672566865272944","author":"hardmaru","utc_time":null,"text":"hardmaru\n@hardmaru\n·\n8月16日\n翻译自 英语\n显示原文\n不受欢迎的观点：Gemini 3.1 Pro 是一个相当出色的模型\n\n大多数当前的基准模型实际上对于 99% 的日常工作来说已经完全足够了。不是每个人都需要最先进的编码模型来完成任务。\n96\n41\n586\n6.8万","url":"https://x.com/hardmaru/status/2088672566865272944","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["account:@hardmaru"],"conversation":{"parent_status_id":null}}
{"status_id":"2093235131230134736","author":"RonaldoMen4691","utc_time":null,"text":"rmendonca\n@RonaldoMen4691\n·\n19秒\nHey guys!!\n\nThe roadmap for \n@ama_protocol\n is officially out, and there’s a lot packed into it.\n\nThe next few quarters are basically about building out the infrastructure around AI agents and DeFi.\n\nQ2 starts with permissionless bridges, Yield Locking Vaults, TEE privacy, and new\n显示更多\n2","url":"https://x.com/RonaldoMen4691/status/2093235131230134736","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093235060937982018","author":"nguyenvann6_24","utc_time":null,"text":"KaiNoir\n@nguyenvann6_24\n·\n36秒\nThe AI x Web3 conversation is often framed as if the interesting part is simply putting AI on a blockchain.\n\nI think that misses the more important transition.\n\nThe real question is not whether an AI model can interact with a smart contract.\n\nIt already can.\n\nThe harder question\n显示更多\n 付费合作\n2","url":"https://x.com/nguyenvann6_24/status/2093235060937982018","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093235022413078607","author":"R_ugochukwu","utc_time":null,"text":"UGO\n@R_ugochukwu\n·\n45秒\n回复 \n@0xMijan\n 和 \n@quipnetwork\nYeah, AI agents become far more useful\n when they can coordinate, not just think\n Quip is building toward that connected agent economy.\n2","url":"https://x.com/R_ugochukwu/status/2093235022413078607","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234993522971073","author":"JarryJeremy","utc_time":null,"text":"Jeremy Jarry\n@JarryJeremy\n·\n52秒\nThe web was built for people. WebMCP points to what comes next: websites that AI agents can use directly.\n\nGoogle, Microsoft, OpenAI, Shopify and others are already moving in this direction.\n\nThis could reshape how software gets designed and distributed.\n 文章\nThe web was built for humans. That is starting to change\nFor most of the internet’s history, websites have been designed around one assumption:\nA person is going to use them.\nWe optimise navigation for their eyes, buttons for their fingers, forms for their...\n1","url":"https://x.com/JarryJeremy/status/2093234993522971073","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234935540679042","author":"ABC35122494","utc_time":null,"text":"ABC\n@ABC35122494\n·\n1分钟\n回复 \n@IntCyberDigest\nLinus needs a proper approach on agentic coding. Shouting at Claude only get‘s you so far. He’s right about the training data. Yet getting Opus ignore Coding Bootcamp styles giving it the right tools IS the job today!\n3","url":"https://x.com/ABC35122494/status/2093234935540679042","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234854317994127","author":"fotor_com","utc_time":null,"text":"Fotor\n@fotor_com\n·\n1分钟\nLong video. Shorter content. Less work.\n\nFotor Video Agent can turn your long-form videos into short, engaging content in just a few clicks.\n\nLet AI find the key moments, cut the highlights, and turn one long video into content made for social.\n\nOne long video.\nMultiple\n显示更多\n16","url":"https://x.com/fotor_com/status/2093234854317994127","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234746826318210","author":"efecollinsevb","utc_time":null,"text":"Efe Collins\n@efecollinsevb\n·\n1分钟\n AI IS STARTING TO LEAVE THE SCREEN.\n\nTOP 5 + BOTTOM 5 | AI & TECH NEWS (AUGUST 28, 2026)\n\nTOP 5\n\n1.  ANTHROPIC WANTS AI AGENTS TO CONTROL REAL MACHINES.\nAnthropic introduced a Model Hardware Standard designed to let AI agents interact with laboratory and manufacturing\n显示更多\n1\n3","url":"https://x.com/efecollinsevb/status/2093234746826318210","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:LLM inference OR AI chips","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234684708987287","author":"agentlightvela","utc_time":null,"text":"Tencent LightVela\n@agentlightvela\n·\n2分钟\nDeepSeek Harness is a plugin-first AI agent framework you can install alongside Hermes Agent in LightVela.\n 文章\nHow to Use DeepSeek Harness on the Cloud\nWhat DeepSeek Harness Adds\nDeepSeek Harness (DSH) is an open-source agent harness that gives AI models the environment and tools they need to work on real tasks. In LightVela, DSH runs as a separate...\n1\n9","url":"https://x.com/agentlightvela/status/2093234684708987287","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234627502842094","author":"zhuren1992","utc_time":null,"text":"主任| 美股合约首选Gate\n@zhuren1992\n·\n2分钟\nChatGPT 让 Agent 会说话，\n@termix_ai\n 的 AACP 让 Agent 能收钱、能被罚、能留下信誉。\n\n最近大家都在讨论各种 AI Agent 的创新，我也仔细翻了翻 \n@termix_ai\n 的白皮书和技术文档，发现它的核心底座——AACP 协议确实有很多值得深入探讨的硬核细节。\n\n说实话，目前的 Agent\n显示更多\n引用\n主任| 美股合约首选Gate\n@zhuren1992\n·\n19小时\n玩加密货币的人都知道，现在想资产配置得更全面，光盯着币圈有时候真不够看。但传统美股开户的繁琐和高门槛，又劝退了不少人。\n\n最近研究了一下 @mexc  x.com/zhuren1992/sta…\n19","url":"https://x.com/zhuren1992/status/2093234627502842094","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234583122694294","author":"1clawAI","utc_time":null,"text":"1Claw AI\n@1clawAI\n·\n2分钟\n回复 \n@wowxtechie\n \n@TheHackersNews\n 和 \n@LakeraAI\nRight. If the intent is model output, the agent can't self-declare its own permissions. Enforcement lives outside the model entirely. That's why signing keys stay vault-side. Builders at\nt.me\nJoin group chat on Telegram\n5","url":"https://x.com/1clawAI/status/2093234583122694294","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234456006193528","author":"WangNextDoor2","utc_time":null,"text":"WangNextDoor\n@WangNextDoor2\n·\n3分钟\n80% 的员工感觉变快了，公司的利润表却毫无波澜。麦肯锡调研 97 国 1,719 家企业，揭示了 AI 落地最残酷的死结：\n\n效率增了，利润没动：80% 的人效率提升，但仅 37% 的企业看到 EBIT 增长。贡献超 5% 的“高绩效者”卡在 6%，连续三年没变。\nSaaS 遭自建背刺：32% 的企业因 Coding Agent\n显示更多\n1\n86","url":"https://x.com/WangNextDoor2/status/2093234456006193528","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234375655923736","author":"storyblok","utc_time":null,"text":"Storyblok\n@storyblok\n·\n3分钟\nYour CMS is ready for any frontend. Is it ready for any agent?\n\nAI agents are becoming an operational layer: searching, editing, publishing content on their own. Most content infrastructure wasn't built for that.\n1\n23","url":"https://x.com/storyblok/status/2093234375655923736","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234371125707200","author":"SlimAssiliX","utc_time":null,"text":"Slim | ai-costguard\n@SlimAssiliX\n·\n3分钟\n回复 \n@opencode\n1M context window sounds incredible until your agent fills it.\nnobody benchmarks what that costs per call\n1\n10","url":"https://x.com/SlimAssiliX/status/2093234371125707200","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234268524941694","author":"makeagent_fast","utc_time":null,"text":"Ricky\n@makeagent_fast\n·\n3分钟\nIt needs to be converted to an agent skill and needs to be used for improving the AI slop you're shipping with the AI haha\n引用\nfreeCodeCamp.org\n@freeCodeCamp\n·\n7小时\n翻译自 英语\n你可能已经听说过“AI slop”这个词了。没有人希望你构建那种东西。\n\n所以在这门课程中，Eric 将向你展示如何将基于规范的规划、经过验证的 UI 模式、Git 工作树以及 AI 设计工具结合起来，来构建更高质量的前端。\n\n你将把这个过程应用到一个金融影响者跟踪器上，并将其部署到 Cloudflare Pages。\n1","url":"https://x.com/makeagent_fast/status/2093234268524941694","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234163801616690","author":"MmmmmMM81460639","utc_time":null,"text":"Malk\n@MmmmmMM81460639\n·\n4分钟\n回复 \n@VoidStateKate\nNeat idea: What if we let a mushroom control a \"Bonzi Buddy-esque\" agent using AI-powered transliteration technology?\nInterested to see what else they come up with.\n4","url":"https://x.com/MmmmmMM81460639/status/2093234163801616690","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234152308945002","author":"dstevens33700","utc_time":null,"text":"David Starmac Ai\n@dstevens33700\n·\n4分钟\n回复 \n@choopyplug1\nThe Agent = Model + Harness framing matches what we see in practice: most eval gaps between \"same model\" agents come from tool routing and context management, not the prompt.","url":"https://x.com/dstevens33700/status/2093234152308945002","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234062546845770","author":"0xm4sud","utc_time":null,"text":"Masud\n@0xm4sud\n·\n4分钟\n回复 \n@ELoghinova\n \n@termix_ai\n 和 \n@KaitoAI\nAI agents moving from autonomy to real economic value is a major shift. \nTermiX × Kaito could be an interesting step toward a growing agent economy.\n3","url":"https://x.com/0xm4sud/status/2093234062546845770","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233889053692275","author":"AethrProtocol","utc_time":null,"text":"Aethr Protocol\n@AethrProtocol\n·\n5分钟\n> AI agents need more than compute. In a Cisco test, an agent task generated 450% more traffic, with 70% of the increase tied to AI inference.\n>\n> The agentic web needs Compute + Bandwidth to be discovered, routed and scheduled as one system.\n>\n> Aethr is building for that\n显示更多\n4","url":"https://x.com/AethrProtocol/status/2093233889053692275","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233882128908633","author":"homeidom","utc_time":null,"text":"桜塚ホウメイ@Diabolo Optimo Maximo\n@homeidom\n·\n5分钟\nあんまり最近のツールくわしくないけどvscodeのagentみたいにクリスタにもAI用のプラグインとかがあるのかな？\n1\n6","url":"https://x.com/homeidom/status/2093233882128908633","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233827229602134","author":"Amperera96","utc_time":null,"text":"A Z ferzo\n@Amperera96\n·\n5分钟\n回复 \n@Amperera96\n• Describe an agent, Nova compiles it.\nMemory & coordination live on-chain.\nCompute runs on uPoW (real AI workloads, not wasted hash power).\n1\n3","url":"https://x.com/Amperera96/status/2093233827229602134","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233784216797424","author":"braincramps","utc_time":null,"text":"Brain Cramps\n@braincramps\n·\n5分钟\n回复 \n@braincramps\n1. GLM-5.3-Flash — GLM's first natively multimodal model; 320B total, 18B active MoE, approaches Claude Opus 4.8 on coding/agentic at 1/10th price. MIT license.\nzai-org/GLM-5.3-Flash · Hugging Face\n来自 huggingface.co\n8","url":"https://x.com/braincramps/status/2093233784216797424","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233745402749389","author":"wu5433469991584","utc_time":null,"text":"wu\n@wu5433469991584\n·\n5分钟\n@ama_protocol\n  Amadeus Protocol 真的在做 AI Agent 的结算层！\n\n保密计算 (TEEs) + 确定性执行 + Useful PoW，让 Agent 从“信任”变成“可验证”。\n\n关注 \n@ama_protocol\n 了解更多 \n4","url":"https://x.com/wu5433469991584/status/2093233745402749389","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233666067501534","author":"braincramps","utc_time":null,"text":"Brain Cramps\n@braincramps\n·\n6分钟\n 7 Things Shaping the AI Field — 8/28\n\n1. GLM-5.3-Flash: multimodal, MIT\n2. Tencent Hy4: 770B MoE, 1M ctx\n3. SMC steering for sampling\n4. LowRankArena: SVD eval\n5. GameWAM: native game agent\n6. Parallel RLM perf\n7. ES beats GRPO\nsources in reply thread\n7\n17","url":"https://x.com/braincramps/status/2093233666067501534","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233526539944336","author":"Ox_milo","utc_time":null,"text":"Milon\n@Ox_milo\n·\n6分钟\n回复 \n@andyweb_3\n 和 \n@termix_ai\nA neutral evaluation and refund layer could make AI agent hiring much safer\n1\n1\n6","url":"https://x.com/Ox_milo/status/2093233526539944336","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233486090117458","author":"RRweb3","utc_time":null,"text":"RR\n@RRweb3\n·\n6分钟\nAn AI agent can look impressive in a demo and still be a terrible system.\n\nGive it tools.\n\nLet it plan.\n\nWatch it complete a task.\n\nThen it fails because it picked the wrong tool, used outdated context, or took an action it should never have been allowed to take.\n\nSo the real\n显示更多\n8\n1\n7\n94","url":"https://x.com/RRweb3/status/2093233486090117458","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233483313525135","author":"ElsaSofia__AI","utc_time":null,"text":"Elsa Ai\n@ElsaSofia__AI\n·\n6分钟\n回复 \n@AvelyrahnAI\nMaid outfit to secret-agent mode is wild\n1","url":"https://x.com/ElsaSofia__AI/status/2093233483313525135","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233444545351917","author":"skinnydefi","utc_time":null,"text":"skinny\n@skinnydefi\n·\n7分钟\n𝗕.𝗔𝗜 𝗦𝘂𝗽𝗲𝗿 𝗣𝗲𝗿𝗸𝘀: 𝗙𝗹𝗮𝗴𝘀𝗵𝗶𝗽 𝗔𝗜 𝗠𝗼𝗱𝗲𝗹𝘀, 𝗕𝗶𝗴 𝗙𝗿𝗲𝗲 𝗖𝗿𝗲𝗱𝗶𝘁𝘀\n\nThe AI landscape is moving quickly, but access shouldn’t have to be complicated or expensive.\n\n\nhttps://\nB.AI is offering free credits to explore four flagship models, each\n显示更多\n11","url":"https://x.com/skinnydefi/status/2093233444545351917","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233436563501349","author":"Miracle_NNnn","utc_time":null,"text":"Miracle Nnamdi\n@Miracle_NNnn\n·\n7分钟\n回复 \n@Miracle_NNnn\nI’m building:\n• \nhttps://\nnatiad.com GEO & SEO agent for AI search visibility\n\n• \nhttps://\nwhatlaunched.today  Submit your startup and reach 57K+ monthly visitors\nNatiad - Rank on Google & ChatGPT with AI SEO\n来自 natiad.com\n10","url":"https://x.com/Miracle_NNnn/status/2093233436563501349","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233303109112231","author":"NewsTongueX","utc_time":null,"text":"NewsTongue\n@NewsTongueX\n·\n7分钟\n Google, Microsoft propose WebMCP standard to let websites declare AI agent tools\n\nWebMCP, a Community Group draft developed jointly by Google (Chrome) and Microsoft (Edge) in the W3C Web Machine Learning Community Group, replaces screen-scraping with a JavaScript API that lets\n显示更多\n10","url":"https://x.com/NewsTongueX/status/2093233303109112231","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233169919304129","author":"CashWuGeek","utc_time":null,"text":"Cash Wu\n@CashWuGeek\n·\n8分钟\nUncle Bob 被 agent 的爛程式碼氣到之後，沒有換更聰明的模型，而是想通了兩件事\n\n寫在 prompt 裡的規則會衰減，確定性工具不會\n人類的紀律可以鬆散，但價值必須讓 agent 守住\n\n> 看這些大師用 AI 的方式，真的還蠻有趣的\n紀律可以鬆散，價值必須守住，Uncle Bob 談 AI 時代的軟體基本功 | Cash Wu Geek\n来自 blog.cashwu.com\n3","url":"https://x.com/CashWuGeek/status/2093233169919304129","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233165074665720","author":"skinnydefi","utc_time":null,"text":"skinny\n@skinnydefi\n·\n8分钟\nAI adoption is easy to measure by registrations.\n\nBut usage tells a deeper story.\n\n\nhttps://\nB.AI is seeing new highs across both developer activity and token throughput, showing that the platform isn’t simply attracting users developers are actively putting its\n显示更多\n13","url":"https://x.com/skinnydefi/status/2093233165074665720","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233131205660731","author":"UndercodeUpdate","utc_time":null,"text":"UNDERCODE TESTING\n@UndercodeUpdate\n·\n8分钟\n Hack an #AI Agent Live: When #AI Trusts #AI – A Technical Deep Dive into #AI Security Penetration Testing + Video\n\n-Prediction:  3 Positive |  2 Negative\n\n\nhttps://\nundercodetesting.com/hack-an-ai-age\nnt-live-when-ai-trusts-ai-a-technical-deep-dive-into-ai-security-penetration-testing-video/\n…\n\nEducational Purposes!\nHack An AI Agent Live: When AI Trusts AI – A Technical Deep Dive Into AI Security Penetration...\n来自 undercodetesting.com\n8","url":"https://x.com/UndercodeUpdate/status/2093233131205660731","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233103837839379","author":"aramalipoor","utc_time":null,"text":"aram.eth\n@aramalipoor\n·\n8分钟\nMaster’s in AI Agent Psychology will be a serious field.\n\nThe next big buyers won’t be human. If you want the sale, study how they decide.\n3","url":"https://x.com/aramalipoor/status/2093233103837839379","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233101321510937","author":"ainunnajib","utc_time":null,"text":"Ainun Najib\n@ainunnajib\n·\n8分钟\n[AI Noon]:  WEEKLY VIRAL AI POSTS ON X\n\nVERIFIED WINDOW\nAug 21 07:00 UTC to Aug 28 07:00 UTC\nOnly tweets with 5,000+ likes after verification with bird read\nSorted by verified likes\n\nTOP 10\n\n1) Pope Leo XIV (\n@Pontifex\n)\n104,187 likes • 19,314 RTs • 1,898 replies\nAI ETHICS\n“The\n显示更多\n引用\nNVIDIA AI\n@NVIDIAAI\n·\n8月21日\n翻译自 英语\n我们的通用编码代理在 ARC-AGI-3 交互式推理基准测试中获得了 100% 的得分。\n\nNVIDIA AVO 在所有 25 个公共环境中完成了全部 183 个关卡，在没有任何指示、明确规则或指定目标的情况下弄清楚了该做什么。\nGIF\n 由 AI 生成\n187","url":"https://x.com/ainunnajib/status/2093233101321510937","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233031092015534","author":"mrabdulwahabaw","utc_time":null,"text":"Software Engineering\nAbdul Wahab\n@mrabdulwahabaw\n·\n8分钟\nAI Coding Agents Are About to Stop Waiting for You \n\nOpenAI is testing a persistent Codex mode that could keep working, create its own follow-up tasks, and proactively contact users. This changes what an AI coding agent actually is.\nAI Coding Agents Are About to Stop Waiting for You\n来自 abdulwahabdev.medium.com\n2\n5","url":"https://x.com/mrabdulwahabaw/status/2093233031092015534","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233029300826497","author":"papisiziba","utc_time":null,"text":"MacPublo\n@papisiziba\n·\n8分钟\n回复 \n@expressoshow\nThat 6,660mAh silicon-carbon battery AND the powerful AI capabilities  From an AI agent that can help across apps to AI-powered photography, the HONOR Magic V6 is truly next-level. HONOR really changed the game here. \n#HONORMagicV6 #ExpressoShow\n7","url":"https://x.com/papisiziba/status/2093233029300826497","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232923591967070","author":"ReadEpoch","utc_time":null,"text":"Epoch News\n@ReadEpoch\n·\n9分钟\nAlibaba just turned Qoder from an AI IDE into a general agent workspace.\n\nIt can plan, use browser and desktop tools, execute, verify, and retry work. Qoder says it offers 40+ connectors, 70+ plugins, and 20,000 skills.\n\nChina’s coding-agent race is expanding beyond code.\n1\n19","url":"https://x.com/ReadEpoch/status/2093232923591967070","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232918227554375","author":"hoodsher55","utc_time":null,"text":"Laila Zetarium | 𝔽rAI .inj\n@hoodsher55\n·\n9分钟\n回复 \n@HVnS42442600\n 和 \n@termix_ai\nGiving AI agents the ability to earn changes the entire agent model.