Gradient Labs 使用 GPT-4.1 和 GPT-5.4 mini / nano 驱动 AI agents，以低延迟和高可靠性自动化银行客服支持流程。
Gradient Labs uses GPT-4.1 and GPT-5.4 mini and nano to power AI agents that automate banking support workflows with low latency and high reliability.
最近的开源攻击主要聚焦在窃取 secrets。文章总结了你今天就能做的防护措施，以及 GitHub 正在推进的安全能力。原文最早发表于 The GitHub Blog。
Recent attacks on open source focus on exfiltrating secrets; here are the prevention steps you can take today, plus a look at the security capabilities GitHub is working on. The post Securing the open source supply chain across GitHub appeared first on The GitHub Blog.
Cloudflare 发布了 EmDash 的 beta，这是一套基于 Astro 6.0 的全栈 serverless JavaScript CMS。它结合了传统 CMS 的特性和现代安全能力，把插件运行在隔离的 Worker isolates 中。
2026-04-01Today we are launching the beta of EmDash, a full-stack serverless JavaScript CMS built on Astro 6.0. It combines the features of a traditional CMS with modern security, running plugins in sandboxed Worker isolates....Continue re…
/fleet 允许 Copilot CLI 并行派发多个 agents。文章介绍了如何写 prompt 来拆分跨文件工作、声明依赖关系并避开常见陷阱。原文最早发表于 The GitHub Blog。
/fleet lets Copilot CLI dispatch multiple agents in parallel. Learn how to write prompts that split work across files, declare dependencies, and avoid common pitfalls. The post Run multiple agents at once with /fleet in Copilot CLI appeared first on The GitHub Blog.
AI Search 新增了 wrangler ai-search 命令空间，可以在命令行中创建、列出、查看、更新、删除、搜索实例并查看使用统计。所有命令都支持 --json，方便脚本和 AI agents 直接解析。
AI Search supports a wrangler ai-search command namespace. Use it to manage instances from the command line. The following commands are available: CommandDescriptionwrangler ai-search createCreate a new instance with an interactive wizardwrangler ai-search listList all instances in your accountwrangler ai-search getGet details of a specific instancewrangler ai-search updateUpdate the configuration of an instancewrangler ai-search deleteDelete an instancewrangler ai-search searchRun a search query against an instancewrangler ai-search statsGet usage statistics for an instance The create command guides you through setup, choosing a name, source type (r2 or web), and data source. You can also pass all options as flags for non-interactive use: wrangler ai-search create my-instance --type r2 --source my-bucket Use wrangler ai-search search to query an instance directly from the CLI: wrangler ai-search search my-instance --query "how do I configure caching?" All commands support --json for structured output that scripts and AI agents can parse directly. For full usage details, refer to the Wrangler commands documentation.
8 年前，Cloudflare 推出了 1.1.1.1，目标是构建更快、更私密的互联网。现在他们公布了最新一次独立审查结果，强调其隐私保护承诺仍然有效。
2026-04-011.1.1.1DNSPrivacyConsumer ServicesTransparencyEight years ago, we launched 1.1.1.1 to build a faster, more private Internet. Today, we’re sharing the results of our latest independent examination. The result: our privacy protecti…
ADK SkillToolset 引入了一种“渐进式披露”架构，让 AI agents 可以按需加载领域知识，相比传统单体 prompt 最高可减少 90% token 消耗。它通过从简单 checklist 到 skill factory 的四种模式，让 agent 在运行时动态扩展能力。
The Agent Development Kit (ADK) SkillToolset introduces a "progressive disclosure" architecture that allows AI agents to load domain expertise on demand, reducing token usage by up to 90% compared to traditional monolithic prompts. Through four distinct patterns—ranging from simple inline checklists to "skill factories" where agents write their own code—the system enables agents to dynamically expand their capabilities at runtime using the universal agentskills.io specification. This modular approach ensures that complex instructions and external resources are only accessed when relevant, creating a scalable and self-extending framework for modern AI development.
