本期共 6 条 AI 资讯，涵盖实用技巧 2 条、行业资讯 3 条、项目推荐 1 条。推荐关注：美团tabbit国际版免费接入GPT-5.5/Claude Opus 4.8等旗舰模型；NSA局长：Mythos数小时内攻破其几乎所有机密系统；三星电子向全球员工部署ChatGPT Enterprise和Codex。
美团做AI入口的野心藏不住了，tabbit国际版一口吞下三家旗舰模型还全免费，对不想开多订阅的用户是实用降本方案，趁未收费值得试试。
我去，GPT 5.5， Claude Opus 4.8， Gemini 3.5flash，最新旗舰版全接入，完全免费不用单独开订阅！
美团最近悄悄上了一个叫 tabbit 国际版的应用，里面接了好几家顶级模型，Claude、Gemini、GPT，都是最新旗舰版。
国内这边也没落下，Kimi、GLM、MiniMax 都在。
唯一的坑：一定要下国际版，
国内版只有国内模型，御三家不接入，
这东西的逻辑很简单，美团想做 AI 入口，先砸钱把模型阵容堆满，把用户量拉起来，
现阶段一分钱不花就能同时用几家最贵的模型，之前不想在各家之间来回切，又不想开一堆订阅的人，这个方案挺省心的，
趁现在还没收费，先用着，谢谢兄弟@fengdu2077挖掘到这么好的工具！亲测有效！
### 引用推文
> 风渡fengdu：分享一个免费用 Claude 、Gemini、GPT的方法,都是可用他们的最新旗舰模型。 除了御三家,还有国内的明星模型:Kimi-2.6、GLM-5.1、MiniMax-M3 只要下载一个美团的「tabbit 国际版」,记住要用国际版的才有御三家。 用国内版的只有国内的模型。
前NSA局长说Mythos几小时内攻破军方系统，5天破解MacOS，把零日漏洞发现速度从月压到天，安全行业的威胁模型要重写，而且背后是AI自动化攻击的可怕威力。
来源：X：AI Safety Memes (@AISafetyMemes) 正文： 美国国家安全局（NSA）的负责人（！）表示，Mythos“几乎攻破了我们所有机密系统，不是几周，而是几小时。”
这怎么就不是全世界最大的新闻呢？
[引用 @AISafetyMemes]：Mythos 在 5 天内攻破了 MacOS
为什么这很重要：
### 引用推文
> AI Notkilleveryoneism Memes ⏸️：Mythos cracked MacOS in 5 days WHY THIS MATTERS: - It takes Google Project Zero - the most prestigious bug-finding team in the world - ***6 months*** per zero-d...
Samsung Electronics brings ChatGPT and Codex to employees | OpenAI
June 21, 2026
Company
Samsung Electronics brings ChatGPT and Codex to employees
The deployment is one of our largest to date.
ChatGPT Enterprise and Codex available to all Samsung Electronics employees in Korea and all Device eXperience (DX) employees worldwide
Samsung’s global deployment is one of OpenAI’s largest enterprise launches ever
Samsung Electronics to use ChatGPT and Codex across its operations, from R&D and manufacturing to marketing, corporate functions, and other areas of its business
Samsung Electronics is deploying ChatGPT Enterprise and Codex to its employees around the world as it accelerates AI adoption across the company. Under the agreement, ChatGPT and Codex will be made available to all Samsung Electronics employees in Korea, and all employees worldwide in its Device eXperience (DX) division. This represents one of OpenAI’s largest enterprise deployments to date.
Samsung Electronics plans to use ChatGPT and Codex for technical and non-technical work, across a broad range of functions, including software development, marketing, product development, and manufacturing, to enhance employee productivity and problem-solving capabilities. Codex started as a tool for software development, but it’s increasingly useful for more kinds of work.
With ChatGPT, Samsung Electronics employees perform a wide range of knowledge-based tasks more efficiently, including searching for and analyzing information, drafting documents, developing ideas, and interpreting data. ChatGPT Enterprise provides enterprise-grade capabilities that enable organizations to use AI securely and effectively, including data protection, user and access management, and security controls. These capabilities allow Samsung Electronics employees to use advanced AI within the company’s security policies and governance framework.
Codex can enhance not only developer productivity through tasks such as writing, reviewing, and debugging code, but also the productivity of non-technical teams in their day-to-day work. For example, employees can use Codex to turn ideas into working software, internal tools, websites, and automated workflows.
More than 5 million people now use Codex every week for technical and non-technical workflows and roles. Codex weekly active users in Korea have grown nearly 800% since February 1, 2026.
“This historic deployment for OpenAI is particularly significant because Samsung Electronics, a global leader in technology and manufacturing, is embracing AI not as a tool limited to certain teams or functions, but as a core platform for improving how employees around the world work and innovate,” said Harrison Kim, General Manager of OpenAI Korea.
“OpenAI will work closely with Samsung Electronics to help its employees level up with ChatGPT and Codex to move from ideas to execution faster, solve complex problems, and create new products and services.”