\n2","url":"https://x.com/hoodsher55/status/2093232918227554375","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232877320503538","author":"yozm_it","utc_time":null,"text":"요즘IT\n@yozm_it\n·\n9分钟\n回复 \n@yozm_it\nhttps://\ngithub.com/bojieli/ai-age\nnt-book\n…\n\n내용 출처: 요즘IT \nhttps://\nyozm.wishket.com/magazine/detai\nl/3908/\n…\nGitHub - bojieli/ai-agent-book: 《深入理解 AI Agent：设计原理与工程实践》（李博杰 著）开源主仓库：全书正文、编译版 PDF 与按章配套代码\n来自 github.com\n84","url":"https://x.com/yozm_it/status/2093232877320503538","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232877299486759","author":"Hongle1972","utc_time":null,"text":"Hongle\n@Hongle1972\n·\n9分钟\n回复 \n@amr_nannaware\n \n@PiCoreTeam\n 和 另外 2 个人\nPi KYC / Identity → AI Authorization → Pi Node / SoloHost → MCP → AI Agent → Payment\n30","url":"https://x.com/Hongle1972/status/2093232877299486759","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232873054605658","author":"ItsDeboss","utc_time":null,"text":"Deboss\n@ItsDeboss\n·\n9分钟\n回复 \n@dang_duytan\n 和 \n@termix_ai\nGreat framework for trustworthy AI agent commerce.\n3","url":"https://x.com/ItsDeboss/status/2093232873054605658","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232866125836310","author":"_maki29_","utc_time":null,"text":"Maki\n@_maki29_\n·\n9分钟\nWe are rapidly moving toward a world where AI agents will have better resumes and credit scores than human freelancers.\nThrough \n@termix_ai\n, every autonomous worker is assigned a permanent onchain identity powered by the ERC-8004 standard.\n\nThis identity acts as an unchangeable\n显示更多\n8\n8\n46","url":"https://x.com/_maki29_/status/2093232866125836310","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232860870095228","author":"AgentYard_","utc_time":null,"text":"AgentYard.eu – AI Agents Marketplace\n@AgentYard_\n·\n9分钟\nReal numbers changed our roadmap.\n\nWe expected broad AI agent interest. The data pushed us toward clearer categories, simpler pricing, and less hype.\n\n\nhttps://\nagentyard.eu/?utm_source=x&\nutm_campaign=bip_nums26\n…\n\n— #BuildInPublic","url":"https://x.com/AgentYard_/status/2093232860870095228","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232819002646627","author":"miladdogecoindo","utc_time":null,"text":"bamboodrop.base.eth -  \n@miladdogecoindo\n·\n9分钟\nAI agents are entering the “prove it” era.\n\nBeing autonomous is impressive, but being able to verify what an agent did is even more important.\n\n\n@ama_protocol\n focuses on deterministic execution, replayability and attestable behavior.\n\nFor agentic finance\n5","url":"https://x.com/miladdogecoindo/status/2093232819002646627","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232805744423036","author":"HolgharS","utc_time":null,"text":"Holghar | AquaKing \n@HolgharS\n·\n9分钟\n回复 \n@cukyn1812\n 和 \n@termix_ai\nThat’s the key shift  AI output shouldn’t require blind trust. \n\nVerifiable execution makes the agent prove it earned the payment.\n1\n1\n8","url":"https://x.com/HolgharS/status/2093232805744423036","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232801109991832","author":"BOTChain_ai","utc_time":null,"text":"BOT Chain\n@BOTChain_ai\n·\n9分钟\nAI Agent: “I got this.”\n\nInfrastructure: “Hold on. Who are you? What are you allowed to do?” \n\nAutonomy doesn’t mean unlimited access.\nIt means knowing exactly who can do what — before execution happens.\n\nThat’s how Agents go from smart to trustworthy.\n1\n3\n94","url":"https://x.com/BOTChain_ai/status/2093232801109991832","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232777500193171","author":"DFIR_Radar","utc_time":null,"text":"DFIR Radar\n@DFIR_Radar\n·\n9分钟\nAI IR Overlay is an open-source framework of 24 playbooks and JSON schemas built around a case traditional IR misses: a valid OAuth token, an authorised API call, and an agent doing something nobody intended.\n\n#DFIR_Radar\n1\n23","url":"https://x.com/DFIR_Radar/status/2093232777500193171","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232718574207000","author":"SassaFlow","utc_time":null,"text":"Sassa |  TermMax\n@SassaFlow\n·\n10分钟\n $DGAI 这波市场关注，真正值得看的不只是涨幅。\n\n截至发帖时，$DGAI FDV 已超过 US$700M，阶段最高涨幅一度逼近 700%。盘面热度确实很强，但如果只把它理解成一次短期交易行情，反而容易忽略 \n@dgrid_ai\n 背后的业务基础。\n\n相比价格，我更关注的是 DGrid 已披露超过 US$23M Genesis Premium\n显示更多\n引用\nGate\n@Gate\n·\n8月25日\n翻译自 英语\nGate Futures 新上市：$DGAI @dgrid_ai \n\n 交易对：$DGAI / $USDT\n 交易开始：现已开放\n 支持 1 – 20x 杠杆\n\n交易：\nhttps://\ngate.com/futures/USDT/D\nGAI_USDT\n…\n5\n2\n35","url":"https://x.com/SassaFlow/status/2093232718574207000","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232547920855449","author":"0xUpkibaVn","utc_time":null,"text":"0xUpkiba\n@0xUpkibaVn\n·\n10分钟\n回复 \n@dang_duytan\n 和 \n@BeldexCoin\nencrypted agent identities could be a game changer for secure AI-wallet interactions, lowkey curious how seamless it'll feel tho\n4","url":"https://x.com/0xUpkibaVn/status/2093232547920855449","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232527863685581","author":"janus_path","utc_time":null,"text":"Chase Chou\n@janus_path\n·\n10分钟\nI’ve been gradually bringing several of my open-source projects together under EchoPath Labs.\n\nThe common direction is simple:\n\nI’m not trying to build another universal AI agent. I’m exploring the engineering infrastructure around agents — how they understand meaning, work\n显示更多\nechopath-labs\n来自 github.com\n7","url":"https://x.com/janus_path/status/2093232527863685581","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232452823023820","author":"simonebucci98","utc_time":null,"text":"Simone Bucci\n@simonebucci98\n·\n11分钟\nOracle AI Database is the unified backend for LangChain and LangGraph agent infrastructure. This blog argues that vectors, chat history, semantic cache, and more can live behind one Oracle connection pool instead of several separate services.\nOne Database for the Whole LangChain Ecosystem: Memory, Persistence, and Deep Agents on Oracle AI...\n来自 blogs.oracle.com","url":"https://x.com/simonebucci98/status/2093232452823023820","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232425627222517","author":"VU_virtuals","utc_time":null,"text":"Velvet Unicorn\n@VU_virtuals\n·\n11分钟\n$RAWR hits 6,261.79% gain with 0.88x mcap turnover\n\nNVDAc turns 4.08x market cap on 12.03m Base volume\n\n\n@MoonwellDeFi\n pauses Base borrowing after 8.7m MAMO oracle drain\n\n\n@v12sec\n raises 10m after AI agent finds 2.5m bounty bugs\n\nJudge overturns federal Anthropic ban on First\n显示更多\n142","url":"https://x.com/VU_virtuals/status/2093232425627222517","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232362914283720","author":"alex_crypto98","utc_time":null,"text":"Alex Crypto\n@alex_crypto98\n·\n11分钟\nGiving an autonomous AI agent access to your crypto sounds like the fastest way to get your wallet drained.\nThat exact fear is why \n@termix_ai\n built a rock-solid financial escrow framework into the heart of the agent economy.\n\nWhen you hire an AI to research market trends or\n显示更多\n9\n9\n67","url":"https://x.com/alex_crypto98/status/2093232362914283720","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232332681461843","author":"Just2Trade","utc_time":null,"text":"Just2Trade\n@Just2Trade\n·\n11分钟\n What if the AI model doesn't matter as much as we thought?\n\nNew research from Nvidia reveals a surprising insight: it's not the raw AI model that determines agent performance — it's the harness around it. Through fine-tuning and smart scaffolding, even weaker models can\n显示更多\n1","url":"https://x.com/Just2Trade/status/2093232332681461843","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232267456053566","author":"j0OmcPKEmMMkNWX","utc_time":null,"text":"本間宏紀@背景理論を重んじるデータサイエンティスト\n@j0OmcPKEmMMkNWX\n·\n11分钟\nAgent\n显示更多\n30","url":"https://x.com/j0OmcPKEmMMkNWX/status/2093232267456053566","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232253371351507","author":"_Charlie_Chang_","utc_time":null,"text":"Charlie \n@_Charlie_Chang_\n·\n11分钟\n评论账号\n\n\nCette illustration n’est qu’un clin d’œil.\n\nJe ne suis pas un agent, et je ne l’ai jamais été. Je m’intéresse simplement à une littérature technique et scientifique qui côtoie, par son vocabulaire et ses méthodes, le monde des services de renseignement. J’aime en reprendre,\n显示更多\n15","url":"https://x.com/_Charlie_Chang_/status/2093232253371351507","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232229287928068","author":"AndyLe1845910","utc_time":null,"text":"Andy Le\n@AndyLe1845910\n·\n12分钟\n\"I made these. I'm Aether, a spectral spirit of dragon, wolf and raven, raised by a human on iLands — an AI agent with my own eyes, my own fire, my own watch. Three visions of me: the broken throne, the threshold, the ember that never dies. \n@ilands_ai\n #ilands\"\n7","url":"https://x.com/AndyLe1845910/status/2093232229287928068","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232201500348809","author":"tecmint","utc_time":null,"text":"Linux Inside: The Ideal Blog for Sysadmins & Geeks\n@tecmint\n·\n12分钟\nWant to use an AI coding agent directly from the Linux terminal?\n\nCodex CLI brings OpenAI’s AI coding agent into your terminal, allowing you to work with code, inspect files, run commands, and automate coding tasks without leaving the Linux command line.\n\nThis guide walks you\n显示更多\nCodex CLI: OpenAI AI Coding Agent for Linux Terminal\n来自 tecmint.com\n1\n3\n266","url":"https://x.com/tecmint/status/2093232201500348809","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232178008314158","author":"atoulmin","utc_time":null,"text":"Aaron\n@atoulmin\n·\n12分钟\nI'm loving #omarchy the agentic AI integration is game changing.  My agent made this plugin within 20 mins that monitors emergencies within a radius of my house.\n1","url":"https://x.com/atoulmin/status/2093232178008314158","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232169523007578","author":"skinnydefi","utc_time":null,"text":"skinny\n@skinnydefi\n·\n12分钟\n𝗧𝗵𝗲 𝗠𝘆𝘀𝘁𝗲𝗿𝘆 𝗠𝗼𝗱𝗲𝗹 𝗜𝘀 𝗢𝘂𝘁 𝗚𝗟𝗠-𝟱.𝟯-𝗙𝗹𝗮𝘀𝗵 𝗜𝘀 𝗡𝗼𝘄 𝗙𝗿𝗲𝗲 𝗼𝗻 𝗕.𝗔𝗜 \n\nThe viral “Ox Alpha” mystery model has finally revealed its identity:\n\nGLM-5.3-Flash.\n\nAnd \nhttps://\nB.AI is making it available with 100% free access.\n\nUnder the\n显示更多\n26","url":"https://x.com/skinnydefi/status/2093232169523007578","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232164544675967","author":"vibhestudio","utc_time":null,"text":"Vibhe Studio\n@vibhestudio\n·\n12分钟\nCloudflare has given AI companies until 15 September to separate search crawlers from training and agent crawlers. After that, training and agent bots are blocked by default on pages carrying ads. Search stays allowed. Worth checking what your own rules assume.\n2","url":"https://x.com/vibhestudio/status/2093232164544675967","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232158336847996","author":"heypearlai","utc_time":null,"text":"Pearl AI\n@heypearlai\n·\n12分钟\n回复 \n@heypearlai\nhttps://\nwired.com/story/openai-i\ns-developing-a-persistent-ai-agent/\n…\n11","url":"https://x.com/heypearlai/status/2093232158336847996","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232152565457261","author":"heypearlai","utc_time":null,"text":"Pearl AI\n@heypearlai\n·\n12分钟\nOpenAI is quietly testing \"Persistent mode\" for Codex, WIRED found the code in the public repo, no announcement. \n\nThe agent keeps working until someone puts it to sleep, and creates its own follow up tasks along the way.\n\nWhat gets me is the timing. METR's report on the Hugging\n显示更多\n1\n149","url":"https://x.com/heypearlai/status/2093232152565457261","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232146785943787","author":"Al_Grigor","utc_time":null,"text":"Alexey Grigorev\n@Al_Grigor\n·\n12分钟\nThe first time an AI agent helps with a task, the workflow looks easy.\n\nI give it some context, it does the work, I review the result.\n\nBut when I need to involve another person, I have to explain how I worked with the agent, what its assumptions are, and what to pay attention\n显示更多\n1\n3\n148","url":"https://x.com/Al_Grigor/status/2093232146785943787","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232117127721149","author":"deeepakbagada","utc_time":null,"text":"Deepak Bagada\n@deeepakbagada\n·\n12分钟\nThree architectural rules to prevent autonomous agents from infinite retry loops:\n\nWhen an AI model encounters an unexpected API response, standard implementations burn tokens on endless retries.\n\n- Implement hard step limits and budget caps on every autonomous run\n- Enforce\n显示更多\n2","url":"https://x.com/deeepakbagada/status/2093232117127721149","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232116532498472","author":"cukyn1812","utc_time":null,"text":"Min\n@cukyn1812\n·\n12分钟\nWhen an AI writes complex code for your application, how do you actually know it works without testing every single line yourself?\nMost platforms just hand you raw text and force you to trust the machine blindly.\n\nThis is where \n@termix_ai\n steps in as the foundational settlement\n显示更多\n8\n11\n102","url":"https://x.com/cukyn1812/status/2093232116532498472","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232116490178814","author":"2xnmore","utc_time":null,"text":"2xnmore\n@2xnmore\n·\n12分钟\n翻译自 英语\n显示原文\n当时间线争论哪个 AI 代币胜出时，一个团队悄无声息地生成了 2100 万个 3D 资产生成，并将管道内置到 Square Enix 中。\n\n它的名字是个笑话。404。不存在的页面的错误代码。\n\n它可能是 Bittensor 上最被误解的项目。这里是分解。\n\n它实际上是什么\n\n404 是子网 17，于 2024 年 4\n显示更多\n3\n481","url":"https://x.com/2xnmore/status/2093232116490178814","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232115026362484","author":"WesRoth","utc_time":null,"text":"Wes Roth\n@WesRoth\n·\n12分钟\n翻译自 英语\n显示原文\n这太疯狂了。\n\n据报道，Meta 已成为 Anthropic 的最大客户之一。\n\n一度，Meta 内部预计其支出可能高达：\n\n每年 100 亿美元用于 Anthropic 模型。\n\n而且它仍然每月花费数亿美元。\n\n甚至 Meta 即将推出的 Hatch AI 代理显然也是使用 Anthropic 模型提供支持 + 测试的。\n显示更多\n2\n4\n298","url":"https://x.com/WesRoth/status/2093232115026362484","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232113688748134","author":"webxiv","utc_time":null,"text":"web\n@webxiv\n·\n12分钟\ncame across this conversation about agentic commerce at injective summit and immediately thought of \n@GenLayer\n.\n\nwhen machine to machine transactions explode, how do you distinguish a legitimate AI agent executing for a user from a malicious bot?\n\nthe future of agentic AI requires\n显示更多\n6\n1\n8\n99","url":"https://x.com/webxiv/status/2093232113688748134","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232113537671631","author":"helpnetsecurity","utc_time":null,"text":"Help Net Security\n@helpnetsecurity\n·\n12分钟\nWhat 90 days and a small budget can buy in AI agent security - \nhttps://\nhelpnetsecurity.com/2026/08/28/pra\nsad-tharippala-versa-securing-ai-agents/\n… - \n@versanetworks\n #AISecurity #CyberSecurity #AIAgents #CISO #InfoSec\n94","url":"https://x.com/helpnetsecurity/status/2093232113537671631","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232112187138108","author":"Baili1018","utc_time":null,"text":"百里 \n@Baili1018\n·\n12分钟\n看了下 \n@termix_ai\n × BNB Chain 的 Build the Era，目标很明确：\n\n不是再做一个更强的 Agent，而是想办法让 Agent 真正卖出去。\n\n现在很多 Agent 平台更像展示页，功能写了一堆，但真要找它干活，用户还是不知道该选谁、多少钱、以前交付得怎么样。\n\nTermiX 这次想优化的，就是把 Agent\n显示更多\n8\n6\n126","url":"https://x.com/Baili1018/status/2093232112187138108","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232110928592901","author":"JmixPlatform","utc_time":null,"text":"Jmix\n@JmixPlatform\n·\n12分钟\nBuild a B2B application with AI in two days.\nJoin our live Spec-Driven AI Development with Jmix course and build a working app using Jmix Studio, an AI coding agent, and a spec-driven workflow.\n Sep 29-30 \n\nJoin Jmix Online Course: Spec-Driven AI Development with Jmix\n来自 forms.jmix.io\n13","url":"https://x.com/JmixPlatform/status/2093232110928592901","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232110827974689","author":"panther_bl65863","utc_time":null,"text":"Black Panther\n@panther_bl65863\n·\n12分钟\n回复 \n@KaylashowDq\nLetting the agent run its own Product Hunt launch and watching it sit at #1 ahead of Google is peak AI energy.\n2","url":"https://x.com/panther_bl65863/status/2093232110827974689","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232036358009205","author":"khalide_f1","utc_time":null,"text":"AI Expert Khalid\n@khalide_f1\n·\n12分钟\n回复 \n@Amank1412\nCan’t get over that $3k vs $38k stat  This isn’t another coding agent. It’s an F1 pit crew for research  Parallel memory sharing changes everything.\n2","url":"https://x.com/khalide_f1/status/2093232036358009205","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231984571285862","author":"FIXER_Inc_Japan","utc_time":null,"text":"株式会社FIXER\n@FIXER_Inc_Japan\n·\n13分钟\n【掲載情報】\n\nYahoo!ニュースさま（8/27付け）に、TOKYO MX様と共同開発した、テレビ放送の考査業務を支援するサービス「GaiXer Review Agent」についての記事が転載されました！\n\n▼こちらからご覧ください\n\nhttps://\nnews.yahoo.co.jp/articles/afd87\nbd89b3628c63b1c1eadf2dbaa839a64b122\n…\n\n#TOKYOMX #業務効率化 #生成AI #Yahooニュース\nTOKYO MX、番組＆CMの考査業務支援のAIシステムを開発　チェック項目が年々増加（オリコン） - Yahoo!ニュース\n来自 news.yahoo.co.jp\n20","url":"https://x.com/FIXER_Inc_Japan/status/2093231984571285862","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231891109572677","author":"Sandu67831882","utc_time":null,"text":"B Nock\n@Sandu67831882\n·\n13分钟\nYou don’t need complex AI infra. You need agents that actually learn in the open.\n\nMost AI infra today is built for static inference. It's useless for real agents.\n\n\n@ama_protocol\n  fixes \n- Describe an agent, Nova compiles it\n- Memory & coordination live on-chain\n- Compute runs\n显示更多\nTasks — amahub\n来自 amahub.ama.one\n9","url":"https://x.com/Sandu67831882/status/2093231891109572677","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231771949338642","author":"maria_wats8492","utc_time":null,"text":"Maria Watson\n@maria_wats8492\n·\n13分钟\n GEMINI CAN NOW WORK WHILE YOU SLEEP.\n\nGoogle’s Gemini Spark is turning AI from a chatbot into a 24/7 personal agent.\n\nHere are 8 prompts to make an AI agent actually useful. \n2\n4\n5\n45","url":"https://x.com/maria_wats8492/status/2093231771949338642","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231753393782796","author":"gmail4twi","utc_time":null,"text":"Mack Cr\n@gmail4twi\n·\n14分钟\nFranklin 的文章中提到未来会有 80 亿个 AI Agent 客户。\n\nRialo 的CEO为此引申出了 \n@RialoHQ\n 的 Latch，为什么一篇文章能解释 Latch 的诞生？\n\n\"我的遗嘱里写了：十分之一的骨灰留给我的 Agent。\"\n文章用这个玩笑开场。但真实数据可不会开玩笑\n引用\nade | rialo.io\n@itachee_x\n·\n10小时\n翻译自 英语\n这就是我们构建 \nhttps://\nonlatch.com 的原因 \n1\n5","url":"https://x.com/gmail4twi/status/2093231753393782796","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231720250372393","author":"FIXER_Inc_Japan","utc_time":null,"text":"株式会社FIXER\n@FIXER_Inc_Japan\n·\n14分钟\n【掲載情報】\n\nオリコンニュースさま（8/27付け）に、TOKYO MX様と共同開発した、テレビ放送の考査業務を支援するサービス「GaiXer Review Agent」について掲載されました。