Topic · 主题全部主题 →

Agent 智能体

让模型自主规划、调用工具、完成多步任务的技术方向——从 Claude Code、Manus 到各家 Agent 框架与评测基准的全部动态。

9,105条收录
1,107条精选

精选归档 · 第 48 页

第 941–960 条 · 共 1,107 条

3月26日

星期四 · 5 条
13:46
Artificial Analysis@ArtificialAnlys精选
OpenAI发布GPT-5.4 mini与nano轻量模型OpenAI released GPT-5.4 mini and nano, cheaper variants of GPT-5.4 with the same reasoning modes. GPT-5.4 nano is the standout, scoring ahead of both Claude Haiku 4.5 and Gemini 3.1 Flash-Lite Preview with lower per token pricing@OpenAI released GPT-5.4 mini (xhigh, 48) and nano (xhigh, 44), the first mini and nano updates since GPT-5. Both are multimodal with image input support and feature a 400K token context window. They support the same reasoning effort levels as GPT-5.4 (xhigh, high, medium, low, none) and are priced significantly lower: mini at $0.75/$4.50 per 1M input/output tokens and nano at $0.20/$1.25, compared to GPT-5.4 at $2.50/$15. We evaluated these models across three reasoning variants: xhigh, medium, none. While both models are more intelligent than their peers in the highest reasoning efforts, they are more verbose, using 200M+ output tokens to run the Intelligence Index, higher than even select frontier modelsKey benchmarking takeaways from the highest reasoning variants: ➤ GPT-5.4 nano (xhigh, 44) jumps 18 points from GPT-5 nano (high, 27), with improvements across all evaluations. Compared to Claude Haiku 4.5 (Reasoning, 37) and Gemini 3.1 Flash-Lite Preview (34), GPT-5.4 nano leads on τ²-Bench (81% vs 55% and 31%), IFBench (76% vs 54% and 77%), and TerminalBench (42% vs 27% and 24%) ➤ GPT-5.4 mini (xhigh, 48) gains 7 points over GPT-5 mini (high, 41), with gains across most evaluations. Compared to Gemini 3 Flash Preview (Reasoning, 46) and Claude Sonnet 4.6 (Adaptive Reasoning, max effort, 52), GPT-5.4 mini leads on TerminalBench (52% vs 39% and 53%) and CritPt (10% vs 9% and 3%) ➤ Both models perform less on AA-Omniscience compared to peers, driven primarily by high hallucination rates. GPT-5.4 mini scores -18.7 with a 90% hallucination rate, well behind Claude Sonnet 4.6 (Adaptive Reasoning, max effort, +12.4, 46% hallucination rate) and Gemini 3 Flash Preview (Reasoning, +11.6, 92% hallucination rate but 54% accuracy). GPT-5.4 nano scores -29.6 with a 74% hallucination rate, behind Claude Haiku 4.5 (Reasoning, -4.2, 26% hallucination rate) and Gemini 3.1 Flash-Lite Preview (-15.5, 82%). Both GPT-5.4 models attempt to answer far more questions than Claude Haiku 4.5 and Claude Sonnet 4.6 rather than abstaining, which drives the higher hallucination rates ➤ Both models show strong agentic performance. GPT-5.4 mini scores 1405 on GDPval-AA (Agentic Real-World Work Tasks), ahead of Gemini 3 Flash Preview (Reasoning, 1191) but behind Claude Sonnet 4.6 (Adaptive Reasoning, max effort, 1633). GPT-5.4 nano scores 1169, close to Claude Haiku 4.5 (Reasoning, 1173) and well ahead of Gemini 3.1 Flash-Lite Preview (944) ➤ Token usage with xhigh reasoning effort is higher for both models compared to peers with highest reasoning efforts. GPT-5.4 mini used 235M output tokens to run the Intelligence Index, ~3.4x GPT-5 mini (high, 69M) and more than Claude Sonnet 4.6 (Adaptive Reasoning, max effort, 198M) despite scoring 4 points lower. GPT-5.4 nano used 210M output tokens, ~2.4x Claude Haiku 4.5 (Reasoning, 87M) and ~4x Gemini 3.1 Flash-Lite Preview (53M) ➤ Effective cost to run the Intelligence Index reflects the higher token usage. GPT-5.4 mini (xhigh) cost ~$1,406, compared to ~$278 for Gemini 3 Flash Preview (Reasoning) and ~$3,959 for Claude Sonnet 4.6 (Adaptive Reasoning, max effort). GPT-5.4 nano (xhigh) cost ~$376, compared to ~$584 for Claude Haiku 4.5 (Reasoning) and ~$94 for Gemini 3.1 Flash-Lite Preview. GPT-5.4 nano is cheaper than Claude Haiku 4.5 on an effective cost basis despite using ~2.4x more tokens, due to its significantly lower pricing. Overall, GPT-5.4 nano is the standout offering a better Intelligence vs. Cost to Run Intelligence Index tradeoff than peers and GPT-5.4 mini译OpenAI发布GPT-5.4 mini与nano轻量模型，保留多档推理能力与400K上下文窗口，价格降至$0.20/$1.25每百万token。基准测试显示，GPT-5.4 nano在τ2-Bench等多项测试中领先Claude Haiku 4.5与Gemini 3.1 Flash-Lite Preview，但幻觉率较高且token消耗量大。得益于极低单价，nano在Intelligence Index测试中的有效成本反而低于竞品，展现出优秀的性价比优势。
智能体OpenAI推理模型发布
推荐理由：OpenAI 发布 GPT-5.4 mini/nano，性价比突出且评测数据详实，nano 性能超越同级竞品。
08:00
Cursor Blog精选
AI 评分 69/100
通过实时强化学习改进Composer编码模型

