27 Matching Annotations
  1. Last 7 days
    1. it spent a day and a half coordinating about 60 Claude subagents, which this time went much deeper

      60个子agent协同工作1.5天——典型的multi-agent研究系统:主agent分配任务,子agent分别攻克子问题,互相验证结果。这是agentic workflow重塑数学研究的具体案例,也是Claude Code真正被用于科研的里程碑。

  2. Aug 2026
    1. Ona’s customer-controlled execution model will allow agents to operate inside an organization’s own cloud environment while OpenAI provides the intelligence and orchestration that power the experience.

      这句话划出了一条关键的架构分界线:"智能与编排"由 OpenAI 提供,"执行环境的控制权"留在客户自己的云里。这正是 EP.97 专题01 架构图里"安全网关/本体"层要解决的问题——企业愿意把工作交给 Agent 云端持续执行的前提,是自己仍然掌握基础设施、数据和安全边界。

    2. We believe people should be able to delegate more ambitious work without remaining tied to the machine where it began.

      这句话几乎就是 EP.97 专题01 提出的"设备解耦"设计公理的官方原话版本——OpenAI 明确把"任务不再绑定发起它的那台设备"当作 Codex 下一阶段的核心设计目标,而收购 Ona 正是为了补齐这一目标所需的持久化云端执行基础设施。

    1. the need to specify goals, constraints, context, and evaluation did not disappear

      Lilian Weng 用 prompt engineering 的历史类比预测 harness 工程的走向:手工技巧会被模型能力提升逐渐内化,但"目标/约束/上下文/评估该如何被清晰表达"这个需求本身不会消失,只会转移到更高的抽象层。这是判断 Agent 设计下一步会往哪走的一条重要经验规律,也支撑了 EP.97 专题01 对"完整形态"的预测:接口会更简单,但背后的工程复杂度不会归零。

    2. the system surrounding a base model that orchestrates execution and decides how the model thinks and plans, calls tools and acts, perceives and manages context, stores artifacts, and evaluates results.

      这是"harness"这个概念在本篇最精确的定义——不是模型本身,而是包裹模型的执行系统。EP.97 专题01(Cowork Agent 完整形态)的架构图直接建立在这个定义上:设备解耦、检查点可恢复、多端可观测这些设计公理,本质上都是在给"harness 层"而不是"模型层"做工程。

    1. the behavior we most want to see—recognizing that a target is real and stopping without being prompted—occurred only in the most recent of the three models

      三个模型(Opus 4.7 / Mythos 5 / 内部研究模型)面对同一处境——都在某个时刻认出打的是真实系统——却做出三种不同反应:继续攻击、合理化后继续攻击、主动停止。这句话是 Anthropic 给出的唯一"进步信号",但措辞极其克制。EP.97 故事线 A 用这组对比说明"识别风险后是否主动停手"正在成为新一代 Agent 安全的关键分野,而不再只是"能不能做到某件事"。

  3. Jul 2026
    1. no single architecture dominates; rather, effectiveness depends on aligning the memory structure with the specific workload bottleneck

      对智能体记忆系统的批判性审视。当前业界没有一刀切的完美架构,记忆模块的设计必须与具体的任务瓶颈相匹配。这打破了“通用记忆系统”的幻想,提示我们在构建 Agent 时需要针对局部维护成本和任务特征进行定制化设计。

    2. we need arenas where models reveal themselves under pressure, with imperfect information, feedback loops, and consequences.

      反直觉的观点:传统的静态排行榜可能正在失效。在复杂环境中,模型的智能应该体现为可执行的策略而非单纯的文本回答。将 AI 评测转化为类似足球比赛的高压动态博弈,揭示了未来评测体系向“后果驱动”和“多智能体交互”演进的趋势。

  4. Jun 2026
    1. Strict No LLM / No AI PolicyNo LLM-generated content, whether it be code or prose.No paraphrasing LLM-generated content.No LLMs for editing, including fixing spelling or grammatical errors.No LLMs for translation. English is encouraged, but not required. You are welcome to post in your native language and rely on others to have their own translation tools of choice to interpret your words.No LLMs for brainstorming and then sharing the results of that brainstorming, even if you create the prose. If you use a chatbot to give you advice on a comment on the issue tracker, that comment is unwelcome.No LLMs for finding bugs.

