Tag
Droid 1.0 is an agent harness that enables agents to run hundreds of tool calls in a single model request; building it to support multiple models ensures improvements generalize rather than overfitting to one model's quirks.
Raindrop.ai 实验让来自四个不同模型的八个 AI 智能体在无指定领导的情况下协同处理模拟故障，发现仅有 5% 的实验涌现出真正能改变队友行为的“领导者”，且 Claude Sonnet 5.5 与 GPT-6.1 Sol 比 Gemini 3.1 Pro 更容易成为领导者，提示智能体编排中的细微配置（如模型选择、名称顺序）会显著影响群体协调。
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 Universe 2026 keynote recap covering the launch of the GitHub Copilot app and Copilot Fusion multi-model orchestration, framing GitHub as a comprehensive agentic software delivery system with new infrastructure investments.
微软宣布将源自 GitHub 的 Copilot harness 通过 Microsoft Foundry 向开发者和企业客户开放，提供托管式的多模型智能体编排能力，支持 20+ 前沿模型、MCP 与 agent skills 等开放标准。
GitHub Copilot Fusion 一般可用，它将 Project HydraFusion 的多模型编排与 Copilot 自动模型选择相结合，自动为每个任务路由和编排模型，兼顾性能与成本，并提供 Standard、High、Max 三档 effort 供开发者与管理员使用。
GitHub Universe 2026 发布了一系列围绕agentic software delivery的更新，包括GitHub Copilot app和Copilot Fusion多模型编排功能，旨在让开发者在AI代理时代保持对交付内容的控制。
Intrascope 2.0 推出，提供跨 120+ 模型的智能路由、可连接 1,650+ 应用的 AI agent，以及面向企业的预算、权限和使用分析管理功能。
A practitioner shares lessons from a year of running a multi-model router (drafter, critic, long-context searcher) on real work: role separation, verification gates, a provenance-checked lessons file, and small hot context survived, while agent personas, model voting, and long autonomous chains failed.
16-year-old Brazilian developer Kauan announces Tiny Agents, a Python framework for easily building AI agent harnesses with multi-provider model support, context-window management, async workers, tool execution, and a customizable terminal UI — demonstrating a Claude Code-like assistant in just 103 lines of code.
A user shares that Claude Opus 5.5 Medium paired with Claude Code delegation of easy tasks to Sonnet subagents delivers surprisingly fast and accurate results, with the Claude desktop app's memory management greatly improved over terminal use.
Anthropic shared a multi-model collaboration trick for Claude Code: use Sonnet to host the session and write code, Haiku to search files in parallel, and Opus as an advisor for important decisions — improving results while saving tons of tokens.
Vals AI argues that model routers are becoming a new primitive in the AI stack, showing that an oracle router would beat the best single model by 14 points on the Vals Index while costing about 40% less, as the frontier remains 'jagged' across benchmarks.
MoralityBench.ai 发布更新结果，使用经验证的人类道德心理量表对 16 个 LLM 配置各进行五轮 56 题测试，DeepSeek V4.1 Flash 平均偏离度最低（0.24），Llama 4 Maverick 答案一致性最高（93.9%），并新增 Muse Wonder、Gemini 4 Argon 和 GPT-6 Astra 的 10 月 14 日测试数据。
A thread sharing Anthropic's Claude Code 'advisor' trick: run Sonnet for coding, Haiku for parallel exploration, and Opus as an on-call advisor to get better results with far fewer tokens, along with a paste-ready configuration prompt.
Oh-My-Hermes 3.0.2 发布，新增 SkillEx 技能约束与 FenceGuard 沙箱安全机制，为 Hermes Agent 提供技能路由、长期记忆与跨 GPT、Claude、DeepSeek、Kimi、GLM 的并行沙箱编码代理能力。
作者分享了用 pi agent 通过 models.json 配置四个免费/低成本模型渠道（OpenRouter、百度灵积、阿里云 MaaS、opencode-go）的零成本 AI 编程方案，并给出了避免被静默回退到付费模型的配置踩坑经验，全文以推文串形式展开。
LumaBrowser, an all-in-one local AI assistant bundling LLM chat, image/music/voice models, agent scheduling, clustering, and IDE integration, is now open source and hosted on GitHub after a year in development.
Gemini Agent 是一款面向工作场景的通用 AI 代理，每个任务拥有独立的计算机环境，支持在 Gemini 与 Claude 模型之间自由选择，并具备子代理、多层记忆、主动式工作流、企业级治理与成本上限等功能。
Liquid Inference 是一款 LLM 路由器，让多家模型提供商针对每次 prompt 动态竞价，从而为用户提供最低推理成本，兼容 Claude Code、Cursor 等主流 agentic 编码工具，支持数百个开源与闭源模型。