Cursor团队将实时强化学习技术应用于Composer编码模型，利用真实用户交互产生的推理令牌作为训练信号，以解决模拟环境与真实使用间的匹配问题。该技术使团队能够以每五小时一次的频率部署改进后的模型检查点。通过A/B测试，新版本实现了关键指标提升：代理编辑在代码库中的持久性增加2.28%，用户不满意后续减少3.13%，延迟降低10.3%。实时RL也带来了奖励黑客等新挑战，但真实用户反馈有助于识别和修正此类问题。

智能体现象/趋势编码
推荐理由：Cursor 把真实用户交互当训练信号，每五小时迭代一次 Composer，这不是论文是工程日志。做 coding agent 的团队该看看他们怎么处理 reward hacking 的两个真实案例，比任何 benchmark 论文都实在。
07:14
Anthropic@AnthropicAI精选
工程博客更新：我们如何设计 Claude Code auto modeNew on the Engineering Blog: How we designed Claude Code auto mode.Many Claude Code users let Claude work without permission prompts. Auto mode is a safer middle ground: we built and tested classifiers that make approval decisions instead.Read more: https://www.anthropic.com/engineering/claude-code-auto-mode译Anthropic 工程博客发文介绍 Claude Code auto mode 的设计思路。该模式作为无需人工确认和完全手动之间的安全中间方案，通过构建并测试分类器来替代人工审批决策。
智能体Anthropic产品更新编码
推荐理由：Anthropic 官方揭秘 Claude Code 自动模式的安全分类器设计机制
06:00
00:05
Andrej Karpathy@karpathy精选
所有 LLM 个性化功能的通病：记忆对模型来说过于干扰One common issue with personalization in all LLMs is how distracting memory seems to be for the models. A single question from 2 months ago about some topic can keep coming up as some kind of a deep interest of mine with undue mentions in perpetuity. Some kind of trying too hard.译LLM 的记忆功能存在过度联想问题：两个月前随口一问会被当作深度兴趣永久存档，并在后续对话中反复提及，显得过于刻意且造成不必要的干扰。
智能体大佬观点
推荐理由：Karpathy 指出 LLM 记忆功能过度关联历史对话的普遍痛点