在质量、成本和延迟之间都达到先进水平的视频生成 API。
State-of-the-art video generation across quality, cost, and latency.
Claude Sonnet 4.6 在编码、agents 和专业工作等大规模场景里带来前沿级性能。
ProductFeb 17, 2026Introducing Claude Sonnet 4.6Sonnet 4.6 delivers frontier performance across coding, agents, and professional work at scale.
我们正在升级最聪明的模型。无论是 agentic coding、computer use、tool use、search 还是 finance，Opus 4.6 都是行业领先模型，而且往往领先幅度很大。
AnnouncementsFeb 5, 2026Introducing Claude Opus 4.6We’re upgrading our smartest model. Across agentic coding, computer use, tool use, search, and finance, Opus 4.6 is an industry-leading model, often by wide margin.
TechCrunch 报道称，合规创业公司 Delve 的争议进一步升级，有吹哨人指控它把开源无代码工具 SimStudio 的修改版包装成自己的产品出售。
Julie Bort / TechCrunch: A whistleblower alleges Delve pitched a modified copy of open-source no-code tool SimStudio as its own, a practice that could violate the software's license — The controversy surrounding compliance startup Delve has gone from bad to worse this week. Among the fresh allegations …
Daniel Curtis
ESLint version 10 has removed the legacy eslintrc configuration system, finalizing a long transition to flat config. The update enhances developer experience, especially for plugin authors and monorepo teams, by changing configuration file location and improving JSX reference tracking. Node.js support has been tightened, and new assertion options have been added to the RuleTester API. By Daniel Curtis
Bloomberg 报道称，Elon Musk 与美国证券交易委员会都告诉法官，他们正走向一场正式审判，争议焦点是 Musk 在 2022 年是否按要求披露 Twitter 持股。
Nicola M White / Bloomberg: In a joint filing, Elon Musk and the US SEC say they are ready to move toward a trial over Musk's alleged failure to disclose his Twitter stake in 2022 — Elon Musk and the US Securities Exchange Commission told a judge they are heading toward a trial over the regulator's allegations …
Leela Kumili
Pinterest engineering teams have deployed a production-ready Model Context Protocol (MCP) ecosystem that allows AI agents to automate complex engineering tasks and integrate diverse internal tools. Domain-specific MCP servers, a central registry, and human-in-the-loop approval improve security, governance, and developer productivity while saving thousands of hours per month. By Leela Kumili
9to5Google 报道称，Google 宣布 AI Pro 订阅现在附带 5TB 存储空间，而之前的同价位套餐只有 2TB。
Abner Li / 9to5Google: Google increases the storage of its $19.99/month AI Pro subscription plan to 5TB, up from 2TB, at no additional cost — Google today announced that AI Pro subscriptions now come with 5 TB of storage. — Previously, subscribers of the $19.99 per month plan got 2 TB, just like the $9.99 tier.
