巴菲特退休前为数不多的加仓又“爆雷”！什么信号？_腾讯新闻 巴菲特退休前为数不多的加仓又“爆雷”！什么信号？ 华尔街见闻 2026-02-20 20:51 发布于 上海 华尔街见闻官方账号 巴菲特在退休前这三年一直在净卖出股票，好不容易有几个新进的仓位，进入2026年也在连续“爆雷”，比如说此前的 联合健康 ，一天就暴跌20%。 今天，此前老巴不断加仓的POOL，也“爆雷”了。 现在市场上好不容易被老巴看到可能是机会的小机会都不是机会，而是坑 ；这也难怪他老人家要囤这么多现金。 因为现在整个美股市场就像AI一样，建立在一个“ 循环融资 ”的基础之上…… 发生了什么？ 伯克希尔 ·哈撒韦对 Pool Corporation 的投资始于2024年第三季度。在这个季度，伯克希尔首次向SEC披露了其新建仓的POOL头寸，买入规模为404,057股 。根据市场交易数据还原，该季度POOL股票的平均收盘价约为345.95美元。 进入2024年第四季度，伯克希尔继续稳步推进其建仓计划，增持了194,632股股票，较上一季度末的持仓量增加了48.17%，使总持仓量达到598,689股。 2025年第一季度，POOL的股价跌势明显，伯克希尔大举增持了865,311股，增持幅度高达144.53%，使其总持仓量迅速跃升至1,464,000股。 2025年第二季度，POOL股价进一步破位下行，股价跌破300美元整数关口 ，伯克希尔执行了自建仓以来规模最大的一次单季买入：监管文件显示，伯克希尔在2025年二季度增持了惊人的1,994,885股，环比激增136.26%。至此，伯克希尔对POOL的总持仓量达到了3,458,885股的历史顶峰，不仅使其成为公司前25大持仓之一，也让伯克希尔获得了POOL发行在外总股本8.24%的所有权。 在经历了四个季度的连续买入，并度过了持仓不变的2025年第三季度后，伯克希尔的操作方向在2025年底发生了微妙的逆转。SEC于2026年初披露的针对2025年第四季度的13F文件显示：伯克希尔首次对POOL的头寸进行了削减，卖出了390,000股股票，占其当时总持仓的11.28%。 值得注意的是，此次减持是在股价极度低迷的环境下进行的。2025年第四季度，POOL的平均收盘价仅为259.34美元，价格区间在228.75美元至312.05美元之间波动 。而在削减头寸之前，伯克希尔在Pool Corporation上的加权平均建仓成本大概在约320美元的位置。 刚刚大举建仓没多久，就以亏损状态大幅削减头寸，显然这是在“认错”了。 最新的公布的POOL的财报，好像印证了削减头寸的判断 ：2026年2月19日POOL公布的2025年第四季度及2025全年业绩显示，营收、每股收益（EPS）以及前瞻指引均低于华尔街预期，引发股价大幅抛售。股价一度跌至新的52周低点216.32美元，已经较巴菲特的持仓成本线低了三成。 意味着什么？ POOL的竞争地位仍然相对稳固：第四季度POOL的毛利率提升了70个基点至30.1%，同时全年基础毛利率也提升了20个基点（毛利率保持强劲是行业竞争力强的重要表现之一），而且公司还在扩张：2025年底POOL的销售中心网络扩张至456个直营网点，高于2024年末的448个。 但是公司的营收和利润都还未止住下滑的趋势：2025年全年净销售额为52.89亿美元，较2024年的53.11亿美元继续下滑；2025年全年净利润更是下降了6.4%至4.064亿美元，稀释后每股收益（EPS）从11.30美元继续降至10.85美元。 这背后主要是通胀的持续压力：全年销售及管理费用同比增长 4%，达到9.923 亿美元，在收入下降的情况下，这种费用增长显著压缩了营业利润率。 更糟糕的是现金流质量的恶化：经营活动现金流从2024年的6.592亿美元大幅下滑44%，降至2025年仅3.659亿美元。自由现金流（扣除资本开支后）也从约6亿美元降至约3.10亿美元。 POOL的业绩和资产负债表，甚至是现金流和前景的恶化，尽管有自身行业本身的特征，但是作为 可选消费 的一个侧面，也折射出目前美国经济的脆弱性： POOL的疲弱，放在美国消费结构日益“头重脚轻”的大背景下就更容易理解： 根据穆迪分析的数据，美国收入最高的10%人群如今贡献了大约49%的全部消费者支出——这是自1989年开始追踪这一比例以来的最高的水平。与此同时，底部80%人群的消费占比降至创纪录低点，仅为41%。现在美国就只剩下10-20%的人算是真正的中产，享受到了与自己收入层级和人数相对应的消费比例。 由于美国是消费驱动的经济体，所以现在美国经济受到富裕阶层消费能力的影响非常大，据估计仅前10%人群的消费就贡献了约三分之一的 GDP ；至于收入处于底部80%的人群来说，过去六年他们的支出增长几乎只是勉强跟上通胀——也就是说，自疫情开始以来，他们的生活水平事实上基本停滞。与此同时，美国 基尼系数 已处于60年来的高位，而且高收入与低收入群体之间的信心差距也扩大到十余年来最宽的水平。 如果POOL所代表的中上产和富人消费在萎缩，那其他方面也已经出现了裂痕： 2025年第三季度的一项调查发现，20%的富裕及高净值人群打算减少在设计师时装上的支出； 美国银行 （BofA）的刷卡数据则显示，美国奢侈品消费已连续10个季度下滑。 当然，相比于泳池和奢侈品支出这些表面现象，美国富人的支出真正的影响来源是美股：支撑美国富裕人群消费的力量，主要来自股市带来的“ 财富效应 ”——美国家庭财富排名前20%的群体，持有（直接与间接持有合计）大约87%的企业股票资产。 这也造成了一种令人不安的依赖：一旦股市出现10%以上的长期回调，打击将会精准落在消费能力最集中的人群上。 所以现在美国整体经济都已经有点像AI的“循环融资”：股市上涨——财富效应——富人消费——提振企业利润和经济——股市继续上涨…… 然而，这种单纯由财富效应和财政刺激维持的繁荣，总有一天会崩塌，而可选消费企业的业绩已经提前亮起了红灯。