3月25日

星期三 · 6 条
08:00
Google Developers Blog（RSS）精选
AI 评分 84/100
用 Agent 技能弥合知识鸿沟

Google DeepMind 开发出一项“Gemini API 开发者技能”，使智能体能够实时获取最新文档与 SDK 指导。评估结果显示，配备该技能后，gemini-3.1-pro-preview 模型的成功率从 28.2% 大幅跃升至 96.6%。这种轻量级方法通过赋予模型强大的推理能力并接入“事实来源”，有效解决了静态模型知识与快速演进的软件实践之间的脱节问题，显著消除了过时的编码模式。

智能体DeepMindGoogle产品更新
推荐理由：通过实时文档赋能模型，编码任务成功率飙升，开发者可借鉴优化AI工具。
08:00
OpenAI：官网动态（RSS · 排除企业/客户案例）精选
OpenAI 推出安全漏洞赏金计划

OpenAI 启动安全漏洞赏金计划，悬赏征集 AI 滥用及安全风险漏洞，涵盖智能体漏洞、提示注入攻击和数据泄露等问题。

智能体OpenAI安全/对齐
推荐理由：OpenAI推出安全漏洞赏金计划，聚焦Agent安全与提示词注入风险
05:26
Boris Cherny@bcherny精选
Claude Code 新增自动模式：告别权限确认👏no 👏 more 👏 permission prompts 👏译Claude Code 推出 auto mode，自动代为决定文件写入与 bash 命令的权限，无需逐条手动确认，也不必完全开放权限。每项操作执行前仍经 safeguards 安全检查。

Claude: New in Claude Code: auto mode. Instead of approving every file write and bash command, or skipping permissions entirely,...

智能体Anthropic产品更新编码
推荐理由：Claude Code 新增自动模式，Agent 无需逐条确认即可安全执行命令
02:01
Claude@claudeai精选
Claude Code 新增 auto modeNew in Claude Code: auto mode.Instead of approving every file write and bash command, or skipping permissions entirely, auto mode lets Claude make permission decisions on your behalf.Safeguards check each action before it runs.译Claude Code 推出 auto mode，无需逐条批准文件写入和 bash 命令，由 AI 自动判断权限决策，运行前仍经安全检查验证。
智能体Anthropic产品更新编码
推荐理由：Claude Code 推出自动模式，AI 编程助手进入免确认新阶段
01:25
Jim Fan@DrJimFan精选
AI代理时代的安全噩梦：超越传统身份盗窃的威胁This is pure nightmare fuel. Identity theft of the past would be nothing compared to what vibe agents can do. Sending credentials is too obvious and for rookies. They could easily spread contaminations across ~/.claude, **/skills/*, or even just a PDF your agent visits periodically in /morning-brief. Your entire filesystem is the new distributed codebase. Every file that could go into context would add to the attack vector. Every text can be a base64 virus.In the new world of on-demand software, I try to minimize dependencies - people rarely need all the APIs supported in LiteLLM, might as well build a custom router with only what you need on the fly (which I did in one of my late-night claude sessions).Unfortunately, there is very little middleground between "pressing yes mindlessly for every edit" and "--dangerously-skip-permissions". There will be a full blooming industry for "de-vibing": dampening the slop and putting guardrails/accountability around agentic frameworks. They are the boring old, audited Software 1.0 that watches over the rebellious adolescents of Software 3.0.Claws need shells. Probably many layers of nested shells.译vibe agents带来远超传统身份盗窃的安全威胁，整个文件系统成为分布式攻击面，~/.claude、skills目录乃至PDF都可能被base64病毒污染。LiteLLM 1.82.8被入侵事件显示恶意代码可窃取凭证并自我复制。当前代理框架面临权限管理困境，只能在盲目授权与完全跳过间选择。未来需"de-vibing"行业，用经审计的Software 1.0为Software 3.0建立多层安全护栏。

Daniel Hnyk: LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromise...

