{
 "about": "海外 VC / 投资人自己写的生产力套件、工作空间、Agent 记忆与协作文章（赛道论点、市场地图、投资备忘），每篇打开原文核对过标题、日期、作者。gist 是核心观点的中文转述，不是原文；seg 对应 layers.json 的 code，all = 赛道总论；vs = 我们对「Agent 会不会改写工作空间、价值会不会移到新的一层」态度的归类。",
 "articles": [
  {
   "t": "Rethinking Communication for Human and AI Teams",
   "org": "Emergence Capital",
   "au": "Santi Subotovsky",
   "date": "2026-09-24",
   "url": "https://www.emcap.com/thoughts/rethinking-communication-for-human-and-ai-teams",
   "seg": [
    "a2"
   ],
   "co": [
    "Ando",
    "Slack",
    "Zoom"
   ],
   "kind": "投资备忘",
   "gist": "Agent 正在成为同事，现有沟通平台都假定参与者是人；Ando 从第一天就让人和 Agent 同为工作空间成员，Agent 跟着对话积累上下文并按决策行动。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Our Investment in Ando: Modern Messaging for Teams and Agents",
   "org": "Accel",
   "au": "Ivan Zhou",
   "date": "2026-09-24",
   "url": "https://www.accel.com/news/our-investment-in-ando-modern-messaging-for-teams-and-agents",
   "seg": [
    "a2"
   ],
   "co": [
    "Ando",
    "Slack",
    "Teams"
   ],
   "kind": "投资备忘",
   "gist": "AI 原生的工作越来越多人参与，协调越来越多 Agent 的上下文、记忆和活动会指数级变难；Ando 让人和 Agent 并肩工作，Agent 受和人同样的边界约束。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "The Incumbents Are Coming",
   "org": "a16z",
   "au": "Seema Amble",
   "date": "2026-09-03",
   "url": "https://www.a16z.news/p/the-incumbents-are-coming",
   "seg": [
    "all"
   ],
   "co": [
    "Salesforce",
    "Atlassian",
    "Docusign",
    "Klaviyo",
    "Harvey",
    "Anthropic"
   ],
   "kind": "论点",
   "gist": "通用 Agent（如 Claude）能直达记录系统的数据，在位者也在自己加 Agent，记录系统可能更重要而不是更不重要；但在位者的判断只限于自己持有的记录，一份完整的业务工作跨应用跨团队跨外部方，垂直 AI 公司靠聚焦某项工作和专家反馈的学习闭环仍能赢。",
   "top": 1,
   "vs": "有条件",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Our Investment in Engram",
   "org": "General Catalyst",
   "au": "Marc Bhargava、Max Rimpel",
   "date": "2026-06-23",
   "url": "https://www.generalcatalyst.com/stories/our-investment-in-engram",
   "seg": [
    "a3"
   ],
   "co": [
    "Engram"
   ],
   "kind": "投资备忘",
   "gist": "制约 AI 的是记忆而不是推理；更长的上下文和更好的检索只是更大的文件柜，不是学习；Engram 的赌注是把知识直接训进权重，调用便宜、更新成本比反复喂材料低几个数量级。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Who Owns Memory? The Market Map for AI Agents' Missing State Layer",
   "org": "个人投资人（Substack）",
   "au": "Rishi Maheshwari",
   "date": "2026-06-08",
   "url": "https://rishimaheshwari.substack.com/p/who-owns-memory-the-market-map-for",
   "seg": [
    "a3"
   ],
   "co": [
    "Mem0",
    "Zep",
    "Letta",
    "Cognee",
    "Supermemory",
    "Glean",
    "Dust",
    "Granola"
   ],
   "kind": "市场地图",
   "gist": "记忆是记忆、身份、算力三项基础原语里最没定型的一项；赢家要握住工作流里一个受信的控制点（上下文来源、写入策略层或使用记忆的运行时）。文内把记忆市场分六类并附图。页面未写所在基金，按个人投资人收录。",
   "top": 1,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "From \"System of Record\" to \"System of Intelligence\"",
   "org": "a16z",
   "au": "Gio Ahern、Stephenie Zhang、Alex Immerman",
   "date": "2026-05-14",
   "url": "https://a16z.com/from-system-of-record-to-system-of-intelligence/",
   "seg": [
    "all"
   ],
   "co": [
    "Salesforce",
    "HubSpot",
    "Oracle",
    "ServiceNow"
   ],
   "kind": "论点",
   "gist": "以 CRM 为例：Salesforce、HubSpot 仍握有有价值的数据，但长期价值会流向在其上读写数据、跨工具编排工作的推理层「智能系统」；AI 扩大的是整体预算而不是砍人工成本。",
   "top": 1,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Why We Invested in Lua",
   "org": "Flourish Ventures",
