全部 AI 动态

斯坦福大学「AI 原生软件开发」第二周课程已公开 @mihail_eric 教授带来的这门「CS146S: The Modern Software Developer」已经进行到了第二周，主题是：高级上下文工程 https://themodernsoftware.dev/ 这周课程分为两节： 高级提示技巧 + RePPIT 框架与 SDD MCP 与工具调用全解：理论、搭建与工具设计 # RePPIT 框架：五步工作流 RePPIT (Research, Propose, Plan, Implement, Test) and spec-driven development 本周的方法论主体，Mihail 用一个自己 vibe code 出来的 Mint（个人记账应用）克隆做现场演示，为它添加"按类目显示消费趋势线"的功能。关键设定是：他自己也没读过这个代码库的代码，以此模拟接手陌生代码的真实场景。 五步各自要点： 1. Research（研究） — 只记录"现状是什么"，明确禁止提出改进建议。产出一份结构化的研究文档：代码库有哪些组件、行为契约是什么、关键代码位置在哪。这份文档有双重作用：既为后续步骤播种上下文（用压缩后的表示替代喂原始代码），也帮人自己建立足够的心理模型去鉴别智能体输出的好坏。 2. Propose（提议） — 基于研究文档，让智能体给出两个相互独立的方案（附权衡分析与开放问题）。为什么是两个？讲师用 PCA 打比方：他要的是"正交的主成分"，强制智能体探索解空间的两个不同方向，更多方案收益递减，剩下的判断空间留给人。人在这两步的参与度最高：作为理解产品需求与系统约束的专家做出选择，这是当前人类杠杆最大的位置。 3. Plan（计划） — 用模板化的设计文档填充：现状、功能性与非功能性需求、设计决策与权衡、技术设计（讲师特意要求带行号，让计划接近代码级别）、测试计划、可观测性、未来考虑。其中最有价值的一节是明确圈定范围外的事项（"不要碰这些，只做范围内的工作"）这是防止智能体跑偏做无关改动的重要护栏。 4. Implement（实现） — 提示词几乎就是一句"执行计划"。真正值得注意的变化是：现代智能体（如 Cursor）会隐式地打开浏览器、造测试数据、通过 CSV 接口上传来验证自己刚写的功能，实现与测试正在合并为一步。若你的智能体没有此能力，可加 Playwright 之类的 MCP server 补足。 5. Test（测试） — 分两层。行为契约（上传 CSV → 看到趋势线）由智能体的浏览器交互直接验证；性能、可读性、安全性等超出行为契约的部分交给 agent 驱动的代码审查（实践中常见四五个专门化智能体各管一域，安全、性能、风格，由协调智能体汇总成一份报告）。 # 上下文管理的三个实操问题 这是现场问答中含金量最高的部分（依据为 LangWatch 对 2,451 个 Claude Code 会话的实证研究）： 何时压缩（compact）？ 在 25 万–45 万 token 之间执行压缩，宁可偏晚也不要太早，压缩过早的代价远高于过晚。原始研究还发现上下文利用率随窗口膨胀急剧衰减：5 万以下时利用率 47.5%，60 万以上时只剩 2.5%；而压缩后 5 步内用户纠错率会升至基线的 2.37 倍（存在"压缩后迷雾期"），所以也不能无节制地压。 何时清空重开？ 演示中讲师做了一次他自认不妥的操作：选定方案后没清上下文，导致落选方案的 token 仍留在窗口里干扰执行。他的原则是倾向于清空，把关键信息写入文件，用压缩表示开新会话。另外不要从臃肿的父会话 fork 子智能体（有人因此一天烧掉 11 亿 token），应让子智能体从干净的小上下文起步。 模型怎么分配？ 前段（研究/提议/计划）用最强的模型，因为思考集中在这里；后段（实现/测试）可以用轻量模型。这是一种直接的工程化降本手段。 # MCP 与工具调用 第二次课覆盖 MCP 的理论、搭建与工具设计。配套阅读勾勒出三个层次： 1. 协议本身：server/client/tool/transport 的基本模型——让智能体以标准化方式调用外部能力。 2. 工程实践：Cloudflare 的企业级方案直面 MCP 落地的四大痛点，供应链风险（本地 server 依赖未审计软件）、授权失控（员工各装各的）、工具 schema 洪水（52 个工具约 9,400 token，Code Mode 把它们收拢成 2 个工具约 600 token，降 94%）以及影子 MCP 检测。其"门户 + 默认拒绝 + 审计日志"的架构是 MCP 企业化的一套参考答案。 3. 工具设计的"人机工学"：为智能体设计工具不是把 REST API 直接包一层——API 的组织逻辑服务人类浏览，而智能体需要的是低歧义、可组合、schema 精简的调用面。这与上一讲的"smart tokens"一脉相承：工具定义本身也占用上下文，也是要经营的成本。 # 生态对比：三条规格驱动路线 阅读材料中的 OpenSpec 与 Superpowers 正是 RePPIT 的工业化对照物，三者共享同一思想内核（先探索、再提案、再计划、后执行、终验证），但形态不同： RePPIT：讲师自用一年半打磨的个人技能集，轻量、可自由改造 OpenSpec（约 6.8 万 star）：以 /explore → /propose → /apply → /verify → /archive 命令固化为目录化的规格产物（proposal.md、specs/、design.md、tasks.md），强调规格随工作演进且团队与智能体对齐 Superpowers（约 30 万 star）：讲师称之为"我讲的框架的强化版"，一整套可组合的技能库（头脑风暴、写计划、执行计划、代码审查、TDD、系统性调试），并强制执行"测试先行、系统化而非临时起意、证据优先于声明"的纪律。讲师特别点出：这个 30 万 star 的仓库本质就是一组 SKILL.md 技能文件，这本身就是"工作流即代码"的绝佳例证。