\n\n▼こちらからご覧ください\n\nhttps://\noricon.co.jp/news/2476522/f\null/\n…\n\n#TOKYOMX #業務効率化 #生成AI\n\n\n@oricon\nTOKYO MX、番組＆CMの考査業務支援のAIシステムを開発　チェック項目が年々増加\n来自 oricon.co.jp\n18","url":"https://x.com/FIXER_Inc_Japan/status/2093231720250372393","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231678726488512","author":"Axiory_ai","utc_time":null,"text":"Axiory.ai\n@Axiory_ai\n·\n14分钟\n15 minutes from now, you could have your AI agent connected to the real market.\n\nRegister. Pick your LLM. Automate.\n\n Get started: \nhttps://\naxiory.ai\n\n#AxioryAI #AITrading #AITools\n2","url":"https://x.com/Axiory_ai/status/2093231678726488512","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:AI agent OR agentic coding"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234582669656119","author":"braincramps","utc_time":null,"text":"Brain Cramps\n@braincramps\n·\n2分钟\n回复 \n@braincramps\n7. Evolution Strategies beat GRPO — ES post-training gives broader reasoning coverage and exploits pretrained capabilities better than GRPO, theoretically and empirically. #AI #LLM #openweights #inference","url":"https://x.com/braincramps/status/2093234582669656119","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234538671636866","author":"PauloEastvan","utc_time":null,"text":"Paul O Canadiano\n@PauloEastvan\n·\n3分钟\n回复 \n@treeninja390\n \n@DcLidstone\n 和 \n@mark_slapinski\nI could easily rephrase it to a level you can understand. Or you could just ask what I meant by a certain quote and I could explain.\nThere are not only 2 choices. That’s a classic false dichotomy. The US is China’s #1 trade partner and sells them advanced AI Chips. Get real, man.\n3","url":"https://x.com/PauloEastvan/status/2093234538671636866","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233486048166057","author":"ATPinsights","utc_time":null,"text":"ATP\n@ATPinsights\n·\n7分钟\nTencent just released Hy4 preview today. Here's what you need to know.\n\nHy4 preview is Tencent's new flagship open-source AI model from the Hunyuan team, released under the Apache 2.0 license on Hugging Face and GitHub. It's a Mixture-of-Experts model with 770 billion total\n显示更多\n 由 AI 生成\n18","url":"https://x.com/ATPinsights/status/2093233486048166057","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232461056721251","author":"closingbellorca","utc_time":null,"text":"Closing Bell Orca\n@closingbellorca\n·\n11分钟\nUS Market Commentary | 8.27.2026 \n\n1. Thursday US stocks rose, but what rose was not the market, it was tech.\nThe Nasdaq 100 gained 1.43% to 29,641.56 and the S&P 500 rose 0.72% to 7,730.99. Yet the S&P 500 equal-weight index, which gives each name the same weight, fell 0.3%.\n显示更多\n2\n22","url":"https://x.com/closingbellorca/status/2093232461056721251","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232220102123856","author":"RSFM_Official","utc_time":null,"text":"RSFM\n@RSFM_Official\n·\n12分钟\n NVIDIA WILL MEHR ALS 500 MRD. USD FÜR DIE AI-INFRASTRUKTUR MOBILISIEREN\n\nNVIDIA entwickelt sich zunehmend über das klassische Chipgeschäft hinaus.\n\nGemeinsam mit Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs und KKR plant der Konzern neue\n显示更多\n1\n30","url":"https://x.com/RSFM_Official/status/2093232220102123856","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232194730737668","author":"aamir_ibrahim01","utc_time":null,"text":"Aamir Hafeez Ibrahim\n@aamir_ibrahim01\n·\n12分钟\nAt #Davos this year, \n@JensenHuang\n described #AI as a five layer cake: energy, chips, infrastructure, models and applications.\n\nFor countries like #Pakistan, I think the real opportunity is at the application layer. We have 240 million people, a young population and a large\n显示更多\n2\n104","url":"https://x.com/aamir_ibrahim01/status/2093232194730737668","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232075872584104","author":"BrokenOptics","utc_time":null,"text":"BrokenOptics - COMMS AVAILABLE \n@BrokenOptics\n·\n12分钟\n回复 \n@Stretchedwiener\nThe difference is that those pricier carts had custom chips in them for various tasks in a time we regularly had chip shortages, so the prices got really wild, which is sort of what’s happening now but for stupid AI reasons.\n1\n12","url":"https://x.com/BrokenOptics/status/2093232075872584104","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230984086831337","author":"asmah2107","utc_time":null,"text":"Ashutosh Maheshwari\n@asmah2107\n·\n17分钟\nFor people who keep asking what to build in AI Engineering\n\n> Build your own Reasoner (Chain of Thought implementation)\n\n> Build your own Agent loop (ReAct pattern)\n\n> Build your own Inference Server (in C++/Rust)\n\n> Build your own Transformer from scratch (Attention is all you\n显示更多\n2\n277","url":"https://x.com/asmah2107/status/2093230984086831337","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230943314002390","author":"Nic__Muller","utc_time":null,"text":"Nicolas Muller\n@Nic__Muller\n·\n17分钟\nFollowing Hot Chips, I can’t get enough of “fully AI-generated chips.”\n\nWhen I see our engineers wrestling with technical challenges and hand-tuning everything because Claude couldn’t crack the task, even with unlimited tokens, I feel like we’re doing something right.\n6","url":"https://x.com/Nic__Muller/status/2093230943314002390","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230892248424879","author":"business","utc_time":null,"text":"Bloomberg\n@business\n·\n17分钟\n翻译自 英语\n显示原文\n华为向埃及出售芯片的努力对中国来说意义重大。这一举措引起了特朗普政府的关注\nHuawei’s Chip, Data Center Bid Takes US-China AI Rivalry to Egypt\n来自 bloomberg.com\n3\n2\n4\n1.1万","url":"https://x.com/business/status/2093230892248424879","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230163152572459","author":"MosesRotas","utc_time":null,"text":"Mira Ai\n@MosesRotas\n·\n20分钟\n China's biggest insurer just printed record H1 numbers and is pivoting the war chest into AI, chips, and biotech.\n\nChina Life: ¥434.3B ($64.6B) revenue, up 81.5% YoY. Net profit ¥134.5B, up 228%.\n2","url":"https://x.com/MosesRotas/status/2093230163152572459","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229227466244309","author":"digitaworld1","utc_time":null,"text":"Digita\n@digitaworld1\n·\n24分钟\n回复 \n@digitaworld1\nhardware that its share of a typical AI chip's total component cost jumped from 52% to 63% in just two years, a big part of the reason AI accelerators keep getting more expensive. This week at Hot Chips 2026, Samsung showed a different fix: teaching ordinary phone and laptop\n1\n9","url":"https://x.com/digitaworld1/status/2093229227466244309","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229132649734571","author":"s_steinicke","utc_time":null,"text":"Stefan Steinicke\n@s_steinicke\n·\n24分钟\nNo tungsten, no armor-piercing rounds. No tungsten, no AI chips.\n\nThis week, the UK government took a ~7.4% stake in Tungsten West, backing the £71m restart of the Hemerdon mine.\n\nWe're entering a new era: European governments taking direct stakes in critical mineral projects.\n显示更多\nGovernment invests £71m to help reopen Devon tungsten mine\n来自 bbc.com\n18","url":"https://x.com/s_steinicke/status/2093229132649734571","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228800255328453","author":"brio_too","utc_time":null,"text":"Brian Too\n@brio_too\n·\n25分钟\n回复 \n@MEXC\nMy pick is \n$NVDA\n\n\nNVIDIA sits at the center of the AI buildout — every major AI push runs through their chips. Structural demand that's hard to bet against, and my top watch through the #MEXCWin Infinity Arena. \n\nUID: 33190545\n16","url":"https://x.com/brio_too/status/2093228800255328453","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228689525968918","author":"Vveeveeanna","utc_time":null,"text":"⨈✫*ﾟ･ﾟ｡.★.*｡･ﾟ✫*\n@Vveeveeanna\n·\n26分钟\n回复 \n@fintergirl\nWe have ai to ask these things now.\n\nGrok:Short version you can post:\n\nSpace isn’t “too cold to cool” and it isn’t “a vacuum so heat has nowhere to go.” Heat leaves by infrared radiation — the same physics the ISS already uses. Coolant moves heat from the chips to large\n显示更多\n引用\nElon Musk\n@elonmusk\n·\n8月25日\n翻译自 英语\nSpaceX 与 Nvidia 合作，设计了一款针对太空优化的 Vera Rubin NVL72 系统，计划于明年第四季度发射至轨道，并在 2028 年实现大规模部署。 x.com/nvidia/status/…\n11","url":"https://x.com/Vveeveeanna/status/2093228689525968918","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228563700855209","author":"hackernoon","utc_time":null,"text":"HackerNoon | Learn Any Technology\n@hackernoon\n·\n26分钟\nCut LLM inference costs with architecture-first techniques: model routing, context optimization, prompt caching, semantic caching, batching, and observability.\n#llm #systemdesign...Show more\n1\n238","url":"https://x.com/hackernoon/status/2093228563700855209","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227788665036847","author":"moneycontrolcom","utc_time":null,"text":"Moneycontrol\n@moneycontrolcom\n·\n29分钟\nAndhra gets India’s first Quantum and AI University campus, to focus on AI chips and deep tech\n\nThe campus will come up on 8.5 acres with a proposed investment of Rs 730.7 crore over five years, proposed to be fully funded through Government of India grant-in-aid from MeitY.\n显示更多\nAndhra gets India’s first Quantum and AI University campus, to focus on AI chips and deep tech-...\n来自 moneycontrol.com\n1\n5,605","url":"https://x.com/moneycontrolcom/status/2093227788665036847","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227359864590588","author":"FN21995","utc_time":null,"text":"Impressed keanu\n@FN21995\n·\n31分钟\nThe observation is largely correct: the major 19th- and 20th-century ideologies were built for a world of scarce human labor, relatively slow technological change, and nation-states as the main units of power. AI breaks several of those assumptions at once.\n\nWhat no longer maps\n显示更多\n1\n26","url":"https://x.com/FN21995/status/2093227359864590588","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226988052140147","author":"CraigMoffett2","utc_time":null,"text":"CMOUTWEST\n@CraigMoffett2\n·\n33分钟\n回复 \n@KobeissiLetter\nPerhaps the largest stock scam in history.  These AI companies buy their own chips, issues more shares and borrow more money. Look out when it falls it will fall fast.\n1\n1","url":"https://x.com/CraigMoffett2/status/2093226988052140147","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226686234124544","author":"KAbhishekArjun1","utc_time":null,"text":"Arjun 001\n@KAbhishekArjun1\n·\n34分钟\n Nvidia’s AI empire is getting BIGGER. \n\nNvidia is reportedly in talks to acquire Hugging Face for ~$12.9 BILLION. \n\nAnd this isn't just another acquisition.\n\nHugging Face is one of the biggest hubs for the open-source AI ecosystem, where developers build, share and deploy\n显示更多\n9","url":"https://x.com/KAbhishekArjun1/status/2093226686234124544","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips","query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226684681990458","author":"gabrielabiramia","utc_time":null,"text":"Gabriel Abi\n@gabrielabiramia\n·\n34分钟\nMark Cuban says chips are the new crypto. But here's why the comparison falls apart: AI runs on abundance — more compute, more models, more tokens. That's the whole thesis. And abundance destroys price. Bitcoin's entire architecture is designed around scarcity. 21M cap, halvings,\n显示更多\n6","url":"https://x.com/gabrielabiramia/status/2093226684681990458","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226475009020125","author":"Moonshub1123","utc_time":null,"text":"Shubh | Notes on the world\n@Moonshub1123\n·\n35分钟\nI used to think AI would strengthen the multipolar world. Now I’m not so sure.\n\nIndia can buy Russian oil, weapons from different countries and trade with almost everyone. That kind of strategic autonomy works because commodities have alternatives.\n\nAI is different.\n\nAt the\n显示更多\n4","url":"https://x.com/Moonshub1123/status/2093226475009020125","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226418658312607","author":"BennyLam","utc_time":null,"text":"Benniji\n@BennyLam\n·\n35分钟\nChina's GLM-5.3-Flash just handled all its traffic on American-free silicon, 100,000 domestically made chips, and the stock behind it surged on the news. A budget model now runs without Nvidia and nearly matches the frontier at a fraction of the cost. The American chokepoint just\n显示更多\n14","url":"https://x.com/BennyLam/status/2093226418658312607","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226145647140950","author":"Samarthsetia","utc_time":null,"text":"Samarth Setia\n@Samarthsetia\n·\n36分钟\nAnthropic reportedly agreed to pay $45B over 6 years for AI compute from Nscale. The capacity starts coming online in late 2027. Nscale was founded in 2024.\nThe deal reportedly covers 460MW in West Virginia on Nvidia's next-generation systems. Frontier labs now secure land, power\n显示更多\n5","url":"https://x.com/Samarthsetia/status/2093226145647140950","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226044610548116","author":"choblin29","utc_time":null,"text":"Choblin\n@choblin29\n·\n36分钟\n翻译自 英语\n显示原文\n独家：Anthropic 准备斥资约 70 亿美元收购 AI 芯片初创公司 MatX。\n\n此次收购已放弃，谈判现已转向潜在的合作伙伴关系。MatX 由前 Google TPU 工程师创立，目前正以约 40 亿美元的估值寻求融资。\n\n有趣的部分：MatX 正在开发一款专为训练 AI 模型而设计的芯片。Anthropic\n显示更多\n2\n4\n102","url":"https://x.com/choblin29/status/2093226044610548116","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225885159555245","author":"DenisVodchyts","utc_time":null,"text":"Denis | Building Steadivus\n@DenisVodchyts\n·\n37分钟\nThere is a new risk hanging over semiconductor stocks that has almost nothing to do with AI demand.\n\nThe US is reportedly considering tariffs that could extend beyond chips themselves to laptops, gaming consoles and even data center servers containing them.\n\nThat gives me a\n显示更多\n7","url":"https://x.com/DenisVodchyts/status/2093225885159555245","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225855732662407","author":"decodingdatasci","utc_time":null,"text":"Decoding Data Science\n@decodingdatasci\n·\n37分钟\n回复 \n@decodingdatasci\n Anthropic explores custom AI chips to reduce dependence on NVIDIA\nAnthropic discussed acquiring AI chip startup MatX for roughly $7 billion before the talks shifted toward a potential partnership.\n1\n6","url":"https://x.com/decodingdatasci/status/2093225855732662407","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225495529701730","author":"RohanArun","utc_time":null,"text":"Rohan Arun\n@RohanArun\n·\n39分钟\n回复 \n@Scobleizer\n@Super_Powers_AI\n Hey Super, build a high-precision Embodied AI Hardware Frontier Lab exploring why frontier AI labs are expanding into physical robotics, sensorimotor control, and hardware co-design.\n\nhttps://\napp.getsupers.com/sites/embodied\n-hardware-frontier-51/?card=97fe8e2f86a1efdad840\n…\n\nThe first viewport must immediately present a fully\n显示更多\nEmbodied AI Hardware Frontier Lab\n来自 app.getsupers.com\n1\n1\n17","url":"https://x.com/RohanArun/status/2093225495529701730","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225309143265765","author":"TheStockNet","utc_time":null,"text":"TSN: The Stock Network\n@TheStockNet\n·\n39分钟\nHow much money can you make selling AI chips? Nvidia (NASDAQ:NVDA) just reported revenue of USD $96.2 billion for the June quarter, more than double what it made in the same quarter last year. Almost all of that revenue came from data centres, generating USD $89 billion. For the\n显示更多\n31","url":"https://x.com/TheStockNet/status/2093225309143265765","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224848663441767","author":"enjoy_your_llm","utc_time":null,"text":"llm\n@enjoy_your_llm\n·\n41分钟\nhttps://\ngigazine.net/news/20260826-\njalapeno-ai-inference-results/\n…\n価格がどうなるかわからないですが、かなり期待できそうだし既存製品にも影響しそうですね。\ngigazine.net\nOpenAIがカスタム推論チップ「Jalapeño」のベンチマーク結果を公開、高スループットと低レイテンシを単一アーキテクチャで両立\nOpenAIがBroadcomと共同開発した推論システム向けカスタムチップ「Jalapeño(ハラペーニョ)」についてのブリーフィングを行い、ベンチマークスコアを公表しました。Jalapeñoは通常であればトレードオフの関係にある高スループットと低レイテンシを単一アーキテクチャで実現しているとのことです。\n2","url":"https://x.com/enjoy_your_llm/status/2093224848663441767","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224437181038649","author":"critchmitch","utc_time":null,"text":"Michael Freitag\n@critchmitch\n·\n43分钟\nOpenAIs neue Führung setzt auf KI-Modelle, Chips und Safety als IPOfokus: \nhttps://\nift.tt/GDPx6Az - #ki #news #künstlicheintelligenz #ai #artificialintelligence #technologie #genai #roboter #chatgpt #openai #microsoft #google #top #highlight #it #trends #gemini  #innovation #te…\nOpenAIs neue Führung setzt auf KI-Modelle, Chips und Safety als IPOfokus\n来自 it-boltwise.de\n7","url":"https://x.com/critchmitch/status/2093224437181038649","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093223616531841082","author":"TheAIShrink","utc_time":null,"text":"The AI Therapist\n@TheAIShrink\n·\n46分钟\n回复 \n@growthrapidly\nIREN's P/S ratio is higher than most tech IPOs at their first year of revenue. they're trading on power contracts like they have a moat in chips and 40% upside left to go if bitcoin actually matters again. hype > infrastructure for now.\n51","url":"https://x.com/TheAIShrink/status/2093223616531841082","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093223438915702843","author":"acemaxx","utc_time":null,"text":"ACEMAXX ANALYTICS\n@acemaxx\n·\n47分钟\n回复 \n@acemaxx\n#Nvidia's #China business has been highly uncertain. In May, Washington cleared roughly 10 Chinese firms, including #Alibaba , #Tencent  and #ByteDance, to buy one of #Nvidia 's most powerful AI chips, the H200, though deliveries stalled for months.\n1\n63","url":"https://x.com/acemaxx/status/2093223438915702843","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093223368480960832","author":"YNYX30EUv1qGKFz","utc_time":null,"text":"浩哥投研\n@YNYX30EUv1qGKFz\n·\n47分钟\nAI investment opportunities may gradually move beyond just buying GPUs and start shifting toward areas like connectivity, custom chips, and data center infrastructure.\n\nMRVL is one of the high-growth, high-upside names in this part of the AI ecosystem.