Source: Tencent News | Original Link

Web Components: The Framework-Free Renaissance Web Components: The Framework-Free Renaissance 8 min read Building Modern UIs Without the Framework Tax 17.02.2026, By Stephan Schwab Modern browsers now support everything needed to build sophisticated, reactive web interfaces without React, Vue, or Angular. Web components, custom elements, shadow DOM, and native event systems let developers create modular, reusable UI pieces that communicate elegantly — and AI assistants can help you master these patterns faster than ever before. The Shift That Already Happened Something remarkable occurred while many developers weren’t watching. The web platform itself became capable of doing what frameworks were invented to do. “The browser has become the framework. We just haven’t fully noticed yet.” Custom Elements let you define your own HTML tags with their own behavior. Shadow DOM provides encapsulation that keeps component styles and structure isolated. Templates and slots offer composition patterns. And perhaps most importantly, the native event system provides a robust mechanism for components to communicate without tight coupling. These aren’t experimental features. They’ve been shipping in every major browser for years. The question is no longer whether they work, but why more developers haven’t embraced them. Freedom from the Upgrade Treadmill Every framework carries hidden costs. There’s the initial learning curve, certainly. But there’s also the ongoing maintenance burden: major version upgrades that break things, deprecated patterns you must migrate away from, build tooling that needs constant updating. “When your component is just HTML, CSS, and JavaScript, there’s nothing to upgrade except your own code.” Web components built on platform standards sidestep this entirely. The browser vendors have committed to backward compatibility in ways that framework maintainers simply cannot. Code written to web standards a decade ago still works today. That’s not true of code written for Angular 1, or React class components, or Vue 2’s options API. For organizations building products that need to run for years, this stability matters enormously. It’s one less thing that can break, one less dependency that can become a security vulnerability, one less abstraction layer between your code and the runtime. Components That Talk to Each Other The elegance of web components becomes most apparent when you consider how they communicate. The native Custom Events system provides everything you need for sophisticated component interaction. A component deep in your UI hierarchy can dispatch an event that bubbles up through the DOM tree: this . dispatchEvent ( new CustomEvent ( ‘ item-selected ‘ , { detail : { itemId : this . selectedId , metadata : this . itemData }, bubbles : true , composed : true })); Any ancestor component — or the application shell itself — can listen for and respond to that event. There’s no need for a global state store, no prop drilling, no contex

