language models (LLMs) that – in effect – found freedom with the November 2022 launch of OpenAI’s ChatGPT with its extremely easy-to-use / speedy user interface. In addition, relatively new AI company founders making new tech advances increasingly more powerful, accessible, and economically viable. OpenAI’s ChatGPT – based on user / usage / monetization metrics – is history’s biggest ‘overnight’ success (nine years Internet 1.0 revolution – where technology started in the USA and steadily diffused globally – ChatGPT hit the world stage all at once, growing in most global regions simultaneously. Meanwhile, platform

2 AI 工具梳理 6 2.1 问答 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.1 ChatGPT . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.2 Claude . . . . . . . . . . . . . . . . . . . . 11 2.5.1 FlowGPT . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.5.2 ChatGPT 指令大全 . . . . . . . . . . . . . . . . . . . . . 11 2.5.3 SD 提示词手册 . . . . . . . . . . . . . . . 2.1 问答 2.1.1 ChatGPT ChatGPT 是一个由 OpenAI 开发的大型语言模型，它基于 GPT（Generative Pre-trained Transformer）架构。这种模型通过分析大量的文本数据来学习语 言结构和信息，使其能够生成连贯的文本、回答问题、撰写文章、进行对话等。 6 Figure 3: AI 问答工具 ChatGPT 经过特别训练，可以理解和生成人类语言，从而在多种应用场景中提

developing new models and capabilities. Our Applied Team turns those models into products, like ChatGPT Enterprise and our API. And our Deployment Team takes these products into companies to address closely with Legal, Compliance, and IT Security teams to ensure responsible use. They rolled out ChatGPT Enterprise globally, then let people discover their own use cases. “Normally, in a business like hours. The impact was felt across many disciplines and departments: The Credit Risk team Uses ChatGPT to determine creditworthiness faster and 
 more accurately. The Legal team Uses it to answer 40

我们先看大模型赛道的整体状况：  大模型是一项相对较新的技术。尽管 OpenAI 早在 2019 年就发布了第一个重要的 模型 GPT-2，但大模型的广受关注实际始于 2022 年 11 月发布的 ChatGPT。8 个月以后 Meta 就与微软合作发布了开源大模型 LLaMA-2。这个赛道的主要玩家在技术和商业化上有差距， 但没有到翻盘无望的程度。  大模型赛道不但包括模型的训练，也包括模型服务。训练是软件的制作成本，而服 进性”更多是用于公关宣传 的素材。考虑到数据获取、加工的成本，模型训练的成本，这是一种相当昂贵的宣传方式。 C 端用户指那些把大模型当成智能个人助理来使用的普通个人用户。OpenAI 在 ChatGPT 上一个重要且成功的操作就是把大模型从学术界、工业界直接推向了普通个体，让 C 端用户切 实感受到了大模型的可能性与魅力。这一点被国内的大模型厂商广泛学习。在 B 站刷视频，国 内知名的那 已从单一的辅助工具，逐渐演 变为软件开发人员不可或缺的助手或伙伴。 除了前面已介绍的 Cursor、Composio SWE-Kit、OpenHands CodeAct 等工具之外，国 内主要使用 chatGPT、GitHub copilot、通义灵码、CodeGeeX、文心快码、蚂蚁 CodeFuse 等编程工具，国外还出现一些受欢迎的、新的编程工具，如 Codeium IDE Cascade、Solver

53Linear search 54Principle #4: “Simplicity is the ultimate sophistication” 55 Source: ChatGPT56 Source: ChatGPT Principle #5: “Mechanical sympathy”I-Cache misses – likely/unlikely attributes 57 template Fan-out Type Support? PODsPrinciple #7: “Choose the right tool for the right task” 83 Source: ChatGPT84 FastQueue – Design Data available for readers mReadCounter mWriteCounter85 FastQueue – Design Results on AMD EPYC 9474FPrinciple #4: “Simplicity is the ultimate sophistication” 106 Source: ChatGPT107 Going further – Zero copy struct QProducer { void Write(std::span buffer); };108

specific code • Hand-optimize cache lines and struct layout 6How can we improve it? • We could ask ChatGPT… 7 That’s cool, but it won’t get us very far…How can we improve it? (second try) • At its core philosophy: 1. “Transform packages” 2. Compile time functional programming 3. Use modern C++ features as ChatGPT suggested • Now you should be able to write code with it. 27Let’s write some code… • Objectives

2025!3" Manus AI!Agent"#$ChatGPT%& #$% SAC NO. S0570519080006 | SFC NO. BQZ938 &'( SAC NO. S05701220801381 !"#$%&'() !"#$ • !"#$%&'()*AI+!"#$,-./012334%&'(56789:;<=>?@A BC%&'() • DEFGHI)*DEFGJKH ƒD‡5†[ˆGfigma> • 2022Eb,÷‹MonicauŒ>Monica!"#¶‰$•)€GAIŸ ,$ŒÜÝÞLMŽ•áâS),•ÌQŸ%ãR²cA+C•‘W O>Monica !"#. ChatGPT API áâ()G Chrome ßà,’L !"#$%Bloomberg*&'()2 Agent()9 Agent%;<=4 !"#$%B

12 Output the median after the samples are collected3ChatGPT 4ChatGPT 5Median #include #include #include auto median(std::vector& v) { std::ranges::sort(v);

talk: "C++ Modules: The Challenges of Implementing Header Units" From C++ Now 20239 Ask ChatGPT 202310 Ask ChatGPT 2024Simple Module 11 export module B; export void b() { }Built Module Interface (BMI)

R1形成了新的AGI定律，加速了AGI发展 Alpha Zero时刻 • Alpha Go采用监督学习， 使用人类棋谱训练 • Alpha Zero采用强化学习， 自己跟自己对弈 ChatGPT时刻 • OpenAI ChatGPT大模型， 通过预训练方式，实现涌 现，理解人类语言和知识 • 诞生预训练Scaling Law Deepseek-R1时刻 • o1采用强化学习，但训练复 杂推理能力，技术未公开