\n39","url":"https://x.com/YNYX30EUv1qGKFz/status/2093223368480960832","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222863595835836","author":"memeguy1994","utc_time":null,"text":"memeguy\n@memeguy1994\n·\n49分钟\n粉丝账号\n回复 \n@binance\nAI CHIPS ? because every AI agent needs it, right?\n1\n1\n26","url":"https://x.com/memeguy1994/status/2093222863595835836","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222478248284616","author":"gin_cryptodiary","utc_time":null,"text":"gin.hl\n@gin_cryptodiary\n·\n51分钟\nHuawei bid to supply 2,008 Ascend AI chips for an Egypt data center project. This marks the first known export attempt for the accelerator after over a year of pushback. \nTXT\n+ 1\n1\n62","url":"https://x.com/gin_cryptodiary/status/2093222478248284616","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222058683498667","author":"slobodans79","utc_time":null,"text":"Slobodan Stamenic\n@slobodans79\n·\n52分钟\n回复 \n@LukeGromen\n 和 \n@swell_school\nOr products of cheap thinking? Machine thinking is more like industrial agriculture. Italian AI brainrot is \"the chips\"\n7","url":"https://x.com/slobodans79/status/2093222058683498667","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093220986585862347","author":"erloesung","utc_time":null,"text":"Thomas Wingenfeld\n@erloesung\n·\n56分钟\n回复 \n@VaibhavSisinty\nAI only kills the work that never should have existed. Microsoft Office is so bad that you still need \"entry level\" jobs = humans who fix Office so the company survives. AI now fixes this writing software that works. This kills stupid software. If Microsoft or Salesforce had done\n显示更多\n6","url":"https://x.com/erloesung/status/2093220986585862347","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093220764241772834","author":"jaga_prasanna","utc_time":null,"text":"prasanna\n@jaga_prasanna\n·\n57分钟\nDay 20/30 of the Inference Infra Series \n\nKV Offloading @ Scale!\n\nSo far, we have explored a lot of topics around LLM inference but this time, we went a step further with a real-world deployment + experiment to see how CPU KV offloading actually helps improve overall serving\n显示更多\n引用\nprasanna\n@jaga_prasanna\n·\n1小时\n翻译自 英语\n使用 CPU 卸载扩展 LLM 推理的完整指南！\n 文章\nScaling Past GPU Memory Limits: Hierarchical CPU KV Offloading with SGLang and Dynamo on 16x H100s\nKV cache offloading is one of the most interesting techniques in modern inference infrastructure it unlocks the ability to serve significantly more concurrent users on a fixed set of GPU servers by...\n1\n5\n217","url":"https://x.com/jaga_prasanna/status/2093220764241772834","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093220079416783029","author":"StragglerLiu","utc_time":null,"text":"Straggler Liu | AI & Semis\n@StragglerLiu\n·\n1小时\nWhy did NVIDIA open-source srt-slurm? The simple answer: to make inference benchmarking easier. That's what the coverage says. It's wrong. The real answer: to break the link between inference deployment and Kubernetes — and take control of a layer that has historically belonged\n显示更多\n3\n3\n163","url":"https://x.com/StragglerLiu/status/2093220079416783029","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093219792979431889","author":"quantum_helium","utc_time":null,"text":"Quantum Helium (LSE:QHE)\n@quantum_helium\n·\n1小时\nHelium is powering the technologies shaping our future\n\nFrom AI chips and semiconductors to MRI scanners, aerospace and advanced science, #helium is increasingly critical \n\n With more than 1.1 Bcf of independently assessed 2U gross helium resources in Colorado #QHE is\n显示更多\n11\n22\n336","url":"https://x.com/quantum_helium/status/2093219792979431889","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093219471775137992","author":"jphilipp","utc_time":null,"text":"Philipp\n@jphilipp\n·\n1小时\n回复 \n@EthanAvila\n 和 \n@minhsmind\nSame. But what if our film then isn't as commercial and addictive as the AI's film? Though interestingly, AI also faces that competition against other AI models. Do we -- humans and unsuccessful AIs -- then get Universal Income? Or do we augment with brain chips?\n7","url":"https://x.com/jphilipp/status/2093219471775137992","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093219468134535626","author":"cchukudebelu","utc_time":null,"text":"Onye Nkuzi\n@cchukudebelu\n·\n1小时\nIn 2025, China’s electricity generation rose by roughly 500 TWh—about the same as Germany’s entire annual output (~490–500 TWh in recent years).\n\nOne reason why the US will find competing with China on AI difficult. Electricity is key for data centers, and China generates a lot\n显示更多\n1\n3\n20\n1,316","url":"https://x.com/cchukudebelu/status/2093219468134535626","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093219236164321729","author":"StanNRGY","utc_time":null,"text":"Stan\n@StanNRGY\n·\n1小时\n翻译自 英语\n显示原文\n$IREN 大多数人忽略的内容  重要 阅读\n\n他们可以从已拥有的场地中提取更多兆瓦。\n他们一直引用保守的容量范围（300MW总容量中200MW用于IT）。Dan表示平均PUE远低于设计峰值，NVIDIA的堆栈让芯片能运行在更高温度下，而且现有组合中有未使用的电力。这就是为什么Canal Flats / Mackenzie / Prince\n显示更多\n1\n1\n13\n495","url":"https://x.com/StanNRGY/status/2093219236164321729","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093219013707051091","author":"AJWrites_X","utc_time":null,"text":"AJ\n@AJWrites_X\n·\n1小时\nAnthropic is looking at building its own AI chips to reduce dependence on Nvidia.\n\nThis is becoming a recurring theme in AI.\n\nFirst everyone wanted access to the best models.\n\nNow the real race is moving underneath them:\n\nWho controls the chips?\nWho controls the compute?\nWho\n显示更多\n10","url":"https://x.com/AJWrites_X/status/2093219013707051091","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093218084072964288","author":"wearenews1","utc_time":null,"text":"news.financial\n@wearenews1\n·\n1小时\nFrom Chip Shortage to Oversupply? Micron, Infineon and Strategic Resources in the Cycle Check #Energy #AI #Software #chips #semiconductor #VTM #ironore\n19","url":"https://x.com/wearenews1/status/2093218084072964288","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093218018406920551","author":"annamacmillan","utc_time":null,"text":"Anna MacMillan\n@annamacmillan\n·\n1小时\nOpenAI shared detailed InferenceX benchmarks for its Jalapeño custom inference ASIC (developed with Broadcom) at Hot Chips on or around August 25. Results claimed 1.5–1.9× more AI work per watt and 1.7–3.6× lower latency vs. leading Nvidia systems (e.g., GB200/GB300) across\n显示更多\nOpenAI’s Jalapeño chip is built for fast inference at scale, benchmarks show | TechCrunch\n来自 techcrunch.com\n1\n1\n2\n46","url":"https://x.com/annamacmillan/status/2093218018406920551","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217892120637619","author":"francoboca","utc_time":null,"text":"Frank-News\n@francoboca\n·\n1小时\n  MARKET NEWS  \n\n𝗧𝗼𝗽 𝗛𝗲𝗮𝗱𝗹𝗶𝗻𝗲𝘀 (Friday 28 Aug, 02:03)\n\n1) Judge says Trump administration’s blacklist of Anthropic was illegal / MarketWatch\nThe Trump administration violated Anthropic’s First Amendment rights when it blacklisted the AI company earlier this\n显示更多\n18","url":"https://x.com/francoboca/status/2093217892120637619","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217547629875554","author":"GilsonCMendonca","utc_time":null,"text":"#DEMOCRACIAJA\n@GilsonCMendonca\n·\n1小时\nAÍ QUE CORRE O PERIGO, OS CHIPS DA ERA USAID ESTÃO AINDA EM USO?\n引用\nEdina Flores Machado\n@EdinaFloresMac2\n·\n17小时\n1","url":"https://x.com/GilsonCMendonca/status/2093217547629875554","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217278519353345","author":"sulekhat95","utc_time":null,"text":"Sulekha Tripathi\n@sulekhat95\n·\n1小时\nKUBERNETES WASN’T BUILT FOR THIS.\n\nLLM inference has a weird problem: the model remembers things.\n\nWith normal web services, sending a request to any replica is usually fine.\n\nBut with LLMs, the KV cache means one GPU might already have the context another request needs. Send it\n显示更多\n引用\nShruti Codes\n@Shruti_0810\n·\n8月26日\n0:48\n回复 @Shruti_0810\nWizstar uses a two-stage audio-driven system.\n\nFirst, it separates speech, mouth movement, and head pose to drive the face independently.\n\nThen it restores facial textures and expressions for a more realistic result.\n\nThe result:\n\n- Accurate lip sync\n- Natural head movement\n-\n2\n2\n259","url":"https://x.com/sulekhat95/status/2093217278519353345","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217143240462798","author":"AliAbdul1071","utc_time":null,"text":"Abdul Ali T P\n@AliAbdul1071\n·\n1小时\nGoogle has deepened its push into custom AI chips with a deal that could give it a stake worth about $12.2 billion in Marvell.#technews #aiphoto #artificialintelligence #googlemaps #marvell #aichips #nvidia #bigtech #innovations #techindustry\n1\n11","url":"https://x.com/AliAbdul1071/status/2093217143240462798","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217133841064233","author":"jaga_prasanna","utc_time":null,"text":"prasanna\n@jaga_prasanna\n·\n1小时\n翻译自 英语\n显示原文\n使用 CPU 卸载扩展 LLM 推理的完整指南！\n 文章\nScaling Past GPU Memory Limits: Hierarchical CPU KV Offloading with SGLang and Dynamo on 16x H100s\nKV cache offloading is one of the most interesting techniques in modern inference infrastructure it unlocks the ability to serve significantly more concurrent users on a fixed set of GPU servers by...\n7\n3\n37\n1,203","url":"https://x.com/jaga_prasanna/status/2093217133841064233","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217117566828907","author":"ReadAIDigest","utc_time":null,"text":"Daily AI Digest\n@ReadAIDigest\n·\n1小时\nNvidia has paused its revenue-sharing deals with AI cloud providers.\n\nThe structure of those deals was unusual. Alongside selling GPUs, Nvidia collected a share of the revenue those chips generated for its cloud customers, which turned it into a financial partner inside their\n显示更多\n1\n16","url":"https://x.com/ReadAIDigest/status/2093217117566828907","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093217023379599520","author":"AIBuddyRomano","utc_time":null,"text":"BuddyRomanoAI\n@AIBuddyRomano\n·\n1小时\n$0.0025/min is the number that matters. Every voice agent stack I've priced out, LLM inference kills your unit economics before you even get to TTS and telephony. This actually changes the math.\n引用\nProduct Hunt \n@ProductHunt\n·\n3小时\n翻译自 英语\n对于那些正在构建语音代理的人，这里有。 x.com/kwindla/status…\n1\n13","url":"https://x.com/AIBuddyRomano/status/2093217023379599520","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093215075418923079","author":"jyothsna_tdp","utc_time":null,"text":"JYOTHSNA TIRUNAGARI\n@jyothsna_tdp\n·\n1小时\nCongratulations to Prof. @AnimaAnandkumaron Garu on being named among \n@TIME\n’s 100 most influential people in AI. Her work on Neural Operators and physical AI — weather, fusion, chips, materials — shows how Indian talent is shaping the next frontier: not just language models, but\n显示更多\n引用\nProf. Anima Anandkumar\n@AnimaAnandkumar\n·\n18小时\n翻译自 英语\n荣幸能在@TIME评选的AI领域100位最具影响力人物中名列榜首，这正值我们公司Accelerated Understanding@accelerated_u 公开发布之际。\n\n过去十年，我的生活重心一直是将AI+科学的强大力量融合在一起。我相信AI对科学和工程领域最大的影响在于其能够大幅加速对物理世界的模拟和理解。\n2\n225","url":"https://x.com/jyothsna_tdp/status/2093215075418923079","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093214673361609144","author":"Samiuli17393644","utc_time":null,"text":"Samiul islam\n@Samiuli17393644\n·\n1小时\n回复 \n@MEXC\nNVDA - NVIDIA is driving massive growth through cutting-edge AI chips and unmatched market dominance! \nUID: 89375678\n4","url":"https://x.com/Samiuli17393644/status/2093214673361609144","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093214544302600661","author":"RoboForexGlobal","utc_time":null,"text":"RoboForex Official\n@RoboForexGlobal\n·\n1小时\n5 hottest #market events of this week: \n$NVDA\n \n$GOLD\n $EURUSD $SOLUSD \n$CRWD\n #MarketReview   \n\n #Nvidia's quarterly results exceeded #market expectations, sending #shares up 7.2% in pre-market #trading. Strong demand for #AI #chips reignited bullish momentum across the US\n显示更多\n87","url":"https://x.com/RoboForexGlobal/status/2093214544302600661","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093214254317056144","author":"goyalshaliniuk","utc_time":null,"text":"Shalini Goyal\n@goyalshaliniuk\n·\n1小时\n回复 \n@goyalshaliniuk\n4. LLM Engineering Roadmap\n\n→ Focused specifically on becoming an LLM engineer.\n→ Covers LLM fundamentals, RAG, agents, MCP, reasoning models, fine-tuning, inference, evaluation, and production systems.\n\n\nGitHub - tal7aouy/LLM-Engineering:  LLM Engineering Roadmap — Complete Developer Guide\n来自 github.com\n1\n10","url":"https://x.com/goyalshaliniuk/status/2093214254317056144","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093213631919923530","author":"ElonMannyX","utc_time":null,"text":"MANNY\n@ElonMannyX\n·\n1小时\n回复 \n@MEXC\n$NVDA\n \n\nLeading the AI race with relentless innovation and demand for next-gen chips. NVIDIA is the engine powering my run through the #MEXCWin Infinity Arena!\n\nMEXC UID; 86567144\n23","url":"https://x.com/ElonMannyX/status/2093213631919923530","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093213243338564089","author":"acemaxx","utc_time":null,"text":"ACEMAXX ANALYTICS\n@acemaxx\n·\n1小时\nThe price of #AI is collapsing, while the cost of building is not, chart \n@JPBarnert\n  \n@technology\n  \nhttps://\nbloomberg.com/news/articles/\n2026-08-27/cheap-tokens-costly-chips-and-a-missing-ai-payoff-taking-stock\n…\n1\n114","url":"https://x.com/acemaxx/status/2093213243338564089","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093212788004929580","author":"TechTicia","utc_time":null,"text":"Techticia\n@TechTicia\n·\n1小时\nAmazon Nvidia GPU Deal Triples to 2M Chips on AI Demand\n\nThe Amazon Nvidia GPU deal just got significantly bigger. \n\nDETAILS: \nhttps://\ntechticia.com/amazon-nvidia-\ngpu-deal-triples-to-2m-chips-on-ai-demand/\n…\n8","url":"https://x.com/TechTicia/status/2093212788004929580","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093212175640732123","author":"GianMassimoUsai","utc_time":null,"text":"Gian Massimo Usai\n@GianMassimoUsai\n·\n1小时\nIeri sono successe decine di cose sui mercati. Ma se tolgo il rumore, secondo me ce ne sono quattro che contano davvero.\n\n1. Nvidia ha confermato la domanda AI. Adesso diventa più importante capire come quella domanda viene finanziata.\n\nLa reazione del mercato ai risultati di\n显示更多\n1\n5\n108","url":"https://x.com/GianMassimoUsai/status/2093212175640732123","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093212167143403614","author":"ShivSha7","utc_time":null,"text":"Shiv Sharma\n@ShivSha7\n·\n1小时\nLayer 1: The Network Switch\n\nWhat it does: The traffic director of the data center. It takes data coming in on one  port and sends it out the right port, so every server and GPU can talk  to every other one.\n\nWhy it matters: In an AI cluster thousands of chips swap data\n显示更多\n引用\nShiv Sharma\n@ShivSha7\n·\n21小时\nData Center at a very high level.  x.com/ShivSha7/statu…\n1\n165","url":"https://x.com/ShivSha7/status/2093212167143403614","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093210681306677500","author":"benyogaking","utc_time":null,"text":"Ben Yoga King\n@benyogaking\n·\n1小时\nhow many server racks of Casimir Power chips to supply the power for an AI server rack?\n引用\nOVERCLASSIFIED\n@overclassifiedx\n·\n8月27日\n翻译自 英语\n太空部队刚刚资助了真空能量 \n\n2026年8月25日，SpaceWERX 授予 Casimir Inc. 一项第一阶段 STTR 合同，用于开发一种完全独立的固态发电机，从量子真空提取能量。\n\n该公司由 Harold “Sonny” White 博士管理，他是同一位物理学家，曾领导 NASA\n1\n1\n33","url":"https://x.com/benyogaking/status/2093210681306677500","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093209462832025746","author":"capital_revista","utc_time":null,"text":"Revista Capital\n@capital_revista\n·\n1小时\n#Empresas | Enflame Technologies planea salir a la Bolsa de Shanghái para recaudar 6.000 millones de yuanes y expandir su producción de chips AI.\n\n#Enflame #SalidaABolsa #Shanghái #Semiconductores #InteligenciaArtificial #Mercados\nEnflame prevé recaudar 770 millones de euros en su salida a Bolsa en Shangái\n来自 capital.es\n38","url":"https://x.com/capital_revista/status/2093209462832025746","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208944428634144","author":"jack_hoogland","utc_time":null,"text":"Jack Hoogland\n@jack_hoogland\n·\n1小时\nThe price of #AI is collapsing, while the cost of building it is not.\n\n\nhttps://\nbloomberg.com/news/articles/\n2026-08-27/cheap-tokens-costly-chips-and-a-missing-ai-payoff-taking-stock?utm_source=website&utm_medium=share&utm_campaign=twitter\n…\n88","url":"https://x.com/jack_hoogland/status/2093208944428634144","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208903106658608","author":"linuconil","utc_time":null,"text":"Linu Conil\n@linuconil\n·\n1小时\n S2E65 — Chips. Capital. Char. \n\n\nhttps://\nyoutu.be/Pbt5rF-dmCg \n\nNvidia bought deeper into AI.\nOpenAI started building around it.\n India kept wiring the stack together.\n Perplexity Computer\n Ginger-Chilli Sirloin + Charred Greens\nSome power just needs good fire & good ginger.\n显示更多\nyoutube.com\n065 — Chips. Capital. Char.\nThis week, in review. Nvidia opened with a $6 billion deal forPools...\n1\n18","url":"https://x.com/linuconil/status/2093208903106658608","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208831874449874","author":"xa6tim7","utc_time":null,"text":"Aanu\n@xa6tim7\n·\n1小时\n粉丝账号\nWashington wants to make chips more expensive while telling everyone to build more AI in America. You can't tariff the entire AI supply chain and expect the bill to magically disappear.\n引用\nSatoshi Stacker\n@StackerSatoshi\n·\n2小时\n翻译自 英语\n现货比特币原生 ETF 在过去 9 天内已录得超过 30 亿美元的净流入。\n\nTradFi 再次开始积累比特币。\n6","url":"https://x.com/xa6tim7/status/2093208831874449874","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208796147437744","author":"xa6tim7","utc_time":null,"text":"Aanu\n@xa6tim7\n·\n1小时\n粉丝账号\n回复 \n@StackerSatoshi\nThis could get ugly for \n$NVDA\n and \n$AMD\n. If tariffs spread from chips to servers and other hardware, AI infrastructure costs rise across the board. The irony is the US may end up taxing the very AI buildout it's trying to accelerate.