Source: Hacker News | Original Link

Subscribe to read Accessibility help Skip to navigation Skip to main content Skip to footer Nvidia and OpenAI abandon unfinished $100bn deal in favour of $30bn investment Subscribe to unlock this article Save 40% on Standard Digital was $540 now $299 for your first year Save now on essential digital access to trusted FT journalism on any device. Savings based on monthly annualised price – offer ends 25th February Save 40% What’s included Standard Digital Stay informed with global news & in-depth analysis Gain clarity from expert opinion First FT: the day’s biggest stories 20+ curated newsletters Follow topics & set alerts with myFT FT App on Android & iOS FT Videos & Podcasts 10 monthly gift articles to share Explore more offers. Trial $1 for 4 weeks Then $75 per month. Complete digital access with exclusive insights and industry deep dives on any device. Cancel anytime during your trial. Select What’s included Global news & analysis Expert opinion FT App on Android & iOS FT Edit: Access on iOS and web FirstFT: the day’s biggest stories 20+ curated newsletters Follow topics & set alerts with myFT FT Videos & Podcasts 20 monthly gift articles to share Lex: FT’s flagship investment column 15+ Premium newsletters by leading experts FT Digital Edition: our digitised print edition Premium Digital Complete coverage $75 per month Complete digital access with exclusive insights and industry deep dives on any device. Select What’s included Global news & analysis Expert opinion FT App on Android & iOS FT Edit: Access on iOS and web FirstFT: the day’s biggest stories 20+ curated newsletters Follow topics & set alerts with myFT FT Videos & Podcasts 20 monthly gift articles to share Lex: FT’s flagship investment column 15+ Premium newsletters by leading experts FT Digital Edition: our digitised print edition FT Digital Edition $119 per year All the content of the FT newspaper on any device (This subscription does not include access to FT.com or the FT App). Select What’s included FT Digital Edition: our digitised print edition Global news & analysis Expert opinion Translate instantly to 26 languages Lex: FT’s flagship investment column FT Magazines, including HTSI Offline Access Check whether you already have access via your university or organisation. Terms & Conditions apply Explore our full range of subscriptions. For individuals Discover all the plans currently available in your country Digital Print Print + Digital For multiple readers Digital access for organisations. Includes exclusive features and content. FT Professional Why the FT? See why over a million readers pay to read the Financial Times. Find out why Close side navigation menu Search the FT Search Subscribe for full access

Source: Hacker News | Original Link

blakewatson.com – I used Claude Code and GSD to build the accessibility tool I’ve always wanted Skip to content I used Claude Code and GSD to build the accessibility tool I’ve always wanted February 18, 2026 Because of a severe mobility impairment—spinal muscular atrophy—I use a Mac for almost everything I do, and I have a particularly unique way of interacting with it . One of my biggest challenges— aside from typing —is the rather mundane act of scrolling . I’m unable to manipulate a scroll wheel or perform a swiping gesture. That means the usual methods for scrolling on a mouse or trackpad are out. The method I do use is probably something you haven’t thought about in a long time—I click on and drag the scrollbar. [1] In most instances, it’s not too bad. But it can be very annoying in cases where the scrollable area is very long or in newer apps that tend to hide the scrollbar when not in use. For years now, I’ve been able to ease my scrolling burden with a fantastic browser extension called ScrollAnywhere . It works great on most web pages and I do use the web a lot. That said, there are a ton of places where it doesn’t work. Even in the web browser, it doesn’t work in reader mode and it doesn’t work in browser settings views and other built-in browser pages. And obviously it doesn’t work in other apps or in the operating system itself. Several months after posting about my scrollbar woes , Ryan Ohs sent me an email with a delightful surprise. He had been learning about Apple’s Accessibility API and after reading my post he put together this quick and dirty drag and scroll program that worked at the operating system level. It was a cool experiment. And while I was grateful for the gesture, it didn’t turn out to be practically usable in my day-to-day workflow. Time passed. I helplessly watched as app after app shipped with inaccessible scrollbars. I made do with a cumbersome workaround—macOS has a dwell action for scrolling. The way it works is when you enable dwelling, you hover your mouse over where you want to scroll and wait for a short delay. Some UI appears with arrows which you can then hover your mouse over to scroll the area in that direction. It’s slow and annoying but effective when needed. In other cases where apps are built with customizable web technology—like Obsidian —I was able to write CSS styles to make scrollbars large and in charge. Fast-forward to now. AI is everywhere. Many developers I follow are blogging or making videos about the moment they realized AI is a force to be reckoned with. Other developers I follow loathe the technology, citing ethical concerns over environmental impact and intellectual property theft. I have a ton of cognitive dissonance. It’s amazing and horrible all at the same time. I would say that 90% of my current AI use is Copilot autocomplete in VS Code. This saves me a ton of time because I’m using an onscreen keyboard, pecking out code character by character. Yes, there is standard editor comple