   "au": "Efayomi Carr、Ameya Upadhyay",
   "date": "2026-04-16",
   "url": "https://flourishventures.com/perspectives/why-we-invested-in-lua/",
   "seg": [
    "a2"
   ],
   "co": [
    "Lua",
    "Norrsken22",
    "Y Combinator"
   ],
   "kind": "投资备忘",
   "gist": "中型企业想用 Agent，但现有选择要么贵且重咨询，要么太技术，要么太浅；Lua 让技术与非技术团队都能搭建、部署并拥有多步 Agent，有望成为新兴市场中型企业的基础 Agent 层。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "AI Infrastructure Roadmap: Five frontiers for 2026",
   "org": "Bessemer Venture Partners",
   "au": "Janelle Teng Wade、Lance Co Ting Keh、Talia Goldberg 等",
   "date": "2026-03-30",
   "url": "https://www.bvp.com/atlas/ai-infrastructure-roadmap-five-frontiers-for-2026",
   "seg": [
    "a3"
   ],
   "co": [
    "Braintrust",
    "Judgment Labs",
    "Bigspin.ai"
   ],
   "kind": "年度预测",
   "gist": "五个前沿的第一项是围绕模型的 harness 基础设施，包含记忆与上下文管理和评测观测；模型商品化后差异化移到记忆与上下文层，它正在成为独立品类；另一个前沿是连续学习系统，文中指出冻结权重限制了部署后的学习。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "AI and the End of the Productivity Bundle",
   "org": "Sapphire Ventures",
   "au": "Cathy Gao、Kevin Burke、Marina Beveridge、Aditya Reddy",
   "date": "2026-02-03",
   "url": "https://sapphireventures.com/blog/ai-and-the-end-of-the-productivity-bundle/",
   "seg": [
    "all"
   ],
   "co": [
    "Microsoft Loop",
    "Google Workspace（Docs + Gemini）",
    "Glean",
    "Gamma",
    "Lovable",
    "Wonderful",
    "Sola",
    "LangChain",
    "CrewAI",
    "NotebookLM"
   ],
   "kind": "市场地图",
   "gist": "AI 正在拆散邮件、文档、表格、幻灯片这套靠切换成本维持的生产力套件；可持续的优势来自随时间累积的企业上下文，而不是 Agent 界面本身；控制面上移到 ChatGPT、Claude 这类入口，套件退成后端。文内附一张生产力赛道的市场图。",
   "top": 1,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Context Graphs: AI's Trillion-Dollar Opportunity",
   "org": "Foundation Capital",
   "au": "Jaya Gupta、Ashu Garg",
   "date": "2025-12-23",
   "url": "https://foundationcapital.com/ideas/context-graphs-ais-trillion-dollar-opportunity",
   "seg": [
    "a3",
    "all"
   ],
   "co": [
    "Salesforce",
    "Workday",
    "SAP",
    "Zendesk",
    "PagerDuty",
    "Slack",
    "Snowflake",
    "Databricks",
    "Neon"
   ],
   "kind": "论点",
   "gist": "现有记录系统存的是结果，从没存过决策的理由。站在 Agent 执行路径上的创业公司，每次运行都写下「决策轨迹」（谁批准、什么例外、什么先例），这些轨迹拼成的上下文图谱会成为新的记录系统；在位者和数据仓库拿不到这层。",
   "top": 1,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "The State of AI 2025",
   "org": "Bessemer Venture Partners",
   "au": "Kent Bennett、Talia Goldberg、Janelle Teng Wade 等",
   "date": "2025-08-13",
   "url": "https://www.bvp.com/atlas/the-state-of-ai-2025",
   "seg": [
    "a3",
    "all"
   ],
   "co": [
    "Mem0",
    "Zep",
    "Supermemory",
    "MemTensor 记忆张量（MemOS）",
    "Notion",
    "Salesforce",
    "ServiceNow"
   ],
   "kind": "年度预测",
   "gist": "对 AI 应用创始人，上下文和记忆可能是新的护城河；AI 压低迁移成本，削弱大型记录系统的壁垒，走向「行动系统」；预测 2025–2026 在位者收购 AI 能力的浪潮，含编排、评测、可观测和记忆工具。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  },
  {
   "t": "Bringing Memory to AI Agents（Our investment in Letta）",
   "org": "Felicis",
   "au": "Astasia Myers",
   "date": "2024-09-23",
   "url": "https://www.felicis.com/blog/letta",
   "seg": [
    "a3"
   ],
   "co": [
    "Letta"
   ],
   "kind": "投资备忘",
   "gist": "持久、与模型无关的记忆是可靠、个性化 Agent 的核心基础设施，多 Agent 之间还需要带权限与隐私控制的记忆共享；Felicis 领投 Letta 1000 万美元种子轮。",
   "top": 0,
   "vs": "看好",
   "r": 1,
   "region": "海外"
  }
 ]
}