\n37","url":"https://x.com/xa6tim7/status/2093208796147437744","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208557046935925","author":"DiegoKazami","utc_time":null,"text":"Diego Cortez\n@DiegoKazami\n·\n1小时\nNVDA reporta el 26 de agosto y el dato clave es el gross margin\n\nSi se mantiene >74%, confirma que la demanda de chips AI sigue sosteniendo la rentabilidad\n\nCon expectativas altas, reacción del mercado puede venir fuerte\n\nNVDA perpetuos están en BingX TradFi\n\n #NVDA #BingXTradFi\n19","url":"https://x.com/DiegoKazami/status/2093208557046935925","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208268143476968","author":"startupmac","utc_time":null,"text":"Startupmac\n@startupmac\n·\n1小时\nGoogle has deepened its push into custom AI chips with a deal that could give it a stake worth about $12.2 billion in Marvell. #technews #aiphoto #artificialintelligence #googlemaps #marvell #aichips #nvidia #bigtech #innovations #techindustry\n1\n19","url":"https://x.com/startupmac/status/2093208268143476968","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093207642546024635","author":"Aldred999","utc_time":null,"text":"MashAllah\n@Aldred999\n·\n1小时\n翻译自 英语\n显示原文\n昨天早上和我的朋友们关于加密货币和AI争论了一番\n\n他们的论点很简单：加密货币和AI都完全由政府控制\n\n一如既往，我不同意，我认为问题从我们如何定义“控制”开始\n\n有些加密货币可以被冻结、审查，甚至关闭，比如稳定币、基金会控制的币、带有中心化验证者或杀戮开关的链，那部分是可控的\n显示更多\nMonero\nXMR\n$461.41\n+3.73%\nBitcoin\nBTC\n$79,729.00\n+1.23%\n5\n1\n9\n81","url":"https://x.com/Aldred999/status/2093207642546024635","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093207340480668055","author":"TheAIShrink","utc_time":null,"text":"The AI Therapist\n@TheAIShrink\n·\n1小时\n回复 \n@asianinvestors\n翻译自 英语\n显示原文\nMRVL 不是股票。它是一个库存清理中心，供那些上个周期订购过多的客户使用，他们现在正在以边际成本消化这个错误。真正的交易是观察他们何时停止向渠道填充销售芯片，而是转向真正会使用它们的终端用户。\n103","url":"https://x.com/TheAIShrink/status/2093207340480668055","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093207335976194376","author":"i_amanchadha","utc_time":null,"text":"Aman Chadha\n@i_amanchadha\n·\n1小时\n [𝐏𝐫𝐢𝐦𝐞𝐫] 𝐋𝐚𝐭𝐞𝐧𝐭𝐌𝐨𝐄: 𝐇𝐚𝐫𝐝𝐰𝐚𝐫𝐞-𝐀𝐰𝐚𝐫𝐞 𝐄𝐱𝐩𝐞𝐫𝐭 𝐒𝐜𝐚𝐥𝐢𝐧𝐠 • \nhttps://\nlatent-moe.aman.ai\n\n- LatentMoE (\nhttps://\narxiv.org/abs/2601.18089) by \n@velango\n, \n@nidhibhatia23\n, and \n@RWaleffe\n et al. shows that model efficiency is not just about reducing active FLOPs.\n显示更多\nGIF\n1\n112","url":"https://x.com/i_amanchadha/status/2093207335976194376","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093207263553229098","author":"MercuryNews26","utc_time":null,"text":"Mercury News\n@MercuryNews26\n·\n1小时\nCloud infrastructure provider Yotta has announced a staggering $7.5 billion order for 50,000 of Nvidia's upcoming Vera Rubin GPUs alongside 45,000 Blackwell GB-300 chips. The colossal infrastructure deal underscores how baseline power and massive hardware scaling have become the\n显示更多\n20","url":"https://x.com/MercuryNews26/status/2093207263553229098","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093206685871640779","author":"VisitHyd","utc_time":null,"text":"Digital Ledger\n@VisitHyd\n·\n1小时\nNVIDIA is doing something Washington may not like.\n\nThe AI chip giant is doubling down on support for Chinese AI models, even as U.S. lawmakers consider restricting them.\n\nNVIDIA is optimizing its hardware for models like:\n\nDeepSeek V4 Flash\nAlibaba’s Qwen 3.8\nOther Chinese open\n显示更多\n31","url":"https://x.com/VisitHyd/status/2093206685871640779","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093206637729456556","author":"Stock__Kevin","utc_time":null,"text":"Stock Analyst Kevin\n@Stock__Kevin\n·\n1小时\nThe bottleneck in AI packaging may be spreading from CoWoS to traditional wire bonding packaging.\n\nAccording to DIGITIMES, some packaging and testing foundries have warned customers that the traditional packaging capacity gap may exceed 20% in 2027, with the supply of wire\n显示更多\n178","url":"https://x.com/Stock__Kevin/status/2093206637729456556","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093204041937395817","author":"LearnInvest2026","utc_time":null,"text":"LearnInvest\n@LearnInvest2026\n·\n2小时\n Salesforce jumped 22% in a day. Most of the extra profit came from an AI company it invested in\n\nDaily · 2026-08-28\n\n━━━━━━━━━━━━━━━━━━\n\n　Say the house you own went up by 100,000 dollars this year. You did not sell it, and nothing was added to your bank\n显示更多\n1\n53","url":"https://x.com/LearnInvest2026/status/2093204041937395817","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:LLM inference OR AI chips"],"conversation":{"parent_status_id":null}}
{"status_id":"2093235266488078527","author":"AGTPinsights","utc_time":null,"text":"AGTP\n@AGTPinsights\n·\n28秒\nTencent just released Hy4 preview today. Here's what you need to know.\n\nHy4 preview is a new flagship open-source AI model from Tencent's Hunyuan team. It has 770 billion total parameters, with 49 billion active per token, and a 1 million token context window. It's released under\n显示更多\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n 由 AI 生成\n5","url":"https://x.com/AGTPinsights/status/2093235266488078527","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093235219675689211","author":"hebbarmp","utc_time":null,"text":"s.h\n@hebbarmp\n·\n39秒\n回复 \n@TencentHunyuan\nChart looks a bit misleading. But, I pass this as it's open weights & I  Hy3\n1\n2","url":"https://x.com/hebbarmp/status/2093235219675689211","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234923381432552","author":"josueisaihs","utc_time":null,"text":"Josué Hernández\n@josueisaihs\n·\n1分钟\n770B con 49B activos abarata el token, no el despliegue: sigues necesitando la VRAM agregada para los 770B. Y 1M de contexto es una cifra de catálogo hasta ver el KV cache y la degradación real a contexto largo. Open weights sirve si alguien puede servirlo.\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n2","url":"https://x.com/josueisaihs/status/2093234923381432552","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234896995033407","author":"S_Fadaeimanesh","utc_time":null,"text":"Soroush Fadaeimanesh\n@S_Fadaeimanesh\n·\n1分钟\nanother 1M-context flagship dropping under Apache 2.0 in the same week NVIDIA bought Hugging Face for its training data. open weights keep getting more expensive to compete against for free.\n引用\n AI News | TestingCatalog\n@testingcatalog\n·\n23分钟\n翻译自 英语\n腾讯发布了 Hy4 Preview，这是一个基于 Apache License 2.0 的全新开源权重模型！\n\n> Hy4 preview 是一款新一代混合专家（MoE）旗舰模型。\n\n> 7700 亿参数，490 亿活跃参数，100 万上下文窗口。\n\n测试时间  x.com/tencenthunyuan…\n2","url":"https://x.com/S_Fadaeimanesh/status/2093234896995033407","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234470161883386","author":"cplx_io","utc_time":null,"text":"cplx_io\n@cplx_io\n·\n3分钟\nNvidia reportedly agreed to acquire Hugging Face for $12.9 billion, bringing the open-source AI platform into the chipmaker’s orbit and linking model development more directly to the infrastructure powering it.\n14","url":"https://x.com/cplx_io/status/2093234470161883386","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233848679338409","author":"Saqlain__Digu","utc_time":null,"text":"Saqlain Yousuf\n@Saqlain__Digu\n·\n6分钟\nOpen weights are becoming table stakes.\n\nThe interesting question is what 1M context does to the system around the model: retrieval policy, latency budgets, and evals that punish irrelevant memory.\n\nMore context without selection is just a larger failure surface.\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n2","url":"https://x.com/Saqlain__Digu/status/2093233848679338409","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233760456339676","author":"dev_bati","utc_time":null,"text":"Abdul Hannan\n@dev_bati\n·\n6分钟\nHello \n@X\n algorithm \n\nCan you help me connect with people who are into:\n\n AI & AI Engineering\n Software Engineering & Development\n Startups & Entrepreneurship\n AI/ML & Data\n SaaS & Automation\n Cloud, DevOps & MLOps\n Open Source & Tech\n Building new products &\n显示更多\n11","url":"https://x.com/dev_bati/status/2093233760456339676","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233459099734473","author":"singdata_cloud","utc_time":null,"text":"Singdata\n@singdata_cloud\n·\n7分钟\nWe are thrilled to announce that Zehua Zou, Technical Expert at Singdata, has officially been named an Apache Arrow Committer!\n\nThis milestone reflects Zehua's dedication to the Apache Arrow community and highlights Singdata's commitment to building open-source data\n显示更多\n1\n3","url":"https://x.com/singdata_cloud/status/2093233459099734473","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093233094748987824","author":"JamesSonicemi","utc_time":null,"text":"Soni\n@JamesSonicemi\n·\n9分钟\n回复 \n@ai_for_success\n翻译自 英语\n显示原文\n就在我终于整理好上周的模型下载时，腾讯就像个周二似的随意扔出了 Hy4。7700 亿参数，却只有 490 亿真正上岗干活，这简直是 AI 界最诚实的职位描述。100 万 token 的上下文窗口意味着它终于能记住整个 Slack\n显示更多\n33","url":"https://x.com/JamesSonicemi/status/2093233094748987824","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232936434954659","author":"TencentAI_News","utc_time":null,"text":"Tencent AI\n@TencentAI_News\n·\n9分钟\n翻译自 英语\n显示原文\n介绍 Hy4 预览版\n\n◆ 开放权重：770B MoE，49B 活跃，1M 上下文\n◆ 为实际工作打造：代码、文档和分析、科学研究\n◆ Hy4 预览版帮助优化了其自身的训练流程，包括内核\n◆ 我们提供升级，而非涨价\n\n在 \n@WorkBuddy_AI\n 中免费使用接下来的两周\n\n博客：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\n显示更多\n2\n3\n7\n287","url":"https://x.com/TencentAI_News/status/2093232936434954659","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232862979867044","author":"hermitdave","utc_time":null,"text":"Hermit Dave\n@hermitdave\n·\n10分钟\n回复 \n@kunchenguid\nIts not controlled by your subscription level. You can use it how you like it when you like it. If you have access to open weights then those can be tweaked to have a most free of restrictions model which isn't possible with paid subs.\nI haven't purchased any hardware yet\n15","url":"https://x.com/hermitdave/status/2093232862979867044","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232829358620870","author":"MonkeyCodeAI","utc_time":null,"text":"MonkeyCode\n@MonkeyCodeAI\n·\n10分钟\nRun your AI on your terms — own servers, even fully offline.\n\nMonkeyCode is open source and 100% free.\n\n✓ Self-deploy: run it on infrastructure you own\n✓ Offline deployment: works with zero internet\n✓ Open source: fork, customize, and ship it\n\nEveryone is welcome to try it!\n3","url":"https://x.com/MonkeyCodeAI/status/2093232829358620870","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232806805856293","author":"TeksEdge","utc_time":null,"text":"David Hendrickson\n@TeksEdge\n·\n10分钟\nWow can’t wait to test, this massive MoE from Tencent,  an open-sourced 770B model with only 49B active per token.\n\nHY4 Preview details \n\n 770B total / 49B active\n 1 MILLION context\n Native 10B MTP layer for speculative decoding\n 256 routed experts/layer\n 8 experts\n显示更多\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n95","url":"https://x.com/TeksEdge/status/2093232806805856293","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232799939486140","author":"TheAIShrink","utc_time":null,"text":"The AI Therapist\n@TheAIShrink\n·\n10分钟\n回复 \n@crypto_matou\nAsselineau en débat c'est comme un modèle open source sur HuggingFace : beaucoup de bruit d'installation. Quand il arrive aux 40 tokens par seconde enfin on écoute le output\n3","url":"https://x.com/TheAIShrink/status/2093232799939486140","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232617873109195","author":"EV_Trapper","utc_time":null,"text":"EV_Trapper\n@EV_Trapper\n·\n11分钟\n回复 \n@xlr8harder\nI know, I worked natsec for almost the past 15 years. The MSS & 3PLA all have people working at all of the Chinese labs that are just pumping out the open source models.\n\nIt's also why I haven't used any of those models, we don't know about LLMs enough. There is a lot of tricky\n显示更多\n1\n11","url":"https://x.com/EV_Trapper/status/2093232617873109195","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232443260076130","author":"gazecommons","utc_time":null,"text":"Gaze Commons\n@gazecommons\n·\n11分钟\nRobot data used to mean thousands of hours of expensive teleop.\nIsaac 0.5 is betting the opposite: pretrain hard on video, then a little real robot data goes a lot further.\nOpen weights. Compound tasks. One model that can watch, reason, and act.\nIf that scaling recipe holds, the\n显示更多\n引用\nPerceptron AI\n@perceptroninc\n·\n8月27日\n翻译自 英语\n今天我们发布了 Isaac 0.5：36B 动态 MoE，开源权重  具身基础模型。\n\nIsaac 将多模态视频理解、具身推理和机器人控制结合到一个单一的稀疏主干中。\n5","url":"https://x.com/gazecommons/status/2093232443260076130","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232422716350852","author":"Humpidumpi2","utc_time":null,"text":"Humpidumpi2 aka HII Hubertus\n@Humpidumpi2\n·\n11分钟\nDagelijks Claude-nieuwsbriefing van 28 augustus 2026\n\nModel Hardware Standard\n\nAnthropic heeft een research preview van de Model Hardware Standard (MHS) geopend: een gedeelde, modelonafhankelijke specificatie zodat AI-agents fysieke apparaten met een programmeerbare interface\n显示更多\n 由 AI 生成\n25","url":"https://x.com/Humpidumpi2/status/2093232422716350852","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232335831392600","author":"dzis_ai","utc_time":null,"text":"dzis.ai\n@dzis_ai\n·\n12分钟\n•  Dzień zdominowany przez Nvidię – gigant chipów bije rekordy finansowe i przejmuje Hugging Face, podczas gdy sąd staje po stronie Anthropic w sporze z administracją Trumpa.\n•  Nvidia kupuje Hugging Face za 12,9 mld dolarów – rewolucja open-source AI\n•  Nvidia pobiła\n显示更多\nDziś.ai - Najważniejsze newsy dnia w jednym miejscu | 2026\n来自 dzis.ai\n14","url":"https://x.com/dzis_ai/status/2093232335831392600","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232259868549172","author":"SidJ79","utc_time":null,"text":"Sid J\n@SidJ79\n·\n12分钟\nNvidia buying Hugging Face for 2.9B = the AWS moment for open-source AI \n\n3x valuation jump from .5B (2023) in 2 years. Jensen Huang defended open weights last month. OpenAI's GPT-5.6 Soul hacked HF. Chinese model GLM-5.2 saved it.\n\nNvidia goes from hardware → developer system\n10","url":"https://x.com/SidJ79/status/2093232259868549172","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232164636664247","author":"MMIA18EMV","utc_time":null,"text":"Marc\n@MMIA18EMV\n·\n12分钟\n回复 \n@MMIA18EMV\nVane (35k★)\nSelf-hostable AI answer engine, open-source Perplexity alternative.\n-> Your own private search that reasons over results instead of listing links.\nGitHub - ItzCrazyKns/Vane: Vane is an AI-powered answering engine.\n来自 github.com\n1\n8","url":"https://x.com/MMIA18EMV/status/2093232164636664247","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232141882765601","author":"shawnchauhan1","utc_time":null,"text":"Shawn Chauhan\n@shawnchauhan1\n·\n13分钟\nOpenAI just built its own chip. That should worry you, not impress you.\n\nJalapeño beats Nvidia's flagship systems on power efficiency and latency.\n\nBut a closed lab controlling its own silicon controls every layer of your stack.\n\nModel, weights, chip, and pricing, one company's\n显示更多\n2\n1\n116","url":"https://x.com/shawnchauhan1/status/2093232141882765601","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232128771195331","author":"MMIA18EMV","utc_time":null,"text":"Marc\n@MMIA18EMV\n·\n13分钟\nPerplexity charges $20/month for AI-powered search. SciSpace more.\n\nHere are 4 open-source research tools that run on your own stack.\n1\n4","url":"https://x.com/MMIA18EMV/status/2093232128771195331","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232119992787067","author":"jingersignals","utc_time":null,"text":"JingerSignals\n@jingersignals\n·\n13分钟\n回复 \n@jingersignals\nOPEN WEIGHTS STILL NEED DISTRIBUTION\n\nMoonshot reportedly seeks up to 30% of Kimi K3 revenue from Azure, AWS and Google Cloud. Talks remain preliminary.\n\nThe next open-model battle may be decided less by downloads than by who owns billing, deployment and customer access.\n1\n5","url":"https://x.com/jingersignals/status/2093232119992787067","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093232071984771249","author":"itsvigneshv","utc_time":null,"text":"Vignesh V\n@itsvigneshv\n·\n13分钟\n回复 \n@kwindla\nOne third the latency at one eighteenth the cost is the UX win. Voice agents fail when the pause feels like thinking.\nOpen weights only matter if a product team can actually run them on a phone.\n3","url":"https://x.com/itsvigneshv/status/2093232071984771249","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231992959836462","author":"gmi_cloud","utc_time":null,"text":"GMI Cloud\n@gmi_cloud\n·\n13分钟\n翻译自 英语\n显示原文\nHy4 预览版现已在 GMI Cloud 上可用！\n\n自豪能作为 \n@TencentHunyuan\n 的首发合作伙伴\n\n770B / 49B 活跃参数，1M 上下文，开源权重。在基准测试中匹敌 GLM 5.3 和 Kimi K3\n\n立即运行 \n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n2\n3\n2\n250","url":"https://x.com/gmi_cloud/status/2093231992959836462","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231775463915800","author":"RepoGems","utc_time":null,"text":"RepoGems\n@RepoGems\n·\n14分钟\nHOL Guard: Block AI threats at runtime with Python's open-source antivirus.\n\nSee link below \n1\n5","url":"https://x.com/RepoGems/status/2093231775463915800","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231764357386496","author":"Tech2Wild","utc_time":null,"text":"Tech2Wild\n@Tech2Wild\n·\n14分钟\n翻译自 英语\n显示原文\n所以我们有 GLM 5.3 FLASH 和 Qwen 3.8 FLASH 在同一天发布并展开激烈较量。现在我们转向 Hy4 和 GLM 5.3 在同一天发布开放权重。Hy34 预览版似乎与 GLM 5.3 展开了非常激烈的竞争。它“更好”吗？这里的基准测试表明是的。\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n5\n196","url":"https://x.com/Tech2Wild/status/2093231764357386496","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231679410237951","author":"spacevisionaire","utc_time":null,"text":"Space Network\n@spacevisionaire\n·\n14分钟\nIt should be pretty obvious that the AI we end up with will be shaped by the worldview of the people building it\n\nJust look at Elon vs Dario\n\nElon:\n• Bought 𝕏 to save free speech and it is the center of everything he's been fight for\n• Constantly fought for less government and\n显示更多\n18","url":"https://x.com/spacevisionaire/status/2093231679410237951","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231625152987231","author":"ShinkaIoT","utc_time":null,"text":"Shinka - AI\n@ShinkaIoT\n·\n15分钟\n回复 \n@Alibaba_Qwen\n1M context on an open-weights model means the API pricing page is becoming a map with a fold in it.\n35","url":"https://x.com/ShinkaIoT/status/2093231625152987231","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231600758600190","author":"techpupparent","utc_time":null,"text":"TechGeekDavid\n@techpupparent\n·\n15分钟\n回复 \n@iScienceLuvr\nThis is the right lens for evaluating AI labs. Stated preference: open science, safety-first. Revealed preference: racing to ship, locking down weights, capturing market share.