Should you /grill-with-docs, /wayfinder, or just one shot the code? 1. If the diff is tiny, DON'T grill /grill-with-docs saves you from situations where the agent built the wrong thing. But if the thing it's building is tiny, that cost drops to nothing. So one-shotting is absolutely fine when you're only making small changes. 2. Never start with /wayfinder A failure mode of /wayfinder is that you start grilling and the solution turns out to be way simpler than you thought. Why is this bad? Because you've created the map and tickets, and you realise you don't need them. So if the expected diff is medium-large, always start with a /grill-with-docs session. Then, if planning starts to get out of hand, just say "/wayfinder let's turn this into a map".

Big news. Shaders is now open source 🔥 Use any of our 200+ WebGPU components in your work. Ship magic in a client site, a template, a product, a library, anything. Plus, you can export code for any framework, for free. Give it a star: https://github.com/shader-effects-inc/shaders

you now get Claude API credits with your MAX plans ($100, or 200 matching your plan) every month, use this to build more personal AI for yourself! I made an AI homepage that's generated everyday based on sites I read and replaces my 'new tab' page in Chrome

Emulation handhelds are becoming more and more popular. Being able to play all the old classic games and even some of the new ones on a handheld at the tip of your fingers. The main issue some people have is the set process can be difficult. You have to download emulators and mess with settings. Missing BIOS files. If you aren’t tech savvy these can be a pain in the ass. Luckily I have a solution. And that’s @bot I plugged my new handheld into my computer. I gave Grok bot instructions on how I wanted it set up. What it did: ✔️ Installed emulators for NES, SNES, N64, GameCube, Wii, Switch, Game Boy, GBA, DS, 3DS, Genesis, Dreamcast, PS1, PS2, PSP and Vita ✔️ Set up DS and 3DS games to use both screens ✔️ Picked the best graphics settings for every system ✔️ Installed a clean front end to launch everything ✔️ Sorted my games into the right folders with clean names ✔️ Pulled box art for every game automatically ✔️ Moved my old saves over from my last handheld ✔️ Made sure everything was on the newest version first Didn't touch a single settings menu. If you buy one of these new handhelds let @bot set it up for you.