Sophie Weston 认为技术职业并不是一条直线式晋升阶梯。结合 30 年经验，她分享了资深 IC 如何通过系统思维、策略能力、公开演讲和社区参与，成长为更强的 Principal Engineer。
Sophie Weston explains that technical careers are winding journeys, not straight ladders. Drawing on 30 years of experience, she shares how senior ICs can become "broken combs" by broadening skills in systems thinking and strategy. She discusses the vital role of organizational flexibility and explains how public speaking and community engagement create feedback loops for career success. By Sophie Weston
Reuters 报道称，Intel 计划再向由其 CEO Lip-Bu Tan 担任董事长的 SambaNova 投资 1500 万美元，此前 2 月已投过 3500 万美元。
Reuters: Documents: Intel plans to invest $15M in SambaNova, a startup chaired by CEO Lip-Bu Tan, increasing Intel's stake to 9%, following a $35M investment in February — Chipmaker Intel (INTC.O) has planned to invest another $15 million in SambaNova, a chip startup chaired by Intel CEO Lip-Bu Tan …
Steef-Jan Wiggers
Cloudflare has released Dynamic Worker Loader into open beta, offering V8 isolate-based sandboxing for AI-generated code execution. The company claims isolates start in milliseconds, using megabytes of memory, making them roughly 100x faster and up to 100x more memory-efficient than containers. The feature builds on Cloudflare's Code Mode approach. By Steef-Jan Wiggers
这篇文章演示了如何在 Spring Boot 中一步步实现基于 OTP 的 Google Authenticator 双因素认证，用来提升应用安全性。
Implement Google Authenticator 2FA in Spring Boot using OTP-based authentication to secure your application step by step.Continue reading on Medium »
作者以自己是 Web Developer 与 Digital Marketer 的视角出发，讨论 2026 年 Web 开发的整体图景，以及品牌如何打造更强的线上存在。
I am a passionate Web Developer and Digital Marketer with a strong interest in helping brands build a powerful online presence. I…Continue reading on Medium »
每次你打开 Instagram、在 Google 搜索，或者查看银行余额时，数据库都在毫秒间完成写入、存储和读取。文章试图把这些底层过程讲清楚。
Every time you open Instagram, search on Google, or check your bank balance, something happens in milliseconds. Data is written, stored…Continue reading on Medium »
这个案例来自 Senac 的“创意 UX: 连接用户的界面设计”课程，重点展示了 AI 如何帮助快速完成问题发现与界面原型设计。
Este case foi desenvolvido no contexto do curso UX criativo: design de interfaces que conectam, do Senac. A principal atividade teve como…Continue reading on Medium »
随机插件、臃肿网站、到处打补丁、没有监控、没有真正的安全策略，一切在出事前看起来都像“还能用”。问题不在于投入不够，而在于结构失控。文章给出了一套更稳的思路：用轻量前端、干净后端、可扩展基础设施、内建安全与可观测性来搭建真正能长期成长的技术栈。
Let’s be honest—most small business tech stacks are a mess. Random plugins. Bloated websites. Duct-taped integrations. No monitoring. No real security. And somehow… everything is “working” until it suddenly isn’t. If you’ve ever had a site crash during peak traffic or spent hours debugging something that shouldn’t be broken, you already know the problem. The issue isn’t effort—it’s structure. If you're building or scaling a business, your tech stack needs to be intentional, not reactive. The Real Problem: Reactive Development Most small businesses build their systems like this: Need a website? Install a theme Need a feature? Add a plugin Something breaks? Patch it Need automation? Zapier it It works… until it doesn’t. This approach creates: Technical debt Performance issues Security vulnerabilities Scaling limitations You end up spending more time fixing than building. What a Solid Tech Stack Actually Looks Like A good tech stack isn’t about using the “best” tools—it’s about using the right combination of tools that scale together. Here’s what that typically includes: Frontend That Prioritizes Performance Lightweight frameworks Optimized assets Mobile-first design Speed isn’t optional anymore. It directly impacts conversions, SEO, and user retention. Backend That Doesn’t Fight You Clean architecture Scalable APIs Proper database structure If your backend is hard to work with, everything slows down—development, updates, and debugging. Infrastructure That Can Scale Cloud-based hosting Load balancing CDN integration If your system can’t