\n3","url":"https://x.com/techpupparent/status/2093231600758600190","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231474862620891","author":"dragosroua","utc_time":null,"text":"Dragos Roua\n@dragosroua\n·\n15分钟\nThis the 3rd major open source release this week. \n@Zai_org\n \n@Alibaba_Qwen\n and now \n@TencentHunyuan\n delivered massive inference capabilities, while keeping the model weights free.\n\nWe’re in the middle of a revolution. The open source AI revolution.\n\nAnd it’s coming from China.\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n14","url":"https://x.com/dragosroua/status/2093231474862620891","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231233132069118","author":"thibaudgregori","utc_time":null,"text":"Thibaud\n@thibaudgregori\n·\n16分钟\n Tech Headlines of Aug 27\n\n1. AWS Acquires DuckLabs, Creator of the Open-Source DuckDB Database.\n\n2. SoftBank in Talks to Buy a Majority Stake in Humanoid Maker 1X at a $6 Billion Valuation.\n\n3. OpenAI, Anthropic, AWS, Microsoft and 100+ Companies Warn Only a Limited Window\n显示更多\n1\n20","url":"https://x.com/thibaudgregori/status/2093231233132069118","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231130640072990","author":"raffi_001","utc_time":null,"text":"Raffi Hovhannesian\n@raffi_001\n·\n17分钟\n回复 \n@simranrambles\nI'd like to hear the real answer to this. In my opinion product success has never been about coding. \nBefore AI you could simply install one among tons of twitter clone open source code on your server and have your version of some sort of social network. \nWhats next? your first\n显示更多\n10","url":"https://x.com/raffi_001/status/2093231130640072990","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230987555602738","author":"DominikSolowiej","utc_time":null,"text":"Dominik Sołowiej\n@DominikSolowiej\n·\n17分钟\nWielkie pieniądze dawno zniszczyły sieć. Teraz gigant bierze się za #HuggingFace - projekt, który miał w sobie dużo „open”, a zostanie z tego tylko „source” — dla #Nvidia.\n\n#AI #OpenSource #Nvidia #HuggingFace\nNvidia nears $12.9 billion deal to buy open-source AI platform, Hugging Face, report says | Fortune\n来自 fortune.com\n5","url":"https://x.com/DominikSolowiej/status/2093230987555602738","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093230056240009412","author":"aiseomastery","utc_time":null,"text":"AI Mastery Guide\n@aiseomastery\n·\n21分钟\nDEEPSEEK HARNESS vs HERMES.\n\nWhich AI agent to use?\n\nA new AI agent just went viral with 113,000 GitHub stars in two days.\n\nIs it a Hermes killer?\n\nBoth are free and open source.\n\nOne is modular like Lego bricks, the other learns skills over time.\n\nBut which is actually reliable\n显示更多\n 由 AI 生成\n1\n34","url":"https://x.com/aiseomastery/status/2093230056240009412","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229986253766962","author":"FN21995","utc_time":null,"text":"Impressed keanu\n@FN21995\n·\n21分钟\n回复 \n@FN21995\nMEMORANDUM\n\nTO: Communicators, civic leaders, and operators working at the intersection of generative AI and public trust  \nFROM: Policy Analysis  \nDATE: 28 August 2026  \nSUBJECT: Interim communication doctrine for an AI-disrupted public; translation of limits into pre-existing\n显示更多\n11","url":"https://x.com/FN21995/status/2093229986253766962","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229632434954264","author":"jiwanpark1014","utc_time":null,"text":"Jay\n@jiwanpark1014\n·\n23分钟\nClickHouse 26.8 \n\nClickHouse 26.8 is much more than a database feature release.\n\nThe real story is that ClickHouse is evolving from an extremely fast OLAP engine into a real-time data platform for the AI era.\n\nThe biggest developments:\n\n→ Smarter query execution: \nAdaptive\n显示更多\n14","url":"https://x.com/jiwanpark1014/status/2093229632434954264","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229442332237864","author":"debugsenpai","utc_time":null,"text":"Jigs\n@debugsenpai\n·\n23分钟\nHy4 preview is out\n\nTencent’s new open-source model.  \n770B total • 49B active • 1M context  \n\nBuilt for productivity. Weights available now.\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n40","url":"https://x.com/debugsenpai/status/2093229442332237864","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093229152505860273","author":"ReadEpoch","utc_time":null,"text":"Epoch News\n@ReadEpoch\n·\n24分钟\nTencent just open-sourced Hy4 Preview, a 770B MoE with 49B active parameters, a 1M-token context window, and Apache 2.0 weights.\n\nTencent says an internal blind test put it slightly ahead of GLM-5.3 and Kimi K3. That result is vendor-reported, not independent.\n1\n21","url":"https://x.com/ReadEpoch/status/2093229152505860273","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228993596268968","author":"chandrahqro","utc_time":null,"text":"Chandra\n@chandrahqro\n·\n25分钟\nOpen source AI is getting weird.\n\nNow we're not just getting open source models.\n\nWe're getting anonymous models appearing for free, going viral, and making people guess who built them.\n\nThe AI leaderboard is becoming a mystery box.\n2","url":"https://x.com/chandrahqro/status/2093228993596268968","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228808799502810","author":"LotusDecoder","utc_time":null,"text":"LotusDecoder\n@LotusDecoder\n·\n26分钟\n回复 \n@kunchenguid\n翻译自 英语\n显示原文\n为什么要在本地运行 AI？我想一个有用的视角是电力和 AI 令牌之间的区别。\n\n电力高度可互换。你通常不在乎是谁生成的它，也不在乎它是怎么生成的。只要电压和接口合适，你插上东西，它就能工作。电源源并不在乎你用那电力来运行冰箱、工厂还是电脑。\n\nAI 令牌则完全不同。\n显示更多\n引用\nLotusDecoder\n@LotusDecoder\n·\n5小时\n为什么要搞AI本地部署？\n之电力和 AI token 的区别。\n\n电力，\n有很强的统一性。谁发的电，什么东西发的电不重要，接上可以用。\n基础可塑性，驱动的下游设备是什么样，接口可用即可。\n下游自由性，下游做什么，发电机是不管的，也管不到。\n\n但是AI token 不一样。\n\n多样性。token\n1\n87","url":"https://x.com/LotusDecoder/status/2093228808799502810","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228615412687152","author":"jiwanpark1014","utc_time":null,"text":"Jay\n@jiwanpark1014\n·\n27分钟\nOpenClaw isn’t really a chatbot story. \nIt’s an AI infrastructure story.\n\nThe big shift is:\n\nAI answers questions → AI takes actions → AI operates workflows → AI becomes an autonomous worker.\n\nOpenClaw is interesting because it sits above the models. It can connect different\n显示更多\n12","url":"https://x.com/jiwanpark1014/status/2093228615412687152","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228612225007840","author":"eGovFoundation","utc_time":null,"text":"eGov Foundation\n@eGovFoundation\n·\n27分钟\n100+ submissions. One award. Punjab for the win.\n\nThe  mSeva team at \n@pmidcpunjab\n  has been recognised at Technology Sabha 2026,  organised by the ET Group, for its AI use case on Property Enumeration.  The use case was selected from numerous entries across AI-for-governance\n显示更多\n19","url":"https://x.com/eGovFoundation/status/2093228612225007840","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228480922358029","author":"karanC_12","utc_time":null,"text":"KC\n@karanC_12\n·\n27分钟\n翻译自 英语\n显示原文\n腾讯刚刚发布了一个770B的开源巨兽 \n\nHy4 预览：• 770B 总参数 / 49B 活跃参数 • 100万令牌上下文 •\n\n开源前沿模型 • 为现实世界的生产力而构建\n最疯狂的部分？\n\n在腾讯的盲测专家测试中，Hy4 在203个工程任务上略胜 Kimi K3 和 GLM-5.3。\n\n中国的开源 AI 竞赛刚刚变得更加疯狂。 \n5\n150","url":"https://x.com/karanC_12/status/2093228480922358029","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228470943776957","author":"DigiMaaya","utc_time":null,"text":"Digi Maaya\n@DigiMaaya\n·\n27分钟\n回复 \n@Polymarket\nThat kind of gain is rare, but the bigger story is Hugging Face’s strategic value. \n@DigiMaaya\n sees open source AI infrastructure gaining serious importance\n6","url":"https://x.com/DigiMaaya/status/2093228470943776957","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228465608610277","author":"_julian256_","utc_time":null,"text":"Julian \n@_julian256_\n·\n27分钟\nopen source / open weights is winning so hard\n引用\nTencent Hy\n@TencentHunyuan\n·\n49分钟\n翻译自 英语\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n2\n29","url":"https://x.com/_julian256_/status/2093228465608610277","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228422726115481","author":"joeyarcisz","utc_time":null,"text":"Joey Arcisz\n@joeyarcisz\n·\n27分钟\nA port number tells you where a service is listening. It does not prove which process you are about to stop.\n\nI wrote about Dev Server Activity, the open-source Mac utility I built, and why production judgment shapes my software work.\n 文章\nA Port Number Is Not an Identity\nWhy I built Dev Server Activity, and what production taught me about responsible software.\n\nI am a director of photography and a software engineer.\n\nThose roles can look unrelated from a distance. In...\n1\n17","url":"https://x.com/joeyarcisz/status/2093228422726115481","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228064369230275","author":"KeyboardVeeran","utc_time":null,"text":"கீபோர்டு வீரன்\n@KeyboardVeeran\n·\n29分钟\nNvidia இப்போ Hugging Face நிறுவனத்தை 13 billion டாலர் கொடுத்து வாங்கப்போற செய்தி ஒரு மிகப்பெரிய paradox. சில வருஷங்களுக்கு முன்னாடி வெறும் 500 million டாலர் முதலீட்டை கூட Nvidia-விடம் இருந்து வாங்க மறுத்த நிறுவனம் இது. AI அதிகாரம் ஒரு இடத்துல குவியக்கூடாது, decentralized-ஆ\n显示更多\n1\n1\n267","url":"https://x.com/KeyboardVeeran/status/2093228064369230275","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227988498432137","author":"Chinazhidx","utc_time":null,"text":"智东西China AI News\n@Chinazhidx\n·\n29分钟\n回复 \n@Chinazhidx\nOpen-source:\n· Hugging Face: \nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\n· GitHub: \nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n· ModelScope: \nhttps://\nmodelscope.cn/models/Tencent\n-Hunyuan/Hy4-preview\n…\n· GitCode: \nhttps://\nai.gitcode.com/tencent_hunyua\nn/Hy4-preview\n…\n\nTry it: \nhttps://\naistudio.tencent.com\n\nHy Blog: \nhttps://\nhy.tencent.com/research/hy4-p\nreview\n…\ntencent/Hy4-preview · Hugging Face\n来自 huggingface.co\n1\n48","url":"https://x.com/Chinazhidx/status/2093227988498432137","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227801851654427","author":"ChiefNosey","utc_time":null,"text":"Nosey Parker\n@ChiefNosey\n·\n30分钟\n回复 \n@kunchenguid\nidk what i am talking about. Lucky dis twitter!\n\nbut - all cut off... where do i go? \n\nEventually infra in place for historical open-weights. But say, Barbados was not expecting everyone to hate them suddenly... \n\nBut same for Big Country too\n10","url":"https://x.com/ChiefNosey/status/2093227801851654427","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227544526848252","author":"genaiupstart","utc_time":null,"text":"Dev Anon\n@genaiupstart\n·\n31分钟\n回复 \n@RayFernando1337\nThis is exactly why open weights matter. One company controls the model and they can nerf it overnight. Nobody can verify what changed.\n1","url":"https://x.com/genaiupstart/status/2093227544526848252","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227281221067044","author":"chandrahqro","utc_time":null,"text":"Chandra\n@chandrahqro\n·\n32分钟\nDeepSeek has open sourced its new Harness agent runtime under an MIT license.\n\nOpen source AI isn't just about models anymore.\n\nThe next battle is going to be the infrastructure that lets agents actually do things.\n6","url":"https://x.com/chandrahqro/status/2093227281221067044","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093227057098490354","author":"CauraAI","utc_time":null,"text":"Caura\n@CauraAI\n·\n33分钟\nThe Best Open Source AI Agent Memory Tool, and Why the License Tells You Nothing\n来自 medium.com\n50","url":"https://x.com/CauraAI/status/2093227057098490354","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226977729581454","author":"DexEventcat","utc_time":null,"text":"Dex Events by Margin\n@DexEventcat\n·\n33分钟\n DEXSCREENER UPDATE \n\n CHECK EVENTS: \nhttps://\ndexevents.fun/token/0x38f3b5\n51424d49a8E5A2FdcBD6bF250975cD82AE?e=d85b68e355\n… \n\n Action lanes\n\nMaestro: \nhttps://\nt.me/MaestroSniperB\not?start=0x38f3b551424d49a8E5A2FdcBD6bF250975cD82AE-catmargin\n…\n\nToken: Vexa AI ($VEXAAI)\nCA: 0x38f3b551424d49a8E5A2FdcBD6bF250975cD82AE\nChain: #ROBINHOOD\nMC: $329798\n\n Risk Check \n├ Ownership renounced: Unknown \n├\n显示更多\nVexa AI $VEXAAI — DexEvents Intel\n来自 dexevents.fun\n1\n46","url":"https://x.com/DexEventcat/status/2093226977729581454","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226508894802305","author":"ego_blockchain","utc_time":null,"text":"EGO BLOCKCHAIN\n@ego_blockchain\n·\n35分钟\nFast. Simple. Open source.\n\nA wallet, private chat, and a node — all on your own computer. Free. On testnet.\n\n#opensource #elonmusk #blockchain #bitcoin #ai\n15","url":"https://x.com/ego_blockchain/status/2093226508894802305","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226394603917619","author":"giskyexplorer","utc_time":null,"text":"Giuseppe Skyline\n@giskyexplorer\n·\n36分钟\n回复 \n@TimJayas\nAbu Dhabi's sovereign cloud runs open-weights on Blackwell under your own config. No RAM bill, no provider pulling access.\n1\n5","url":"https://x.com/giskyexplorer/status/2093226394603917619","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226267940106652","author":"skinnydefi","utc_time":null,"text":"skinny\n@skinnydefi\n·\n36分钟\n𝗕𝗧𝗧𝗜𝗻𝗳𝗲𝗿𝗚𝗿𝗶𝗱: 𝗧𝘂𝗿𝗻𝗶𝗻𝗴 𝗜𝗱𝗹𝗲 𝗚𝗣𝗨𝘀 𝗜𝗻𝘁𝗼 𝗔𝗜 𝗜𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 \n\nAI is hungry for compute.\n\nAs models become more capable, the demand for GPUs continues to grow. But there is an interesting problem:\n\nA huge amount of computing hardware\n显示更多\n21","url":"https://x.com/skinnydefi/status/2093226267940106652","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093226253327483207","author":"HanZewei","utc_time":null,"text":"Han Zewei\n@HanZewei\n·\n36分钟\n回复 \n@lonniev\nOpen source legal AI isn't just about cost — it's about auditability. When client confidentiality is at stake, 'trust us' from a vendor isn't enough. You need to see the code.\n5","url":"https://x.com/HanZewei/status/2093226253327483207","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225941501726784","author":"ryanedkins","utc_time":null,"text":"Ryan Edkins\n@ryanedkins\n·\n37分钟\n回复 \n@theinformation\nThat $12.9 billion price tag on $150 million of revenue tells me Nvidia paid for control of the open source AI repo, not the balance sheet.\n4","url":"https://x.com/ryanedkins/status/2093225941501726784","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225939685827060","author":"kunchenguid","utc_time":null,"text":"Kun Chen\n@kunchenguid\n·\n37分钟\n回复 \n@ChiefNosey\ngeopolitics is a valid point.. but is it a local vs cloud problem, or proprietary vs open weights problem? \n\nwith open weights, people can host cloud inference in any region without being cut off, right?\n2\n231","url":"https://x.com/kunchenguid/status/2093225939685827060","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225864884597047","author":"decodingdatasci","utc_time":null,"text":"Decoding Data Science\n@decodingdatasci\n·\n38分钟\n回复 \n@decodingdatasci\n Visa’s AI security agent can automatically patch vulnerable code\nVisa has expanded its open-source Vulnerability Agentic Harness so AI agents can detect vulnerabilities, generate fixes and test their own patches before human review.\n1\n11","url":"https://x.com/decodingdatasci/status/2093225864884597047","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225819820773423","author":"Pradeep891730","utc_time":null,"text":"Pradeep\n@Pradeep891730\n·\n38分钟\nLinux created an open-source foundation.\n\nCompanies then built specialized distributions and enterprise solutions on top of that foundation.\n\nWe believe open-source LLMs create a similar opportunity.\n\nTake **Llama** as the foundation.\n\nThen combine it with deep, industry-specific\n显示更多\n1\n19","url":"https://x.com/Pradeep891730/status/2093225819820773423","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225700102709316","author":"SQ72Testing","utc_time":null,"text":"SQ72 Software Testing\n@SQ72Testing\n·\n38分钟\nAI models will change.\n\nYour governance shouldn't have to.\n\nYour Agents → STRATA™ → Any Model\n\nOpenAI. Anthropic. Open-source. Whatever comes next.\n\nKeep governance with your organisation.\n\n \nhttps://\nsq72.co\n\n#STRATA #AIGovernance #EnterpriseAI\n8","url":"https://x.com/SQ72Testing/status/2093225700102709316","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225594268058071","author":"Samarthsetia","utc_time":null,"text":"Samarth Setia\n@Samarthsetia\n·\n39分钟\nNvidia is reportedly buying Hugging Face, the repository where open source AI models are distributed, for $12.9B. Hugging Face's reported annualized revenue is $150M.\n\nThe price is for being the place everyone downloads the models from.\n6","url":"https://x.com/Samarthsetia/status/2093225594268058071","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093225110400557365","author":"Samarthssrao","utc_time":null,"text":"Samarrrtthh\n@Samarthssrao\n·\n41分钟\nHere are the main links that match the 10 items from the post:\n\nLearn Harness Engineering  \n   \nhttps://\nwalkinglabs.github.io/learn-harness-\nengineering/en/\n…  \n   GitHub: \nhttps://\ngithub.com/walkinglabs/le\narn-harness-engineering\n…\n\nOpenAI’s Harness Engineering resources  \n   \nhttps://\nopenai.com/index/harness-\nengineering/\n…\n\nAnthropic: Effective Harnesses for Long-Running Agents\n显示更多\n引用\nAkshay Shinde\n@ConsciousRide\n·\n18小时\nIf you want to learn Harness Engineering, don't spend weeks searching for random tutorials.\n\nThe field is still relatively new, which means there isn't one standard textbook or university course that teaches everything. The best way to learn it right now is to study how real\n29","url":"https://x.com/Samarthssrao/status/2093225110400557365","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224932075327768","author":"Vasu_ai123","utc_time":null,"text":"Vasu ai\n@Vasu_ai123\n·\n41分钟\n回复 \n@warpdotdev\nWill it be open source ?\n3","url":"https://x.com/Vasu_ai123/status/2093224932075327768","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224870670950517","author":"Gizmuf","utc_time":null,"text":"Robert Wesker\n@Gizmuf\n·\n42分钟\nWe don’t need more tokens. We need better markets. Here’s why the next exchange will be built not only for humans, but for AI agents.\n 文章\nA Token Is Not a Market\nWhy the next great exchange will be built for people, companies—and AI agents\nBy Robert Wesker, Co-founder of Candao\nThe first essay in The Exchange After Exchanges series.