OpenAI and Samsung Electronics previously began collaborating in the field of global AI infrastructure. Samsung Electronics is working with OpenAI to supply advanced memory semiconductors required for next-generation AI infrastructure. With Samsung Electronics’ adoption of ChatGPT Enterprise, the relationship between the two companies is expanding beyond AI infrastructure to encompass workforce transformation and company-wide AI adoption.
Also in Korea, Seoul National University recently began providing ChatGPT Edu to all 47,000 members of its community, including students, faculty, and staff, as part of its transition toward becoming an AI-native campus. OpenAI has also worked with Kakao to enable people to ask ChatGPT questions and receive responses directly within KakaoTalk group chats, bringing ChatGPT into Korea’s widely used everyday messaging platform.
Companies across a wide range of industries in Korea, including LG Electronics, LG Uplus, LG CNS, GS E&C, Samsung SDS, TVING, Krafton, Toss, MUSINSA, Korea Zinc, Nexen Tire, and HanaTour are also using ChatGPT Enterprise, OpenAI APIs, and Codex.
2026
元宝的父亲节营销，但合影生成指令写得具体可复现，比普通AI写真教程更接地气，父亲节想整活的可以直接抄作业。
公众号正文需在微信内阅读，站内仅提供摘要。
微软做起了跨中美模型的双向经销商，这事比表面看起来意义大——它可能绕过限制，把最先进的模型分发到原本接触不到的市场，全球 AI 供应链要改写了。
还得是微软会闷声发大财啊，它现在已经成为全球最大的 AI 中间商，oh no ，是最大的中转站，
不仅把ChatGPT卖给中国企业，也把DeepSeek 反向卖给西方客户😁
以上来自彭博社的最新报道，绝对权威可信，看完让我意外的不仅仅是微软把 GPT 卖给中国那部分，后面那一句：微软同时在测试 DeepSeek-R1 和 DeepSeek-V4，准备把这些中国模型卖给西方客户。
好家伙，左手接 GPT 卖到中国，右手接 DeepSeek 卖到西方，这他么不是一家 AI 公司在卖模型，简直就是一个跨中美 AI 模型的双向贸易网络正在成形呀
海里的这份 Deep Agents 教程把 LangChain 生态下的 Agent 开发梳理成了清晰的三层架构，虚拟文件系统、子 Agent、Skills 这些实战点讲得透，正在做协作型 Agent 的开发者终于有个成体系的免费入门路径了。
开源技术教程「Deep Agents 实战」，LangChain 官方认证大使 @zhanghaili0610 出品，他也是「LangChain 实战」「LangGraph 实战」的作者
https://github.com/datawhalechina/deepagents-in-action
教程的核心是：基于 LangChain / LangGraph 生态，面向开发者讲解如何"用好" Deep Agents 这个 Harness 框架来构建真实应用。
核心思想：Agent 开发的"三层架构"
1. Runtime（运行时）：LangGraph，持久化执行、断点恢复、流式输出、人机协作
2. Framework（框架）：LangChain，模型抽象、工具接口、Agent 循环、中间件
3. Harness（套件）：Deep Agents ←本课主角，预置文件系统、任务规划、子 Agent、长期记忆
技术内核：上下文工程
Deep Agents 做法：引入虚拟文件系统，让 Agent 像人类一样工作：
· 需要时才 read_file 按需读取
· 中间结果 write_file 落盘
· 大文件用 offset/limit 局部读取
· 上下文里只保留当前步骤真正需要的信息
· 这个文件系统还是可插拔的--内存、本地磁盘、数据库、远程沙箱、甚至混合路由，都可作为后端。
章节结构（8 章 + 2 准备篇）
· 准备篇 ── AgentSeek 环境搭建、开发技能安装
· 认知篇 ── ch01 三层架构 / ch02 5分钟快速上手
· 核心篇 ── ch03 虚拟文件系统 / ch04 任务规划 / ch05 子Agent / ch06 异步子Agent
· 进阶篇 ── ch07 Skills / ch08 长期记忆
· 规划中 ── Human-in-the-Loop、沙箱执行、流式前端、数据分析Agent、生产部署
四个核心能力的演进脉络值得注意：
· 虚拟文件系统（ch03）-- 六大工具：read_file / write_file / edit_file / ls / glob / grep
· 任务规划（ch04）-- write_todos 让 Agent 拆解并追踪复杂任务
· 子 Agent 委派（ch05-06）-- task 工具派发子任务，ch06 引入异步并行
· Skills 复用（ch07）-- 遵循开放的 Agent Skills 规范，编写的 Skill 可在 Claude Code、Cursor、Codex 等 30+ 工具中通用（"Skills 之于 AI Agent，就像 npm 包之于 Node.js"）
### 引用推文
> Harry Zhang：A gift to the Chinese developer community for anyone building agents on @LangChain: After 2 books on the ecosystem, I open-sourced the third - Deep Agents in Ac...