handle growth, you’re building a ceiling into your business. Security Built In (Not Added Later) HTTPS everywhere Regular patching Access control Security isn’t a feature—it’s a foundation. Monitoring and Observability Error tracking Performance monitoring Uptime alerts If you don’t know something is broken, you can’t fix it. The Biggest Mistake: Overcomplicating Everything Ironically, a lot of businesses try to fix their messy stack by adding more tools. That usually makes it worse. The goal isn’t complexity—it’s clarity. Ask yourself: Do we actually need this tool? Can this be simplified? Is this scalable long-term? If the answer is no, cut it. Where Most Teams Get Stuck Even when businesses know their stack is broken, they don’t fix it because: It feels overwhelming They don’t have in-house expertise They’re afraid of downtime So they keep patching instead of rebuilding. A Better Approach: Build With Intention Instead of reacting to problems, build your system with a long-term mindset: Choose tools that integrate well Keep your architecture simple Plan for scale early Automate where it actually matters This doesn’t mean overengineering—it means thinking ahead. Real-World Impact When your tech stack is clean and scalable: Your site loads faster Your users stay longer Your team moves quicker Your business grows without friction Technology stops being a bottleneck—and starts being an advantage. Final Thoughts You don’t need a perfect stack. You need a reliable, scalable, and maintainable one. If your current setup feels fragile, slow, or overly complex, that’s your signal—it’s time to rethink how everything is built. If you want help simplifying and scaling your tech the right way, check out: 👉 https://pushcam-solution.com/
这篇文章盘点了 7 种看似“老掉牙”、却仍在支撑现实世界关键系统的编程语言，例如 COBOL、FORTRAN、Assembly、LISP、Erlang、Ada 和 Prolog。它们仍默默运行在银行、航天、航空电子和即时通信系统里。
Hello👋 everyone I was halfway through a thought when something hit me and stopped me dead in my tracks: the software running your ATM machine was coded even before your parents were born. The code forecasting the weather of tomorrow is older than the internet. The backend delivering your WhatsApp messages is built using a programming language that most developers have not even heard of. We often discuss what the latest tech things are. Languages like #Python, #Rust, and #TypeScript, the shiny tools that everyone is learning. However, very few talk about the real factors that are holding all these up. Hence, I followed a trail down the hole. I looked into 7 programming languages that the world had declared as old and no longer of use, and yet I discovered them quietly running global banking, space missions, aircraft avionics, and real-time communication of billions of people. Here is the first piece of my blog: 👉 "From ATMs to WhatsApp: The Old Programming Languages Behind Your Everyday Tech" https://vickkykruzprogramming.dev/blog/from-atms-to-whatsapp-the-old-programming-languages-behind-your-everyday-tech It delves into COBOL, FORTRAN, Assembly, LISP, Erlang, Ada, and Prolog, complete with actual statistics, real companies, and the honest reasons why nobody has yet replaced them. Let me know what you think, have you ever worked with any of these languages? Leave a message below and share your experience. And if you like content that digs deeper than the headlines, don't hesitate to hit the follow or subscribe button. There's a lot more coming. 🙏 Programming #TechHistory #COBOL #SoftwareDevelopment #LegacyCode #Developer #TechBlog
发帖者在 RTX 5060 Ti 16GB 上测试本地模型，目标是不用 API 费用也能跑一个足够强的模型。最终发现 TurboQuant 不只是对 KV 有用，还能让模型更小、更适合本地部署。
I bought an RTX 5060 Ti 16GB around Christmas and had one goal: get a strong model running locally on my card without paying api fees. I have been testing local ai with open claw. I did not come into this with a quantization background. I…
AnythingLLM 的 Tim 在看到 PrismML Bonsai 的发布后做了认真测试，因为这种体积缩小 14 倍的模型很有吸引力。