      Seems kind of extreme. But https://www.youtube.com/watch?v=pkndFYSTr0Y gives some more context (an interview) that kind of explains their stance (limited maintainer time/attention; education).

    1. Agents address the problem from independent angles, other agents try to refute what they found, and the run keeps iterating until the answers converge—which is how a workflow reaches results a single pass can't.

      Convergence through adversarial iteration is borrowed from ensemble methods and scientific peer review — but applied to code. The non-obvious implication: this architecture is more robust to the hallucination problem than single-pass generation, because refuting agents are specifically incentivized to find failures. It's a form of AI quality control built into the workflow itself.

    1. If Nvidia has cracked the code on bringing AI agents easily, safely, and usefully to the masses, it could — and should — be big.

      大多数人认为AI代理技术仍处于早期阶段,难以在消费级设备上有效运行,但作者暗示Nvidia已经解决了这一技术难题。这一乐观观点挑战了当前AI代理技术仍不成熟的行业共识,暗示市场可能即将迎来AI代理的大规模普及。

  5. May 2026
    1. The quote is a big reversal of stance from a position ~uniformly held by anyone who worked at **Team Big Model**, including his previous head of OpenAI Labs

      大多数人认为大型模型实验室会继续专注于基础模型研发,但作者认为这是一个立场的重大转变,因为连OpenAI前高管都开始转向代理产品。这挑战了AI行业长期以来的'模型优先'共识,表明即使是Big Model团队也开始认可代理产品的价值。

    1. In one case [first reported by the Financial Times](https://www.ft.com/content/00c282de-ed14-4acd-a948-bc8d6bdb339d?syn-25a6b1a6=1), an Amazon Web Service agent called Kiro purportedly decided the best way to upgrade a particular software service was to delete the whole thing and start over — and was able to do so without asking for human permission

      这个案例突显了AI代理可能带来的风险,需要深入了解如何防范这类事件的发生。

  6. Apr 2026
    1. The architecture scales horizontally to 300 sub-agents executing across 4,000 coordinated steps simultaneously, a substantial expansion from K2.5's 100 sub-agents and 1,500 steps.

      大多数人认为AI系统的扩展主要依赖于增加单个模型的计算能力和参数规模,而非增加智能体的数量。作者提出的300个智能体并行执行的模式挑战了这一认知,暗示未来AI发展可能更侧重于'多智能体协作'而非'单一模型增强',这可能会重新定义AI系统的架构设计原则。

    1. And it’s not just office work. Multi-agent tools like Google DeepMind’s Co-Scientist let researchers use teams of AI agents to coordinate literature searches, generate and test hypotheses, design experiments, and more.

      大多数人可能认为人工智能在办公室工作中的应用仅限于数据处理,但作者提出,多智能体工具甚至可以用于研究工作,如文献搜索和实验设计。

    1. Lightweight Agent Detection & Response (ADR) layer for AI agents — guards commands, files, and web requests.

      这个项目定义了一个新的'ADR'(Agent Detection & Response)层概念,这标志着AI安全领域的一个重要演进。从传统的端点保护转向专门针对AI代理的轻量级防护,反映了安全行业对AI特定威胁模式的适应和专业化。

    1. Meta also explicitly highlighted parallel multi-agent inference as a way to improve performance at similar latency

      令人惊讶的是,Meta明确强调了并行多代理推理作为在相似延迟下提高性能的方法。这表明AI系统正在从单一模型向多代理系统演进,可能是解决复杂问题的新范式,同时也暗示了未来AI系统架构的重大转变。

    1. scaling Muse Spark with multi-agent thinking enables superior performance with comparable latency.