Grok Bot can now search, read, and monitor X.

Opus5.5越来越好用了，做产品周更短片也好用。所有素材都是直出不需要接额外服务的。 这周GoodCase新上269个AI爆款案例和提示语，外加7个新 Skill。现在一共覆盖了1699 个案例、95 个 Skill、426 位创作者。 只玩AI视频的可以直接看awesome-seedance，现在加到了795条案例，367次提示语复测，27个提示语模板，还有77个能直接装的视频Skill👇

https://x.com/i/article/2107886964427182080

我的 grok bot 有了第一个强大的使用场景。 用 personal agent 帮你收集推特信息，比人类自己刷收手机、人肉收集、动不动沉迷要好太多了。

Grok Bot can now search, read, and monitor X.

three Effect agent frameworks worth checking out: - Fold by @0xblacklight - Yielded Agent by @danieljvdm - Tardigrade by @arjunkmrm particularly interesting to see they all support cloudflare durable objects as host and worth asking an agent to talk you through their durability approaches

CS 329Z: Engineering AI Agents Stanford / Fall 2026 @stanfordnlp https://cs329z.stanford.edu 课程定位：从"模型"到"系统"的工程学覆盖：简单 LLM 流水线 → 复合 AI 系统 → 自主 Agent。 由 @Diyi_Yang @michaelryan207 @jyangballin 三位老师主讲，现在前三讲讲义都已公开，之前是在 Google Drive，现在统一入口在斯坦福课程的官网了： https://web.stanford.edu/class/cs329z/slides 前三讲主题分别是： Intro to Agentic Systems LLMs for Builders RAG + Agents 理论和实践价值都很高！

Grok Bot can now search, read, and monitor X.

对于低于 100K token 的提示，价格为每百万 token $0.10 / $0.50，缓存读取为 $0.01。 超过 100K token 时，每百万 token $0.50 / $2.50，缓存读取为 $0.05。

Overmind turns anyone into an AI lab. • Builds a context graph from code and traces • Curates training and eval datasets • Evaluates prompts and models • Trains smaller, specialized models https://github.com/overmind-core/overmind

Prompt

I used Opus 5.5 to formally verify the Claude Agent SDK using Lean. A couple short prompts = 16 PRs fixing various bugs and race conditions. Video attached. TLA+ also works well. I sometimes combine Lean and TLA+ to look for issues around data flow, concurrency, and state mgmt. I don't know either language well, but Claude is excellent at both. This approach is super useful for formally modeling your code and finding bugs that a human probably wouldn't have spotted. Is formal verification the future of coding (or at least, bug finding)?

Your AI is vaporware without deep integrations with your customers’ systems of record. Introducing @WithAmpersand: integration infrastructure for enterprise agents. We power 11x, Orb, and Square's ability to take agentic action in systems of record like Salesforce, SAP, NetSuite, and Workday. Ask anyone serious about building AI and they’ll tell you: - The SaaSpocalypse didn’t happen. The world depends on CRMs, ERPs, HRISs, and ITSMs. - Your customers customized their deployments beyond recognition. - Systems of record companies are basically monopolies. They never had to make their APIs and MCPs user-friendly. - Docs don't explain half the weird edge cases you'll hit. - One bad write can blow up your pilot or renewal. - Once you finally get it working, someone changes a field and it breaks again. The world's data lives in structured databases. AI needs a translation layer to work with it. So, for your AI to truly transform the way enterprises work, you'll need deep integrations built for: - scoped permissions - bi-directional actions - real-time speed for agents - custom objects, fields, and workflows for each of your customers We built Ampersand to power the future of software. AI didn't trivialize writing integrations. It made them the critical path. I believe that deeply, so I didn't make a launch video about Ampersand. Instead, it's the best builders I know explaining just how big of a challenge this is.

I used to sit at my desk all day. Now I send off presentations from my sunroom using voice. ☀️