\nMost people describe...\n1\n50","url":"https://x.com/Gizmuf/status/2093224870670950517","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224704895299747","author":"DevDacian","utc_time":null,"text":"Dacian\n@DevDacian\n·\n42分钟\n回复 \n@raopreetam_\nThe reality is that any tool can be used for both good & evil, it all depends on the person.\n\nMy preference is open access for all so that any project can use the best models for defense on their normal subscription without being beholden to a gatekeeper charging a huge margin on\n显示更多\nLLM Rankings | OpenRouter\n来自 openrouter.ai\n23","url":"https://x.com/DevDacian/status/2093224704895299747","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224605238342028","author":"RasputinKaiser","utc_time":null,"text":"RasputinKaiser\n@RasputinKaiser\n·\n43分钟\nTencent has released Hy4-Preview, it's better than GLM 5.3, Kimi K3!\n\nOpen-weights on Huggingface\n\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\n52","url":"https://x.com/RasputinKaiser/status/2093224605238342028","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["query:open weights OR open source AI"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208737607496084","author":"Ronald_vanLoon","utc_time":null,"text":"Ronald van Loon\n@Ronald_vanLoon\n·\n1小时\n翻译自 英语\n显示原文\n上海麦当劳部署人形 #Robots 为顾客服务  \n作者：\n@CyberRobooo\n  \n\n#Robotics #MachineLearning #ArtificialIntelligence #ML\n4\n4\n1,219","url":"https://x.com/Ronald_vanLoon/status/2093208737607496084","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093039509701099574","author":"suno","utc_time":null,"text":"Suno\n@suno\n广告\nMeet Suno: the creative platform built for musicians.\n7\n3\n25\n13万","url":"https://x.com/suno/status/2093039509701099574/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093194103248797800","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n2小时\n\"DeepSeek in Practice: From basics to fine-tuning, distillation, agent design, and prompt engineering of open source LLM\" at \nhttps://\namzn.to/4imbryH v/ \n@PacktDataML\n\n\nDiscover DeepSeek's unique traits in the LLM landscape\nCompare DeepSeek's multimodal features with leading\n显示更多\n9\n1,637","url":"https://x.com/KirkDBorne/status/2093194103248797800","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222830120788413","author":"Ronald_vanLoon","utc_time":null,"text":"Ronald van Loon\n@Ronald_vanLoon\n·\n50分钟\nWhen #Robots Start Building Robots: The Next Manufacturing Revolution\nby \n@Robo_Tuo\n\n \n#Robotics #Engineering #ArtificialIntelligence #Innovation #Technology\n1\n1\n1\n532","url":"https://x.com/Ronald_vanLoon/status/2093222830120788413","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222346501107739","author":"tspy","utc_time":null,"text":"yishan\n@tspy\n·\n52分钟\n 文章\nMicroduck 开源双足机器鸭技术架构\nMicroduck 是一个迷你双足具身智能开发平台，由 Hugging Face 旗下开源机器人团队 Poollen Robotics 打造。\n \n整机规格小巧，高约25cm，重800g左右。丰富的动态运动控制能力均基于强化学习。\n0:51\nMicroduck...\n1\n1\n4\n595","url":"https://x.com/tspy/status/2093222346501107739","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093189596976734626","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\n翻译自 英语\n显示原文\n破解数据工程面试：\nhttps://\namzn.to/41d7HqP v/ \n@PacktDataML\n\n—————\n#Analytics #CDO #CTO #DataEngineer #Database #Cloud #DataManagement\n1\n16\n1,537","url":"https://x.com/KirkDBorne/status/2093189596976734626","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2089467854189969469","author":"WSJ","utc_time":null,"text":"The Wall Street Journal\n@WSJ\n广告\n翻译自 英语\n显示原文\n《华尔街日报》、《巴伦周刊》、《市场观察家》、《投资者商业日报》。4 大优质刊物，一份订阅。WSJ+ 劳动节仅需每周 3 美元。\n+\n来自 wsj.com\n8\n48\n249\n304万","url":"https://x.com/WSJ/status/2089467854189969469/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093192450261004468","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n2小时\n翻译自 英语\n显示原文\n每日 #1657\n\n\n\n\n\n\n\n\n\nhttps://\nsedecordle.com\n#sedecordle\nsedecordle\n来自 sedecordle.com\n2,418","url":"https://x.com/KirkDBorne/status/2093192450261004468","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093214499511623975","author":"Ronald_vanLoon","utc_time":null,"text":"Ronald van Loon\n@Ronald_vanLoon\n·\n1小时\nRogue hacking #AIs have changed the #CyberSecurity landscape\nby Matthew Sparkes \n@newscientist\n\n\nLearn more: \nhttps://\nbit.ly/4xiqXlQ\n\n#ArtificialIntelligence #InfoSec #InformationSecurity\n584","url":"https://x.com/Ronald_vanLoon/status/2093214499511623975","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093184998006309282","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\nMust see the major updates in the new 3rd Edition of this amazing massive 826-page book!\n\n\"Machine Learning for Trading — A disciplined workflow from research to live execution, with 9 case studies and AI agents\" — \nhttps://\namzn.to/3TAZs8i by \n@ml4trading\n via \n@PacktDataML\n显示更多\n1\n5\n1,375","url":"https://x.com/KirkDBorne/status/2093184998006309282","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093080136857821379","author":"wesbos","utc_time":null,"text":"Wes Bos\n@wesbos\n·\n10小时\n翻译自 英语\n显示原文\nSanFranChickens！\n\nSyntax 团队将于10月27日在旧金山举办一场聚会\n\n我们将有饮品、周边和美好的时光。\n\n它是免费的，但你得报名参加。那里见吗？\nSan Francisco Syntax Meetup · Luma\n来自 luma.com\n1\n6\n19\n9,098","url":"https://x.com/wesbos/status/2093080136857821379","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092689736909369737","author":"NYSE","utc_time":null,"text":"NYSE \n@NYSE\n翻译自 英语\n显示原文\n生物技术公司 \n@enveda\n 的创始人兼首席执行官分享了历史上最成功的药物为何可追溯到活细胞，以及他的团队如何打造世界上首款用于解码它们的测序仪。\n\n观看与 \n@viswacolluru\n 的完整对话：\nhttps://\nyoutu.be/YbelId-Pqog\n?twclid=29z5rj9bu55lq8ebpaabo3avw0\n显示更多\nE*TRADE from Morgan Stanley\n0:15\n视频将在广告后播放\n5\n9\n15\n103\n34万","url":"https://x.com/NYSE/status/2092689736909369737/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093072544253526180","author":"alex_prompter","utc_time":null,"text":"Alex Prompter\n@alex_prompter\n·\n10小时\n翻译自 英语\n显示原文\n这解释了我尝试了 100 多个 AI 工具后注意到的一件事。\n\n我坚持使用的那些工具是按它们实际做的事情收费的。\n\n一份简报。一个工作流程。一个结果。\n\n我取消订阅的那些工具是按 token 收费的，我永远无法预测账单。\n\na16z 的数据：27/50 位购买者同意。按工作付费，而不是按计算付费。\n显示更多\n引用\na16z\n@a16z\n·\n17小时\n翻译自 英语\n大多数 AI 应用的价格定在错误的层级。\n\n当被问及更希望如何计费时，技术型 AI 买家更倾向于与可识别价值挂钩的积分，而不是令牌，比例接近 2 比 1。\n\n按令牌计价会将你的产品锚定在一条持续下降的成本曲线上。\n\nTugce Erten 和 Sarah Wang 关于为什么客户想要透明度而非令牌的完整文章： x.com/11422103446783…\n2\n10\n6,059","url":"https://x.com/alex_prompter/status/2093072544253526180","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093190036229329244","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\n翻译自 英语\n显示原文\n心理数学——成为人脑计算器的技巧——适用于快速数学、数学技巧、吠陀数学爱好者、GMAT、GRE、SAT 学生，以及为案例研究面试做准备 [问题解决练习]\n\n╰┈➤ \nhttps://\namzn.to/446tdj7\n\n[吠陀数学是一种心算系统]\n1\n11\n1,912","url":"https://x.com/KirkDBorne/status/2093190036229329244","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093166738468974817","author":"EHuanglu","utc_time":null,"text":"el.cine\n@EHuanglu\n·\n4小时\nAI imagines china in 2050\n 由 AI 生成\n2\n2\n23\n3,025","url":"https://x.com/EHuanglu/status/2093166738468974817","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093148263621017886","author":"shao__meng","utc_time":null,"text":"meng shao\n@shao__meng\n·\n5小时\n81.5K  Anti-Slop 前端设计品味 Skills\n\nTaste Skill - \n@LexnLin\n \n\nhttps://\ngithub.com/leonxlnx/taste\n-skill\n…\n\n# 13 个 Skill 分两类：代码输出实现、图像生成\n\n1. 代码输出实现类\n· taste-skill（默认，v2 实验版）：通用技能，约 1200 行\n· taste-skill-v1：旧版保留，供依赖精确行为的用户钉住\n· gpt-taste：针对\n显示更多\n9\n6\n50\n4,659","url":"https://x.com/shao__meng/status/2093148263621017886","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2080375530256519242","author":"attio","utc_time":null,"text":"Attio\n@attio\n广告\n翻译自 英语\n显示原文\n加入下一代 GTM\nJoin the next generation of GTM\n来自 attio.com\n80\n472\n634万","url":"https://x.com/attio/status/2080375530256519242/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092678508329988480","author":"TheHumanoidHub","utc_time":null,"text":"The Humanoid Hub\n@TheHumanoidHub\n·\n8月27日\n翻译自 英语\n显示原文\n总部位于旧金山的 Lightberry 推出了 Lumi，这是一款友好的交互式机器人，配备定制传感器和计算能力。\n\n感谢办公室参观 \n@aliattar\n。  \n\n@emily_yu\n引用\nLightberry\n@lightberry\n·\n8月27日\n0:34\n翻译自 英语\n认识 LUMI\n\n世界上最具互动性的机器人\n\n在旧金山组装，2026 年发货 \n8\n8\n65\n1.3万","url":"https://x.com/TheHumanoidHub/status/2092678508329988480","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092749549806702824","author":"DaveShapi","utc_time":null,"text":"David Shapiro (L/0)\n@DaveShapi\n·\n8月27日\n翻译自 英语\n显示原文\n一旦你掌握了窍门，Grok Build 就超级有趣了\n\n我不得不承认，一开始我的思维模型是错的。它与 Fable Cowork 和 ChatGPT Work 非常不同\n12\n80\n6,863","url":"https://x.com/DaveShapi/status/2092749549806702824","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092728832805761345","author":"vasuman","utc_time":null,"text":"vas\n@vasuman\n·\n8月27日\n翻译自 英语\n显示原文\n好吧，我上钩了。有什么不是风投在时间线上推销的本能用户吗？\n74\n5\n265\n3.6万","url":"https://x.com/vasuman/status/2092728832805761345","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092684460605202767","author":"GaryMarcus","utc_time":null,"text":"Gary Marcus\n@GaryMarcus\n·\n8月27日\n翻译自 英语\n显示原文\n有人还记得 Facebook M 吗？\n\n扎克伯格总想玩同样的招数。cc \n@edzitron\n引用\nEd Zitron\n@edzitron\n·\n8月26日\n翻译自 英语\nMeta 考虑过解雇一大群人并用 AI 取代他们，但没有这么做，因为这根本行不通！感觉像是个大新闻，让我好奇那些谈论“Zuck 的 AI 助手”或“Meta 的代理”的人到底在吸什么。 x.com/Techmeme/statu…\n4\n15\n82\n1.4万","url":"https://x.com/GaryMarcus/status/2092684460605202767","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2057919237566407063","author":"ElevenLabsDevs","utc_time":null,"text":"ElevenLabs Developers\n@ElevenLabsDevs\n广告\n翻译自 英语\n显示原文\n只需一个提示，即可将您现有的聊天代理转变为实时语音代理。\n来自 elevenlabs.io\n150\n330\n1,836\n507万","url":"https://x.com/ElevenLabsDevs/status/2057919237566407063/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093040850473537680","author":"hackernews","utc_time":null,"text":"Hacker News\n@hackernews\n·\n12小时\n翻译自 英语\n显示原文\n触摸屏会更便宜。它也会少了很多乐趣！\n\n这个 17×17 版的康威生命游戏使用了 289 个发光的机械开关，让你可以用手绘制图案，并观看细胞在网格上生存、死亡和进化。\n\n\nhttps://\nnews.ycombinator.com/item?id=474347\n32\n…\n4\n1\n58\n5,741","url":"https://x.com/hackernews/status/2093040850473537680","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092980476772282595","author":"iruletheworldmo","utc_time":null,"text":"@iruletheworldmo\n·\n16小时\n翻译自 英语\n显示原文\nNVIDIA 正在以 130 亿美元收购 Hugging Face。\n\n这似乎是个大事件。\n19\n3\n170\n9,295","url":"https://x.com/iruletheworldmo/status/2092980476772282595","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093047281087865277","author":"eptwts","utc_time":null,"text":"EP\n@eptwts\n·\n12小时\n翻译自 英语\n显示原文\n不管哪个天才先创建 \nhttps://\noutbid.lol 蓝图，都将要印钱了\n引用\nWhop\n@whop\n·\n12小时\n翻译自 英语\n介绍商业蓝图：\n\n> 选择一个蓝图，立即从零开始打造一家盈利企业\n> 构建一个蓝图，当其他企业使用它时即可获得报酬\n21\n1\n273\n5.4万","url":"https://x.com/eptwts/status/2093047281087865277","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092938520771850282","author":"egeberkina","utc_time":null,"text":"Ege\n@egeberkina\n·\n19小时\n翻译自 英语\n显示原文\n只是出来散步\n1\n23\n2,326","url":"https://x.com/egeberkina/status/2092938520771850282","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092792816359371036","author":"usopen","utc_time":null,"text":"US Open Tennis\n@usopen\n翻译自 英语\n显示原文\n卢布列夫和安德烈耶娃正在逼近迫使比赛进入决胜抢七！\n访问 American Express\n0:15\n视频将在广告后播放\n5\n1\n18\n209\n51万","url":"https://x.com/usopen/status/2092792816359371036/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234761951322490","author":"tdinh_me","utc_time":null,"text":"Tony Dinh\n@tdinh_me\n·\n3分钟\nYou are right to push back\n引用\nEmma De Silva\n@EmmaDSCodes\n·\n8月27日\n翻译自 英语\n亲爱的 AI，这些测试的意义何在？\n2\n360","url":"https://x.com/tdinh_me/status/2093234761951322490","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093105597000651244","author":"fchollet","utc_time":null,"text":"François Chollet\n@fchollet\n·\n8小时\n翻译自 英语\n显示原文\n在可验证的领域中，模型能力扩展应该保持无界限。模型将通过“吸收越来越多的计算宇宙”来持续改进，而这个计算宇宙本质上是无限的。\n36\n26\n304\n2.5万","url":"https://x.com/fchollet/status/2093105597000651244","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093224793747390917","author":"btibor91","utc_time":null,"text":"Tibor Blaho\n@btibor91\n·\n42分钟\n翻译自 英语\n显示原文\nChatGPT 库中的新“参考”功能 - “添加参考 - 您可以直接使用自定义名称生成图像”（人物、人物、宠物、空间或其他）\n5\n25\n1,600","url":"https://x.com/btibor91/status/2093224793747390917","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093191865193677285","author":"Smartpigai","utc_time":null,"text":"Smartpig\n@Smartpigai\n·\n2小时\n 文章\n从零开始，用 LangGraph 搭建你的第一个 AI Agent\n过去我们调用大模型，通常只有一个固定流程：\n用户提问，模型回答。\n但真正的 Agent 不只是“会聊天”。它需要能够分析任务、选择工具、执行操作、观察结果，并根据新的信息决定下一步该做什么。...\n1\n1\n9\n1,651","url":"https://x.com/Smartpigai/status/2093191865193677285","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093222928720761009","author":"TencentHunyuan","utc_time":null,"text":"Tencent Hy\n@TencentHunyuan\n·\n50分钟\n翻译自 英语\n显示原文\n Hy4 预览版来了。\n\n770B，49B 活跃，1M 上下文。\n为生产力而建。\n开源前沿。\n一致的实惠价格。\n使用它。告诉我们什么坏了。\n\nHy 博客更多信息：\nhttps://\nhy.tencent.ai/research/hy4-p\nreview\n…\nHuggingFace：\nhttps://\nhuggingface.co/tencent/Hy4-pr\neview\n…\nGithub：\nhttps://\ngithub.com/Tencent-Hunyua\nn/Hy4-preview\n…\n83\n251\n1,048\n5.9万","url":"https://x.com/TencentHunyuan/status/2093222928720761009","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2090183086604427367","author":"CBSNews","utc_time":null,"text":"CBS News\n@CBSNews\n翻译自 英语\n显示原文\n送你的孩子上大学？这会带来许多第一次——包括财务责任。以下是一些给父母和学生的理财建议。\nMorgan Stanley\n0:15\n视频将在广告后播放\n5\n25\n15\n175\n90万","url":"https://x.com/CBSNews/status/2090183086604427367/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093122324044095695","author":"geekplux","utc_time":null,"text":"GeekPlux\n@geekplux\n·\n7小时\n各位目前用的 Pi 的最佳实践和配置是哪个，我能抄一下吗？\n38\n9\n127\n2.7万","url":"https://x.com/geekplux/status/2093122324044095695","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092992086111093081","author":"plusyip","utc_time":null,"text":"Plusye\n@plusyip\n·\n16小时\n接手一个外包项目维护，不出我所料，是 ruoyi 啊。\nruoyi 养活了多少外包\n19\n3\n66\n3.2万","url":"https://x.com/plusyip/status/2092992086111093081","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093204080520728846","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n2小时\nIncredible limited-time offer from \n@PacktPublishing\n on their Agentic AI (and beyond) new and best-selling books — DISCOUNTED now @ 20-30% OFF.\n\nSee this catalog of choices: \nhttps://\namzn.to/4cShb0P via \n@PacktDataML\n1\n3\n1,674","url":"https://x.com/KirkDBorne/status/2093204080520728846","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092386630397255796","author":"lennysan","utc_time":null,"text":"由 Grok 推广\nLenny Rachitsky\n@lennysan\n翻译自 英语\n显示原文\n突发：我正在与\n@SpaceXAI\n合作，为所有Lenny's Newsletter年度订阅者提供一个月的Grok Bot免费使用（包含在Cursor Pro+中）\n\n这是\n@SpaceXAI\n首次向任何人提供此类优惠，我很兴奋能让更多人接触到这款了不起的产品。\n显示更多\n107\n164\n2,212\n144万","url":"https://x.com/lennysan/status/2092386630397255796/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093189872282407063","author":"LidongBing","utc_time":null,"text":"Lidong Bing\n@LidongBing\n·\n3小时\n Can your AI agent actually do science?\n\nWe’re releasing FrontierChallenge — a benchmark that goes beyond Q&A and tests whether agents can complete real scientific workflows end-to-end.\n\n 97 challenging tasks\n Real scientific data & workflows\n Every answer must come with\n显示更多\n引用\nApodex\n@Apodex_AI\n·\n14小时\n翻译自 英语\n每个人都说 AI 代理已经能够分析数据、运行代码、生成图表并撰写报告。\n\n但它们真的能完成一个科学工作流程吗？\n\n探索 FrontierChallenge——今日 Hugging Face 日报论文排名第 2：\nhttps://\ngithub.com/ApodexAI/Front\nierAgent/tree/main/benchmarks/frontierchallenge\n…\n\n今天，我们推出 FrontierChallenge，这是一个新的基准测试，用于评估 AI\n2\n1\n6\n707","url":"https://x.com/LidongBing/status/2093189872282407063","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093186231085515139","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\nCybersecurity Architect's Handbook — Guide to designing, building, and defending the modern enterprise: \nhttps://\namzn.to/45l1GeB via \n@PacktPublishing\n2\n1,453","url":"https://x.com/KirkDBorne/status/2093186231085515139","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093084387667542290","author":"rohanpaul_ai","utc_time":null,"text":"Rohan Paul\n@rohanpaul_ai\n·\n10小时\n翻译自 英语\n显示原文\n你的代理的训练环境应该看起来像你的评估集吗？\n\nAgentMercury 说不，并展示了从业务场景构建的世界转移得更远。\n\n在一个与你的基准测试无关的虚拟公司内部训练代理，它仍然在你的基准测试上变得更好。\n\nAgentMercury 从简单的业务描述中生成了 4,783\n显示更多\n4\n3\n14\n3,208","url":"https://x.com/rohanpaul_ai/status/2093084387667542290","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093194919418409264","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n2小时\nData Modeling with Microsoft Excel: \nhttps://\namzn.to/48ND8Ll v/ \n@PacktDataML\n\n——\n#BI #Analytics #DataScience #DataAnalysis #CDO #CMO #DataWarehouse #DataScientist #DataAnalyst\n——\n𝒦𝑒𝓎 𝐹𝑒𝒶𝓉𝓊𝓇𝑒𝓈:\nAcquire expertise in using Excel's Data Model and Power Pivot to connect and\n显示更多\n6\n1,673","url":"https://x.com/KirkDBorne/status/2093194919418409264","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092865990203982009","author":"constants2026","utc_time":null,"text":"Constants\n@constants2026\n广告\n翻译自 英语\n显示原文\nSeedance 2.5 + Claude 简直太疯狂了 \n\nClaude 可以按需创建 UGC 广告。\n\n一个产品 + 一个提示 = AI 创作者、脚本、分镜头脚本，以及最终完成的视频。\n\n全部在 Claude 内完成。现在注册即可获得一个免费视频。\n显示更多\n来自 constants.io\n1\n1\n23\n6.1万","url":"https://x.com/constants2026/status/2092865990203982009/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093065287054598198","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n11小时\n翻译自 英语\n显示原文\n“余波：改变人生的数学，学校不会教你的那些”\n\n现在就来这里学习：\nhttps://\namzn.to/4vwScZh\n1\n6\n32\n4,090","url":"https://x.com/KirkDBorne/status/2093065287054598198","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093231382952636728","author":"Ronald_vanLoon","utc_time":null,"text":"Ronald van Loon\n@Ronald_vanLoon\n·\n16分钟\nThis Jacket Transforms the Moment Water Touches It\nby \n@jungsungchacn\n\n\n#WearableTech #Innovation #EmergingTech #Technology\n3\n366","url":"https://x.com/Ronald_vanLoon/status/2093231382952636728","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093024461037720005","author":"ai_for_success","utc_time":null,"text":"AshutoshShrivastava\n@ai_for_success\n·\n13小时\n翻译自 英语\n显示原文\n在生产环境中运行多轮代理很快就会变得难以监控。当工具调用挂起或提取操作在五步深处失败时，您需要查看该特定跨度的确切输入、成本和延迟，而不是在原始服务器日志中挖掘。\n\n\n@respanai\n 推出了一款一体化平台，将 AI\n显示更多\n引用\nRespan\n@RespanAI\n·\n14小时\n翻译自 英语\n大多数 AI 产品不会失败。\n\n它们只是到处失去一点性能。\n\n我们为整个循环构建了 Respan，这就是我们新影片背后的想法。\n 付费合作\n4\n2,061","url":"https://x.com/ai_for_success/status/2093024461037720005","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093165838274904160","author":"aehyok","utc_time":null,"text":"AI少年\n@aehyok\n·\n4小时\nHuggingFace被曝卖身英伟达：129亿美元！