Tag
Anthropic is open-betaling "message threads" for the Claude Messages API: the server stores conversation history, so each turn developers only send new messages plus the previous reply ID. This significantly cuts latency on long conversations (median TTFT for 500K–1M token requests drops by ~35%), with pricing and rate limits unchanged.
The tweet marvels at how many official MCP and Skill services are now available in China, enough to support a Personal Agent for daily life across scenarios such as commuting, dining, office work, and entertainment, and cites a post compiling a list of domestic MCP/Skill services.
Compiled a list of services in China that have made official MCPs or Skills available (such as Amap, Baidu Maps, Tencent Maps, DiDi, Meituan, Taobao, and Luckin Coffee), along with a detailed breakdown of Amap MCP Server’s capabilities, showcasing the ecosystem of external services that personal agents can call.
Peji is an AI chat product featuring a real browser that users can control, launched on Product Hunt.
The user shared a beat-synced ad Agent they adapted from Sac’s open-source match-cut Skill. Simply enter product information, and it automatically generates and edits a 20-second promotional video. They demonstrated it using the iPhone Duo. It’s ready to use once installed.
DeepSeekHarness-Android updated to v0.1.7, with 7 releases in less than a day. Full integration with 12 third-party model providers including OpenAI, Claude, and Gemini. New sessions default to relaxed permissions to reduce approval prompts. Also fixed image upload EACCES errors and layout issues on narrow screens.
GitHub Copilot Fusion 是一款能自动挑选并组合多个模型的 AI 编码代理，旨在为开发者提供更智能的代码辅助体验。
I built Agent WeChat, an open-source WeChat client with an OpenClaw plugin that lets an AI agent participate in your existing WeChat DMs and group chats using your own account, while you continue using your phone normally.
Developer jackwener has open-sourced Kaze, an AI browser for macOS. Its built-in agent executes tasks in a separate tab group, supports connecting external agents like Claude Code and Codex, and allows handing the browser over to your own agent via MCP.
NousResearch 推出 Hermes Desktop 的 ElevenLabs SFX 插件，支持直接生成音效、循环播放、一键生成变体及音效库管理，面向游戏开发场景。
OpenAI 发布 GPT-6.1 Terra，一款兼顾视觉推理、计算机操作与自然写作的高效模型，API 价格较 GPT-5.6 Terra 低 50%，在 Agents' Last Exam、MMMU-Pro、OSWorld 2.1 和 Terminal-Bench 4.0 等基准上大幅提升，同时显著降低幻觉与事实错误率。
Nanbeige released Nanbeige4.3-33B, an Apache 2.0 teacher model with a 512K context window intended to train smaller agent models like an upcoming 3B variant, reporting strong Terminal-Bench 2.1 results but mixed performance on SWE-Bench Pro.
LangChain announced an October 28th live event demoing its internal design agent, which has shipped 3,000+ assets since August and driven ~4.5M social views, serving as the company-wide design tool for decks, diagrams, ad creatives, and more.
Google open-sourced ARTEMIS, an Android device-operating agent that can be connected to tools like Claude Code, Codex, and Cursor via MCP. It supports automated clicking, swiping, cross-app task execution, and automated testing, achieving over 99% completion rate on the AndroidWorld benchmark.
dsh-context is a DeepSeek Harness plugin that provides full context lifecycle management for agents, including a context insights page, Context/Fleet dashboards, and a /context command — enabling cross-session visualization of usage, cost, and context composition.
An in-depth article on a high-frequency Agent interview question — "Walk me through your RAG pipeline" — systematically covering the principles, trade-offs, and common follow-up questions across the full RAG workflow: offline indexing (parsing, chunking, vectorization, index building) and online QA (retrieval, reranking, context construction).
Google has open-sourced ARTEMIS, enabling AI to perform natural language-driven cross-app automation testing on a real Android phone just like a human. It achieves a 99%+ task completion rate on the AndroidWorld benchmark and natively supports MCP, allowing IDEs like Claude Code, Windsurf, and Antigravity to drive devices directly.
NVIDIA 推出 Nemotron 3.5 Super VL 开放模型（120B 总参数/12B 激活），支持文本、图像、文档和视频理解及工具使用，Nous Research 宣布 Hermes Agent 是使用该模型及 Bolt 模型的最佳平台，Super VL 在 Hermes Bench 上得分 63.44，本地吞吐量最高提升 5 倍。
DeltaStreamer是一个开源智能体，通过观察帧间变化、构建结构化文本记忆并在需要时回溯视觉证据，实现类人化的连续流式视频理解，在OVO-Bench和StreamingBench上超越多种现有方法。
A developer argues that running Qwen3.8FN Q4 locally on a budget eBay-built server is cheaper than nearly all cloud models (except GPT-6 Luna and Claude Haiku 5.5 with caching), enabling longer-running agents that produce better results than cost-prohibitive frontier cloud models.