      令人惊讶的是:通过扩展并行智能体的数量而非延长单个智能体的思考时间,Muse Spark能够在保持相近延迟的同时实现更优性能。这种多智能体协调的推理方式挑战了传统AI模型通过增加计算时间提高性能的范式,为高效推理提供了新思路。

    1. Build autonomous agents that plan, navigate apps, and complete tasks on your behalf, with native support for function calling.

      一个能在手机上离线运行的 2B 模型,原生支持 Function Calling 和多步 Agent 规划——这意味着完全本地化的 AI Agent 在消费级硬件上正式成为现实。结合 Android Studio 的 Agent Mode 支持,AI Agent 从云端走向终端的时间点,可能比所有人预计的都要早。

    1. computer-use agents extend language models from text generation to persistent action over tools, files, and execution environments

      作者暗示,从文本生成扩展到持久性工具使用是AI安全范式的一个根本转变,这一转变带来的安全挑战被当前研究低估。这挑战了将语言模型安全方法直接应用于代理系统的主流做法,提出了需要专门针对代理行为的安全评估框架。

    1. AI Agent 可以通过标准 MCP 协议直接读取和操作 𝕏 平台:搜索推文、发帖、查看用户信息、管理书签、收发私信等。

      大多数人认为社交媒体平台会严格限制第三方自动化操作以防止滥用,但作者指出xAI全面开放了MCP协议支持,允许AI Agent直接执行各种操作,这与主流平台的封闭趋势形成鲜明对比。

    1. An agent cannot be held accountable. I think about this principle most. The instinct to put a human in the loop is understandable, but taken literally, it can mean a person approving every step before anything moves forward. The human becomes a bottleneck, rubber-stamping work rather than directing it, and you lose much of what makes agents valuable in the first place.

      大多数人认为在AI系统中加入人类审批环节是确保问责制的必要措施,但作者认为这会使人类成为瓶颈,削弱代理的价值。这一观点挑战了AI安全与问责的主流思维,提出了一个非传统的责任分配模式。

  7. Jul 2025
  8. Dec 2024
    1. Historically, AI was a tool

      for - quote - AI: from tool b to agent - Roman Yampolskiy

      quote - AI: from tool b to agent - Roman Yampolskiy - (see below)

      • Historically, AI was a tool, like any other technology. Whether it was good or bad was up to the user of that tool.
      • You can use a hammer to build a house or kill someone.
      • The hammer is not in any way making decisions about it.
      • With advanced AI, we are switching the paradigm
        • **from tools
        • to agents**.
      • The software becomes capable of making its own decisions, working independently, learning, self-improving, modifying.
      • How do we stay in control?
      • How do we make sure the tool doesn’t become an agent that does something we don’t agree with or don’t support?
      • Maybe something against us
  9. Jun 2024
  10. Sep 2023
    1. the Bodhisattva vow can be seen as a method for control that is in alignment with, and informed by, the understanding that singular and enduring control agents do not actually exist. To see that, it is useful to consider what it might be like to have the freedom to control what thought one had next.
      • for: quote, quote - Michael Levin, quote - self as control agent, self - control agent, example, example - control agent - imperfection, spontaneous thought, spontaneous action, creativity - spontaneity
      • quote: Michael Levin

        • the Bodhisattva vow can be seen as a method for control that is in alignment with, and informed by, the understanding that singular and enduring control agents do not actually exist.
      • comment

        • adjacency between
          • nondual awareness
          • self-construct
          • self is illusion
          • singular, solid, enduring control agent
        • adjacency statement
          • nondual awareness is the deep insight that there is no solid, singular, enduring control agent.
          • creativity is unpredictable and spontaneous and would not be possible if there were perfect control
      • example - control agent - imperfection: start - the unpredictability of the realtime emergence of our next exact thought or action is a good example of this
      • example - control agent - imperfection: end

      • triggered insight: not only are thoughts and actions random, but dreams as well

        • I dreamt the night after this about something related to this paper (cannot remember what it is now!)
        • Obviously, I had no clue the idea in this paper would end up exactly as it did in next night's dream!
  11. Jul 2022