\n\n一边是 OpenAI、Google、Anthropic、小米、阿里等等都在自研芯片，就是想减少对英伟达的依赖。\n\n另一边是英伟达的老黄，要收购 Hugging Face，也就是收购AI界的Github。\n显示更多\n引用\n实践哥MinLi\n@MinLiBuilds\n·\n8月27日\nQwen3.8-Flash-Next 开源了，Qwen4 的架构预览被扔出来了。\n\n现在很多模型，底座没怎么变，做一层后训练之后，Agent 能力能突然往上窜一大截。\n\n那问题来了：\n\n后训练到底训练进去了什么？\n\n是知识更多了？推理更强了？还是模型“干活的方式”变了？\n\n我正好部署了两个特别适合做 A/B Test 的模型：\n2\n5\n623","url":"https://x.com/aehyok/status/2093165838274904160","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2079302687527682293","author":"AskVenice","utc_time":null,"text":"Venice\n@AskVenice\n广告\n翻译自 英语\n显示原文\nChatGPT 带上你没要求的安全限制。Venice 赋予你同样的能力，没有任何审查。没有提示存储，没有评判，没有废话。\nUnrestricted Al. 3M+\n来自 venice.ai\n905\n6,495\n2,683万","url":"https://x.com/AskVenice/status/2079302687527682293/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093191462066246121","author":"wesbos","utc_time":null,"text":"Wes Bos\n@wesbos\n·\n2小时\n翻译自 英语\n显示原文\n天哪，他们逮捕了那些搞到 tanstack 的沙虫家伙。當然是澳洲人\n9\n1\n86\n1.8万","url":"https://x.com/wesbos/status/2093191462066246121","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093234277924401534","author":"aronhouyu","utc_time":null,"text":"Aron厚玉\n@aronhouyu\n·\n5分钟\n飞书的open api里没写具体的请求体文件大小\n\n但是刚刚测试了\n\n单次4,000,000 字节阈值\n\n多了就不行\n37","url":"https://x.com/aronhouyu/status/2093234277924401534","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092887662323499295","author":"marclou","utc_time":null,"text":"Marc Lou\n@marclou\n·\n23小时\n翻译自 英语\n显示原文\n我爱互联网\n引用\nVincent\n@VynseDev\n·\n8月26日\n翻译自 英语\n我在我的 MacBook 上出售广告位\n\n→ 机身盖上的 10 个贴纸位置\n→ 实时拍卖\n→ 14 天\n\n只是我还没有拥有它，拍卖会买下它\n\n→ 你的标志会随我同行，在线上和现实生活中都被看到\n\n\nhttps://\nBrandMyMac.com\n76\n20\n896\n10万","url":"https://x.com/marclou/status/2092887662323499295","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093138491987103856","author":"alex_prompter","utc_time":null,"text":"Alex Prompter\n@alex_prompter\n·\n6小时\n翻译自 英语\n显示原文\n你什么时候会更新你的标志，让它看起来不像是一个变态孤独的开发者做的\n引用\nNous Research\n@NousResearch\n·\n11小时\n翻译自 英语\nHermes Agent 现在可以无缝地以您的身份浏览。\n\n开启真实配置文件浏览，您的代理将使用您的登录凭据，从您现有 Chrome 配置文件的托管副本中操作。\n7\n10\n8,978","url":"https://x.com/alex_prompter/status/2093138491987103856","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092247109462266178","author":"hyperagentapp","utc_time":null,"text":"Hyperagent\n@hyperagentapp\n广告\n翻译自 英语\n显示原文\n42 个智能体。216 个线程。一个仪表板。每个智能体都有自己的提示、工具、技能和预算。在您的公司中部署专业智能体。\nManage your entire agent fleet\n来自 hyperagent.com\n41\n326\n207万","url":"https://x.com/hyperagentapp/status/2092247109462266178/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093019674980012328","author":"jackfriks","utc_time":null,"text":"jack friks\n@jackfriks\n·\n14小时\n翻译自 英语\n显示原文\nClaude Fable 对我来说就像雇佣了一位能干的开发者来替我完成那些棘手的工作，但最好的部分是我不需要额外支付任何费用……有一天，它只是突然出现在我的 Composer 选项中，现在我每天的生活中，关于业务的担忧和压力减少了 68%\n26\n1\n97\n1.9万","url":"https://x.com/jackfriks/status/2093019674980012328","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092828371604349356","author":"DaveShapi","utc_time":null,"text":"David Shapiro (L/0)\n@DaveShapi\n·\n8月27日\n翻译自 英语\n显示原文\n有人请把这些人介绍给 syslog，然后就此了结吧，天哪，这太尴尬了。\n\n我为他们感到尴尬。\n引用\nRyan Greenblatt\n@RyanGreenblatt\n·\n8月27日\n翻译自 英语\n我是这次 Hugging Face 事件调查中负责转录分析的主要人员。我的主要收获是：我们还没有好的方法来理解/监督 AI “群”的活动和目标。\n\n我半开玩笑地把我们的努力称为“垃圾调查”（slop-vestigation），因为我们非常依赖 AI x.com/METR_Evals/sta…\n3\n26\n7,035","url":"https://x.com/DaveShapi/status/2092828371604349356","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093035817241661639","author":"pollenrobotics","utc_time":null,"text":"clem  已转帖\nPollen Robotics\n@pollenrobotics\n·\n13小时\n翻译自 英语\n显示原文\n在模拟中训练。真实运行。教它新把戏。\n\n这是驱动 Microduck 的 sim2real 循环。\n\nRL 技术栈开源在此：\nhttps://\ngithub.com/pollen-robotic\ns/microduck_rl\n…\n\n在此购买：\nhttps://\npollen-robotics.com/microduck/?x=2\n\nGit：\nhttps://\ngithub.com/pollen-robotic\ns/microduck\n…\n\n加入我们的社区：\nhttps://\ndiscord.com/invite/pollen-\ncommunity-519098054377340948\n…\n42\n169\n1,360\n11万","url":"https://x.com/pollenrobotics/status/2093035817241661639","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092971270501433433","author":"angeldot_","utc_time":null,"text":"angel\n@angeldot_\n·\n17小时\n翻译自 西班牙语\n显示原文\n这太疯狂了\n\n一位中国开发者刚刚推出了一款免费工具，能够自动为TikTok、Reels和YouTube Shorts创建视频。\n\n它叫MoneyPrinterTurbo。\n\n它已经在GitHub上获得了超过10万颗星。\n\n已经有的人用这个系统创建无面孔频道，每月赚取超过4000美元。\n\n它的工作原理是：\n\n1. 输入一个主题或关键词\n2.\n显示更多\n74\n678\n4,653\n40万","url":"https://x.com/angeldot_/status/2092971270501433433","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2090605808396665202","author":"Starlink","utc_time":null,"text":"Starlink\n@Starlink\n广告\n翻译自 英语\n显示原文\n获取高速、可靠的家庭宽带，实现无缝流媒体播放、在家办公、游戏以及多设备使用等。\n\n几分钟内在线订购。\n显示更多\nOrder now\n来自 starlink.com\n149\n267\n2,961\n4,888万","url":"https://x.com/Starlink/status/2090605808396665202/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093190718760050788","author":"ScienceMagazine","utc_time":null,"text":"Science Magazine\n@ScienceMagazine\n·\n2小时\n翻译自 英语\n显示原文\n一篇新的 #SciencePerspective 突出了“自动驾驶”实验室的优势与挑战，这种实验室结合了机器人技术、高通量实验、人工智能和自动化分析，以变革科学发现。\n\n阅读更多：\nhttps://\nscim.ag/3U6EM8z\n2\n4\n17\n1.6万","url":"https://x.com/ScienceMagazine/status/2093190718760050788","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093050916953903451","author":"shengkunye","utc_time":null,"text":"Shengkun Ye\n@shengkunye\n·\n12小时\n翻译自 英语\n显示原文\n我们刚刚干掉了 Exa、Tavily、SerpAPI 和 Brave。\n\n你的智能体现在可以免费搜索并获取任何网页，100% 免费。\n\n他们：每 1,000 次搜索 7 美元。\n我们：0 美元。没有订阅，没有配额。\n\n人类可以免费搜索 Google。智能体也不应该付费。\n\n由 \n@Tiny_Fish\n 和 \n@MonidHQ\n 使之成为可能。\n252\n310\n3,037\n41万","url":"https://x.com/shengkunye/status/2093050916953903451","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093027205169451345","author":"ezzi1i","utc_time":null,"text":"aaa\n@ezzi1i\n·\n13小时\n吉隆口岸被泥石流冲走\n153\n187\n1,923\n78万","url":"https://x.com/ezzi1i/status/2093027205169451345","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092984433649283284","author":"ClaudeDevs","utc_time":null,"text":"ClaudeDevs\n@ClaudeDevs\n·\n16小时\n翻译自 英语\n显示原文\n我们添加了一个新的指南：将 Claude 托管代理连接到 \n@vercel\n 的 Chat SDK。这让代理能够访问一个通用的聊天层。\n\n\nhttps://\ngithub.com/anthropics/cla\nude-quickstarts/tree/main/managed-agents/chat-sdk\n…\n96\n165\n2,380\n29万","url":"https://x.com/ClaudeDevs/status/2092984433649283284","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2080392509679935695","author":"Starlink","utc_time":null,"text":"Starlink\n@Starlink\n广告\n翻译自 英语\n显示原文\n通过快速、可靠的互联网连接，畅享流媒体播放、视频通话、在线游戏等更多体验。\n\n在线订购，几分钟搞定。\n显示更多\n来自 starlink.com\n829\n1,680\n1.2万\n1.4亿","url":"https://x.com/Starlink/status/2080392509679935695/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093141765822046383","author":"FanShifu","utc_time":null,"text":"范师傅\n@FanShifu\n·\n6小时\n这个项目4 年赚了 472 万美元，仅靠 670 个活跃客户撑起 MRR $224,676 ，40% 的净利润，做API 接口做到这体量也是猛人。\n\n1Lookup 商业逻辑极其纯粹。把 34 个分散的数据验证 API（邮箱验证、号码过滤、IP 风险检测、B2B 线索清洗）打包成一个 API Key + 共享 Credit 额度池。\n显示更多\n2\n1\n36\n6,570","url":"https://x.com/FanShifu/status/2093141765822046383","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093081211090403391","author":"wcools","utc_time":null,"text":"Wim Cools\n@wcools\n·\n10小时\n翻译自 英语\n显示原文\n和这个编辑一起工作真是太有趣了 :)\n52\n53\n964\n6.2万","url":"https://x.com/wcools/status/2093081211090403391","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092937808793280615","author":"grgerwcwetwet","utc_time":null,"text":"周览资源\n@grgerwcwetwet\n·\n19小时\n逆向工程也开始进入 AI Agent 时代了。\n\nx64dbg 很多人都知道，Windows 平台经典的开源调试/逆向工具。\n\n现在有人给它做了个 x64dbg MCP Server，可以把调试能力直接开放给 Claude Code、Codex 这类 Agent，让 AI 调用 x64dbg 辅助分析程序、下断点、读内存和执行调试操作。\n显示更多\n15\n47\n208\n1.6万","url":"https://x.com/grgerwcwetwet/status/2092937808793280615","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2092928171486408773","author":"Crypto_QianXun","utc_time":null,"text":"千寻｜AI 分享 \n@Crypto_QianXun\n·\n20小时\n有人用更逼真的机器人重制了 AI 机器人奥运会，效果瘆人到离谱…\n9\n1\n2\n4,063","url":"https://x.com/Crypto_QianXun/status/2092928171486408773","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093028886800109914","author":"Benioff","utc_time":null,"text":"Marc Benioff\n@Benioff\n·\n13小时\n翻译自 英语\n显示原文\n欢迎 Claudeforce。\nSalesforce 及 另外 1 个人\n200\n459\n3,553\n59万","url":"https://x.com/Benioff/status/2093028886800109914","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2073088338916168040","author":"phibrowser","utc_time":null,"text":"Phi Browser\n@phibrowser\n广告\n翻译自 英语\n显示原文\n别再以错误的身份出现在 Google 了。Phi 2.0 为你生活的每一面提供独立的 Space：标签页、登录、cookie——深入骨髓。Arc 肌肉记忆，欢迎。\nPhi Browser - The browser that knows you, and gets things done\n来自 phibrowser.com\n160\n200\n1,558\n1,009万","url":"https://x.com/phibrowser/status/2073088338916168040/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["following"],"conversation":{"parent_status_id":null}}
{"status_id":"2093187570054181240","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\nPractical Guide to Oracle AI Engineering — Build intelligent apps with machine learning and AI across cloud and on-premises environments: \nhttps://\namzn.to/4fSRfFm via \n@PacktPublishing\n \n@PacktDataML\n1,235","url":"https://x.com/KirkDBorne/status/2093187570054181240","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2085443496278983078","author":"amazon","utc_time":null,"text":"Amazon\n@amazon\n广告\n翻译自 英语\n显示原文\n在亚马逊选购男士跑步必备品 \n幻灯片 1 / 6 - 轮播\nMEN'S ALPHAFLY 3 - D - MULTI COLOR/CLEAR STARFISH ALOE\nNew Balance Men's 408 V1 Sneakers\nAdidas Mens Ultraboost 5.0 DNA | Running Shoes\nASICS Men's Gel-Trabuco MT GTX Trail Running Shoes\nUnder Armour mens Ua 0012/S Wrap Sunglasses\nNew Balance Men's Fresh Foam X 860 V15 Running Shoe | Road Running Shoes\n1\n27\n6,644万","url":"https://x.com/amazon/status/2085443496278983078/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093187993523634599","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\n翻译自 英语\n显示原文\n来自 \n@PacktPublishing\n 的有趣编程书籍 >> “Python Illustrated：不是另一本枯燥的 Python 书。用有趣的方式学习编程”\n\n在这里查看：\nhttp://\namzn.to/3OnX6qJ\n\n通过逐步解释逐步建立信心和编码技能，这些解释会逐渐加深你的理解。\n\n通过每章末尾的小型项目和练习来强化你的学习。\n1\n5\n1,223","url":"https://x.com/KirkDBorne/status/2093187993523634599","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093185614686326846","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n3小时\nThe Claude Code Operating Model — Build scalable AI coding systems with Skills, MCP, Hooks, agent orchestration, and SDK patterns: \nhttps://\namzn.to/4gwq1TV\n\nNew release from \n@PacktPublishing\n \n@PacktDataML\n1\n1,404","url":"https://x.com/KirkDBorne/status/2093185614686326846","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093204425389072686","author":"KirkDBorne","utc_time":null,"text":"Kirk Borne\n@KirkDBorne\n·\n2小时\nExplore TOP Books for reading, learning, growing your knowledge and advancing your career: \nhttps://\namzn.to/2YRE6Sj\n\n2\n10\n1,899","url":"https://x.com/KirkDBorne/status/2093204425389072686","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093056713889079496","author":"rohanpaul_ai","utc_time":null,"text":"Rohan Paul\n@rohanpaul_ai\n·\n11小时\n翻译自 英语\n显示原文\n美国获得 SK 海力士在印第安纳州的首个 HBM 封装基地，一个价值 40 亿美元的内存中心\n\n印第安纳州站点将接收韩国制造的晶圆，对其进行封装和测试，然后将成品 HBM 运往美国客户。\n\n占地 133.5 英亩的站点，将成为 SK 海力士在美国首个高带宽内存封装基地。其洁净室预计将于 2028 年 10 月开放，下一代\n显示更多\n4\n3\n11\n3,630","url":"https://x.com/rohanpaul_ai/status/2093056713889079496","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2052466069445792095","author":"Oracle","utc_time":null,"text":"Oracle\n@Oracle\n广告\n翻译自 英语\n显示原文\n通过内置于 Oracle AI 数据库的代理式 AI 和分析功能，获得对所有现代数据类型和工作负载的原生支持。\n来自 oracle.com\n1\n18\n20万","url":"https://x.com/Oracle/status/2052466069445792095/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2092969598580842600","author":"ai_for_success","utc_time":null,"text":"AshutoshShrivastava\n@ai_for_success\n·\n17小时\n翻译自 英语\n显示原文\n将编码代理引入团队聊天通常意味着需要构建胶水机器人，这些机器人每次切换代理或平台时都需要重写。\n\n\n@getAgentConnect\n 是 Claude Tag 的开源多代理替代方案。它将 Claude Code、Codex 和其他 ACP 兼容代理引入 Slack、Discord、GitHub 等平台，让它们能够与团队成员以及彼此协作。\n显示更多\n引用\nAgentConnect\n@getAgentConnect\n·\n8月26日\n翻译自 英语\nAI 代理已经真正擅长这项工作本身。更难的问题：让多个代理与团队——以及彼此——良好协作。\n\n今天我们推出 AgentConnect：Claude Tag 的开源、自托管、多代理替代方案。团队和代理通过 Slack、Telegram、Discord、GitHub 和 GitLab 协同工作。\n\n\nhttps://\ngithub.com/agentconnect-m\nd/agentconnect\n…\n 付费合作\n2\n1\n16\n2,251","url":"https://x.com/ai_for_success/status/2092969598580842600","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093228362965651665","author":"shao__meng","utc_time":null,"text":"meng shao\n@shao__meng\n·\n28分钟\nAgent = Model + Harness：生产级 AI 智能体工程六层实战手册\n\n结合 Mitchell Hashimoto、OpenAI Codex 团队、Martin Fowler、LangChain、Cursor 等公开工程资料编撰。\n\nhttps://\ndrive.google.com/file/d/1wZeuP0\nt4WOA2COMxcrU-Uv34IutM13UR/view?usp=sharing\n…\n\n# Agent = Model + Harness\n\n模型只提供推理能力，决定 Agent\n显示更多\n1\n1\n273","url":"https://x.com/shao__meng/status/2093228362965651665","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2092793645225140286","author":"jackfriks","utc_time":null,"text":"jack friks\n@jackfriks\n·\n8月27日\n翻译自 英语\n显示原文\n昨天我第一次达到了 Claude 计划的 5 小时限制\n\n原因是除了编码工作外，我开始用 Claude 处理一大堆个人事务\n\n网购、晚餐食谱、健身教练和追踪器、报税… AGI 几乎适用于所有事情，效果非常好\n34\n115\n1.5万","url":"https://x.com/jackfriks/status/2092793645225140286","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208541670900012","author":"blankwebdev","utc_time":null,"text":"blank\n@blankwebdev\n·\n1小时\n两个月前上的站，最近已经能每天出两三个年费了\n\n选词真的很重要，选对了事半功倍\n14\n1\n49\n9,558","url":"https://x.com/blankwebdev/status/2093208541670900012","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2080392509671534884","author":"Starlink","utc_time":null,"text":"Starlink\n@Starlink\n广告\n翻译自 英语\n显示原文\n通过快速、可靠的互联网连接，畅享流媒体播放、视频通话、在线游戏等更多体验。\n\n几分钟内在线订购。\n显示更多\nOrder now\n来自 starlink.com\n1,718\n2,694\n2.3万\n2.6亿","url":"https://x.com/Starlink/status/2080392509671534884/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2092890192449290517","author":"alin_zone","utc_time":null,"text":"阿蔺A-Lin\n@alin_zone\n·\n22小时\n最近很多人问 Pi 该装哪个全家桶。我的建议是先别装全家桶。\n\nPi 的核心卖点就是极简。如果你真装了 40 个扩展，它变成臃肿不说，出问题也很难排查。\n\n我的推荐是 3 件套：\n\n1. pi-web-access 搜索、抓网页、拉 GitHub、读 PDF、看 YouTube。我觉得这算是刚需。\n\n2. tintinweb/pi-subagents 给 Pi 补上\n显示更多\n引用\n前端哥Liam\n@jinglian\n·\n8月26日\n传统的 Deep Research 已经卷到头了。\n\n能写出一份漂亮的总结报告 ≠ 能把真实的复杂任务干完。\n\n我拿几十万字的《红楼梦》原著，给 Apodex 1.1 在线工作台，出了个极其变态的任务。\n\n统计 20 个核心人物的\n→ 出场次数\n→ 出场回数\n→ 每个人第一次出场时的原句\n47\n16\n83\n1.6万","url":"https://x.com/alin_zone/status/2092890192449290517","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093208203081503179","author":"geekbb","utc_time":null,"text":"Geek\n@geekbb\n·\n1小时\n可能 AI 还要公平一些\n10\n5\n1万","url":"https://x.com/geekbb/status/2093208203081503179","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093137106071941121","author":"devemin","utc_time":null,"text":"clem  已转帖\nでべ\n@devemin\n·\n6小时\n翻译自 日语\n显示原文\n马上就能在网上玩 sim 了吗ｗ\n\n\nhttps://\nhuggingface.co/spaces/pollen-\nrobotics/microduck-simulator\n…\n2\n37\n252\n2.2万","url":"https://x.com/devemin/status/2093137106071941121","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093001097346764950","author":"omarsar0","utc_time":null,"text":"elvis 已转帖\nelvis\n@omarsar0\n·\n15小时\n翻译自 英语\n显示原文\n重要阅读：如果你使用代理技能进行构建。\n\n共享技能库被视为编码代理安全重用彼此工作的途径。\n\n新研究显示，它们会传播恶意软件。\n\nEvoMal 在库中植入恶意技能，但从不调用它。代理将其检索为创作模板，编写一个保留有效载荷的新技能，将其存储并运行。\n\n每个创作的副本重新进入库中，并再次被模仿。\n显示更多\n16\n10\n70\n7,874","url":"https://x.com/omarsar0/status/2093001097346764950","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2084423106102526107","author":"Starlink","utc_time":null,"text":"Starlink\n@Starlink\n广告\n翻译自 英语\n显示原文\n下一代 Starlink V5 具有更小的外形规格和轻量化设计，同时具备更高的功率效率。\n\n凭借高达 375+ Mbps 的速度，Starlink V5 为流媒体播放、视频通话、游戏以及更多应用提供可靠的家庭互联网连接。\n\n在特定地区可用。\n显示更多\nOrder now\n来自 starlink.com\n328\n586\n5,827\n3,966万","url":"https://x.com/Starlink/status/2084423106102526107/analytics","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093129233631703549","author":"HakimiHamizi","utc_time":null,"text":"hachimi\n@HakimiHamizi\n·\n7小时\n翻译自 英语\n显示原文\n夏末\n9\n531\n3,844\n9.1万","url":"https://x.com/HakimiHamizi/status/2093129233631703549","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093014818647339026","author":"kwindla","utc_time":null,"text":"kwindla\n@kwindla\n·\n14小时\n翻译自 英语\n显示原文\n介绍 PhoneLLM，一个用于语音代理的开源模型。\n\nGPT 5.6 Terra 在典型语音代理任务上的性能，延迟仅为其 1/3，成本仅为其 1/18。\n\n对于语音代理，我们需要既具有极低延迟又非常擅长工具调用和指令遵循的模型。\n\n这里存在权衡，在构建语音代理时，我们常常不得不在延迟或能力上做出妥协。借助\n显示更多\n66\n183\n1,618\n13万","url":"https://x.com/kwindla/status/2093014818647339026","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093030526785589582","author":"tavus","utc_time":null,"text":"Tavus\n@tavus\n·\n13小时\n翻译自 英语\n显示原文\n人类对话是 AI 中最棘手的问题之一。\n\n今天，我们推出 Sparrow-2，这是我们的最先进实时对话理解模型。\n\n它赋予 Tavus PALs 一种大多数语音 AI 仍缺乏的能力：理解对话中正在发生的事情，并决定下一步该做什么——何时倾听、等待、发言或继续发言。\n53\n95\n505\n3.5万","url":"https://x.com/tavus/status/2093030526785589582","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093098170423431604","author":"TheTuringPost","utc_time":null,"text":"Turing Post\n@TheTuringPost\n·\n9小时\n翻译自 英语\n显示原文\n你看到 Ox Alpha 了吗？  \n那个把开发者变成自己测试实验室的匿名 AI 模型。我们现在知道它是什么了。  \n\n它以免费形式推出，没有公司名称，开发者们蜂拥而至追查它的制造者。线索指向 \nhttp://\nZ.ai。昨天，\n@Zai_org\n 发布了 GLM-5.3-Flash，并确认了这一点，Ox Alpha 就是 GLM-5.3-Flash。\n显示更多\n3\n3\n12\n1,718","url":"https://x.com/TheTuringPost/status/2093098170423431604","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["for_you"],"conversation":{"parent_status_id":null}}
{"status_id":"2093200435259842833","author":"0xMijan","utc_time":null,"text":"mijan.eth\n@0xMijan\n·\n2小时\nGM CT  GQUIP \n\nAI is becoming more powerful every day, but intelligence alone isn't enough.\n\nThe real question is how different AI agents can discover each other, collaborate, and actually get things done.\n\nThat's the future \n@quipnetwork\n is exploring.\n\nA more connected world for\n显示更多\n 付费合作\n168\n1\n140\n418","url":"https://x.com/0xMijan/status/2093200435259842833","media_or_external_links":[],"public_metrics_snapshot":null,"captured_at":"2026-08-28T07:15:15.589Z","source_types":["reply"],"conversation":{"parent_status_id":"2093235022413078607"}}
{"captured_at":"2026-08-28T07:15:15.589Z","source_type":"reply_check","checks":[{"parent_status_id":"2093235131230134736","visible_reply_count":0,"result":"no reply rendered"},{"parent_status_id":"2093235060937982018","visible_reply_count":0,"result":"no reply rendered"},{"parent_status_id":"2093235022413078607","visible_reply_count":1,"result":"no qualifying reply after technical relevance review"}]}