剛剛，90後新CEO對釘釘動刀：這五個人“向我彙報”

在本教學中，我們建置了一套完整的 Crawlee for Python 工作流程，涵蓋環境設定、本機網站生成、靜態爬取、動態爬取、結構化擷取及下游資料處理。首先，我們配置相容的 Crawlee 執行環境，包含固定版本的 Pydantic 支援、Playwright 瀏覽器安裝、持久化儲存目錄，以及 Colab 安全執行處理。接著，我們產生一個擬真的本機示範網站，包含產品頁面、文件頁面、部落格內容、內部連結、robots.txt 規則、JSON-LD 元資料，以及 JavaScript 渲染的商品目錄。透過 BeautifulSoupCrawler，我們執行快速的遞迴 HTML 爬取，並擷取頁面標題、元資料、文字預覽、對外連結、產品屬性、文件標題、程式碼區塊與部落格標籤。

Nous Research has added a Blank Slate setup mode to its open-source Hermes Agent. It inverts the usual onboarding. Instead of a fully loaded default, you start with almost nothing. Hermes Agent is the self-improving agent framework from Nous Research. It runs on your own machine. The team announced the new mode on X. Blank Slate now joins two existing options: Quick Setup and Full Setup. TL;DR Blank Slate boots an agent with everything off except provider & model, File Operations, and Terminal. Web, browser, code execution, vision, memory, delegation, cron, skills, plugins, and MCP stay disabled. It writes an explicit platform_toolsets.cli list plus agent.disabled_toolsets to pin the surface. Nothing you skipped loads later — not even after hermes update. Re-enable anything with hermes too

NVIDIA Research 發布了 SpatialClaw，這是一個無需訓練的空間推理框架。該框架針對視覺語言模型（VLM）持續存在的弱點——難以判斷物體的位置、相對關係及在三維空間中的移動方式。SpatialClaw 不重新訓練模型，而是改變代理程式用來呼叫感知工具的動作介面。研究團隊指出，介面才是瓶頸所在。他們的解決方案是將程式碼作為動作介面。在 20 項基準測試中，SpatialClaw 達到平均 59.9% 的準確率，比近期發布的空間代理 SpaceTools 高出 11.2 個百分點。

{"id":"39ef5947-b77a-4904-bf03-ff6264f08dc4","object":"response","model":"deepseek-v4-flash","output":[],"stop_reason":"max_output_tokens","usage":{"input_tokens":154,"output_tokens":200,"total_tokens":354}}

Back to Articles MosaicLeaks: Can your research agent keep a secret? Enterprise Article Published June 18, 2026 Upvote - Alexander Gurung agurung Follow ServiceNow Rafael Pardinas rafapi-snow Follow ServiceNow TL;DR Deep research agents increasingly combine private local documents with external tools like web retrieval, creating a privacy risk: an agent's external queries may leak sensitive information. MosaicLeaks proposes a new deep-research task with multi-hop questions that interleave public and private information. Across the models we tested, agents frequently leaked private information, and training only for task performance made it worse. We propose a mosaic-leakage-aware RL training method, Privacy-Aware Deep Research (PA-DR), which raises strict chain success (the share of chains

這篇消息聚焦「騰訊老兵+大廠00後新銳，碼上飛想做的不只是AI Coding」。原始導語提到：已接入華為鴻蒙生態 從 AI 情報角度來看，這類內容值得關注其背後的技術進展、產品落地、產業競爭與後續市場影響。