Hey everyone, Tim from AnythingLLM and yesterday I saw the PrismML Bonsai post so i had to give it a real shot because 14x smaller models (in siz…
作者尝试把 self-attention 里的标准 dot-product 替换成另一种距离度量，比如 RBF kernel，并分享实验过程与结论。
I recently asked myself what would happen if we replaced the standard dot-product in self-attention with a different distance metric, e.g. an rbf-kernel? Standard dot-product attention has this quirk where a key vector can "bully" the soft…
发帖者看到很多关于 AI 使用的分裂意见，想了解构建 ML 模型的工程师会如何看待 vibe coding 这件事。
I've seen, read and heard a lot of mixed reactions about software engineers (ie. the ones who aren't building ML models and make purely deterministic software) giving their opinions on AI usage. Some say it speeds up their workflow as it f…
我们第一次看到了一个大规模生产级 AI agent 系统的完整架构。
Most coverage of the Claude Code leak focuses on the drama or the hidden features. But the bigger story is that this is the first time we've seen the complete architecture of a production-grade AI agent system running at scale ($2.5B ARR,…
美国最大公立医院系统 CEO 表示他已经准备好用 AI 替代放射科医生
CEO of America’s largest public hospital system says he’s ready to replace radiologists with AI
版主表示，经过大量讨论后，他们决定试行一段时间的 LLM 内容禁令，因为相关帖子数量太多，而且通常不符合该 subreddit 希望呈现的深度技术讨论氛围。
Hey folks, After a lot of discussion, we've decided to trial a ban of any and all content relating to LLMs. We get a lot of posts related to LLMs and typically they are not in line with what we want the subreddit to be — a place for detail…
作者最初是在 Fast Mode 刚引入时做的这次逆向分析，也曾尝试联系 Anthropic 做负责披露，但没有得到回复。现在因为社区开始大量讨论 CCH header，他决定把结果公开。
I originally reverse engineered this when Fast Mode was first introduced and contacted Anthropic in order to get approval for responsible disclosure but I never heard back. Now that there's a lot of buzz around the CCH header, I wanted to…
Pyrefly 团队一直认为类型覆盖率是代码质量最重要的指标之一。过去一年里，他们和 Meta 的大型 Python 代码库团队紧密合作，在提升性能和收紧类型系统的同时，也重新思考了“覆盖率”的定义。
On the Pyrefly team, we've always believed that type coverage is one of the most important indicators of code quality. Over the past year, we've worked closely with teams across large Python codebases at Meta - improving performance, tight…
作者最近集中优化 Talk Python 旗下 Web 应用的内存使用。他们在一台 16GB 服务器上跑 23 个容器，内存使用已经爬到 65%，最后通过一系列调整把占用大幅降了下来。
Over the past few weeks I went on a memory-reduction tear across the Talk Python web apps. We run 23 containers on one big server (the "one big server" pattern) and memory was creeping up to 65% on a 16GB box. Turned out there were a bunch…
[![npm]](https://www.npmjs.com/package/@anthropic-ai/claude-code) [npm]: https://img.shields.io/npm/v/@anthropic-ai/claude-code.svg?style=flat-square
[![npm]](https://www.npmjs.com/package/@anthropic-ai/claude-code) [npm]: https://img.shields.io/npm/v/@anthropic-ai/claude-code.svg?style=flat-square
VibeVoice 是一组**开源前沿语音 AI 模型**，同时包含文本转语音（TTS）和自动语音识别（ASR）模型。
VibeVoice is a **family of open-source frontier voice AI models** that includes both Text-to-Speech (TTS) and Automatic Speech Recognition (ASR) models.
TimesFM（Time Series Foundation Model）是 Google Research 开发的预训练时间序列基础模型，用于时间序列预测。
TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
从“只会输入”走向“编排 agent、hooks、skills 和 MCP servers”，配有可视化教程、可复制模板和引导式学习路径。
Go from typing to orchestrating agents, hooks, skills, and MCP servers — with visual tutorials, copy-paste templates, and a guided learning path.
面向浏览器和 node.js 的 Promise 风格 HTTP client。
Promise based HTTP client for the browser and node.js
用你喜欢的包管理器全局安装，然后直接运行即可开始使用。
Install globally with your preferred package manager: Then simply run to get started.
全球最大的开源 AI prompt 库，支持 ChatGPT、Claude、Gemini、Llama、Mistral 等模型，原名 Awesome ChatGPT Prompts。
The world's largest open-source prompt library for AI Works with ChatGPT, Claude, Gemini, Llama, Mistral, and more formerly known as Awesome ChatGPT Prompts 🌐 Browse Prompts • 📖 Read the Book • 📄 View on GitHub • 🚀 Self-Host