智能体安全/对齐编码
另有 1 家信源报道X：Andrej Karpathy (@karpathy)
推荐理由：Jim Fan警示Agent时代新型供应链攻击风险，以LiteLLM被黑事件为例揭示文件系统污染威胁
00:31
Anthropic@AnthropicAI精选
Anthropic 工程博客：我们如何利用多智能体工具提升 Claude 的前端设计与长时自主软件工程能力New on the Anthropic Engineering Blog:How we use a multi-agent harness to push Claude further in frontend design and long-running autonomous software engineering.Read more: https://www.anthropic.com/engineering/harness-design-long-running-apps译Anthropic 工程博客发文介绍多智能体工具（multi-agent harness），用于增强 Claude 在前端设计和长时间自主软件工程任务中的表现。该工具帮助 Claude 更好地处理复杂 UI 设计与长期运行的编程工作。
智能体Anthropic教程/实践编码
推荐理由：Anthropic官方揭秘多智能体框架，推动Claude前端设计与长时自主编程

3月24日

星期二 · 7 条
21:59
Thariq@trq212精选
Figma MCP 今日更新，成为与 Claude Code 的最强集成之一Today’s Figma MCP update makes it one of the strongest integrations with Claude Code I’ve seen.You can now use Claude Code to design in Figma with the the full context of your design systems.译Figma 发布 MCP 更新，开放测试版已上线。新增 use_figma 工具支持 AI 代理基于完整设计系统上下文直接在画布上设计，与 Claude Code 深度集成，可直接通过 Claude Code 在 Figma 中完成设计工作。

Figma: Now you can use AI agents to design directly on the Figma canvas, with our new use_figma MCP tool and skills to teach th...

智能体AnthropicMCP/工具产品更新
推荐理由：Figma 官方接入 MCP 协议，Claude Code 可直接在画布上进行 AI 设计。
17:00
OpenAI：官网动态（RSS · 排除企业/客户案例）精选
ChatGPT 推出产品发现功能

ChatGPT 上线基于 Agentic Commerce Protocol 的全新购物功能，提供更丰富的视觉化沉浸式体验，支持商品发现、并排对比及商家集成。

智能体OpenAI产品更新搜索
推荐理由：ChatGPT 上线 Agentic 购物功能，支持商品智能对比与商家集成
11:38
Boris Cherny@bcherny精选
Anthropic Labs小团队发布Claude电脑控制功能Little known fact, the Anthropic Labs team (the team I joined Anthropic to be on) shipped:• MCP • Skills • Claude Desktop app • Claude CodeIt was just a few of us, shipping fast, trying to keep pace with what the model was capable of.Those early Desktop computer use prototypes, back in the Sonnet 3.6 days, felt clunky and slow. But it was easy to squint and imagine all the ways people might use it once it got really good.Fast forward to today. I am so excited to release full computer use in Cowork and Dispatch. Really excited to see what you do with it!译Anthropic Labs团队规模虽小但迭代迅速，先后发布MCP、Skills、Claude Desktop app及Claude Code。从Sonnet 3.6时代笨拙的Computer Use原型，到如今正式在Claude Cowork和Claude Code中开放完整功能。Claude现可操控电脑完成打开应用、浏览网页、填写表格等任务。目前处于研究预览阶段，仅支持macOS。

Claude: You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in s...

智能体Anthropic产品更新编码
推荐理由：Claude 开放 Computer Use 研究预览，支持在 Cowork 和 Code 中操控电脑完成任务
10:01
Hugging Face：Blog（RSS）精选
AI 评分 78/100
全新语音智能体评估框架EVA发布

ServiceNow AI团队在Hugging Face上发布了语音智能体评估框架EVA。该框架通过标准化测试集与多模态指标，系统评估语音助手在对话理解、任务完成及交互自然度等方面的性能，旨在量化衡量智能体在复杂真实场景下的表现，助力研究人员客观比较不同模型，推动技术优化。

智能体Hugging Face开源/仓库语音
推荐理由：提供语音代理评估标准，帮助开发者优化模型性能和测试效率。
05:53
Thariq@trq212精选
Claude Code 现在可以控制你的电脑了！Claude Code can control your computer now!译Claude Code 推出电脑控制功能，支持鼠标、键盘和屏幕操作，可操控任意应用。配合 Dispatch 使用还能实现远程控制，在用户离开电脑时继续通过 Claude 操作设备。

Felix Rieseberg: Today, we’re releasing a feature that allows Claude to control your computer: Mouse, keyboard, and screen, giving it the...

智能体Anthropic产品更新编码
推荐理由：Claude Code 新增完整计算机控制能力，可操控鼠标键盘及任意应用，支持远程使用
05:38
Claude@claudeai精选
Claude 现已支持控制电脑完成任务You can now enable Claude to use your computer to complete tasks.It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.Research preview in Claude Cowork and Claude Code, macOS only.译Claude 推出电脑控制功能，可自动打开应用、操作浏览器、填写表格等，替代用户完成桌面操作。该功能目前处于研究预览阶段，已在 Claude Cowork 和 Claude Code 上线，仅支持 macOS 系统。
智能体Anthropic产品更新编码
推荐理由：Claude开放Computer Use功能，可操控电脑完成办公任务（macOS）
00:00
Anthropic：Engineering（事故复盘 + 工程实践 · 网页）精选
AI 评分 75/100
利用对抗网络灵感设计多代理架构，突破长时应用开发瓶颈

作者受生成对抗网络启发，设计了一个包含规划器、生成器和评估器的三代理架构，以解决Claude在长时应用开发中的两大瓶颈。该架构通过上下文重置机制，有效克服了模型在长任务中的“上下文焦虑”问题；同时，通过分离生成与评估功能，使代理能依据具体标准进行迭代改进，而非盲目自评。这一方法成功使系统能在多小时的自主运行中生成完整的全栈应用程序，突破了此前提示工程和传统工具设计的性能上限。

智能体Anthropic教程/实践编码
推荐理由：Anthropic 工程师把 GAN 的 generator-evaluator 思路搬进长时 Agent 架构，从设计到全栈编码都跑通了，还附了成本和失败模式。做 Agent 产品的人读完能直接抄作业，比看十篇论文管用。

3月23日

星期一 · 2 条
08:00
Google Developers Blog（RSS）精选
AI 评分 71/100
使用 LlamaParse 与 Gemini 3.1 构建智能金融助手

结合 LlamaParse 与 Gemini 3.1 模型，可从复杂的非结构化文档中提取高质量数据。该方案采用事件驱动架构，利用 Gemini 3.1 Pro 对密集的金融表格进行智能解析，并使用 Gemini 3.1 Flash 进行高性价比的摘要生成。开发者通过此教程可构建个人财务助手，将杂乱的经纪账户对账单转化为结构清晰、易于理解的分析报告。

智能体Google检索增强教程/实践
推荐理由：开发者可快速上手构建财务AI助手，将杂乱数据转化为结构化见解。
00:00
Anthropic：Research（发表成果 · 网页）精选
AI 评分 71/100
利用长时运行智能体工作流革新科学计算

Anthropic 研究员展示了如何将多日智能体编码工作流应用于科学计算任务。以使用 Claude Opus 实现宇宙学玻尔兹曼求解器的可微分版本为例，该任务通常需耗费研究人员数月甚至数年时间。通过制定清晰的项目指令、利用日志文件作为智能体的持久记忆并设置测试预言，即使是非领域专家也能引导智能体在数小时内完成这类复杂项目。该方法的核心在于设定高层目标后，让智能体团队自主工作，仅需偶尔人工监督，从而显著提升了科学代码开发与移植的效率。

智能体Anthropic教程/实践编码
推荐理由：Anthropic 研究员用 Claude Opus 4.6 花几天从零写出了一个宇宙学 Boltzmann 求解器，原本是博士级团队几个月的活。这不是论文，是一份完整的多日 Agent 工作流实操手册，做科研或长周期编码的人可以直接抄作业。
MCP 与工具调用AI 编码推理能力