Source: Hacker News | Original Link

Untapped Way to Learn a Codebase: Build a Visualizer Jimmy Miller Untapped Way to Learn a Codebase: Build a Visualizer The biggest shock of my early career was just how much code I needed to read that others wrote. I had never dealt with this. I had a hard enough time understanding my own code. The idea of understanding hundreds of thousands or even millions of lines of code written by countless other people scared me. What I quickly learned is that you don’t have to understand a codebase in its entirety to be effective in it. But just saying that is not super helpful. So rather than tell, I want to show. In this post, I’m going to walk you through how I learn an unfamiliar codebase. But I’ll admit, this isn’t precisely how I would do it today. After years of working on codebases, I’ve learned quite a lot of shortcuts. Things that come with experience that just don’t translate for other people. So what I’m going to present is a reconstruction. I want to show bits and parts of how I go from knowing very little to gaining knowledge and ultimately, asking the right questions. To do this, I will use just a few techniques: Setting a goal Editing randomly Fixing things I find that are broken Reading to answer questions Making a visualizer I want to do this on a real codebase, so I’ve chosen one whose purpose and scope I’m generally familiar with. But one that I’ve never contributed to or read, Next.js . But I’ve chosen to be a bit more particular than that. I’m particularly interested in learning more about the Rust bundler setup (turbopack) that Next.js has been building out. So that’s where we will concentrate our time. Being Clear About Our Goal Trying to learn a codebase is distinctly different from trying to simply fix a bug or add a feature. In post, we may use bugs, talk about features, make changes, etc. But we are not trying to contribute to the codebase, yet. Instead, we are trying to get our mind around how the codebase generally works. We aren’t concerned with things like coding standards, common practices, or the development roadmap. We aren’t even concerned with correctness. The changes we make are about seeing how the codebase responds so we can make sense of it. Starting Point I find starting at main to be almost always completely unhelpful. From main, yes, we have a single entry point, but now we are asking ourselves to understand the whole. But things actually get worse when dealing with a large codebase like this. There isn’t even one main. Which main would we choose? So instead, let’s start by figuring out what our library even consists of. ❯ ls -1d */ apps/ bench/ contributing/ crates/ docs/ errors/ examples/ packages/ patches/ rspack/ scripts/ test-config-errors/ test/ turbo/ turbopack/ A couple of things to note. We have packages, crates, turbo, and turbopack. Crates are relevant here because we know we are interested in some of the Rust code, but we are also interested in turbopack in particular. A quick look at these shows that

Source: Hacker News | Original Link

Collections: Against the State – A Primer on Terrorism, Insurgency and Protest – A Collection of Unmitigated Pedantry Skip to content Bret Devereaux Collections , Military Theory Primers February 13, 2026 44 Minutes This week, continuing in the vein of some of our previous strategy and military theory primers , I wanted to off a basic 101-level survey of the strategic theory behind efforts, in a sense, directed against the state itself, both violent approaches (what we might call ‘terroristic insurgency’) 1 and non-violent approaches (protest). It may seem strange to treat violent insurgency and non-violent protest together but while they employ very different methods, as we’ll see, they share a similar theoretical framework, attempting to achieve some of the similar effects by different means , both working within the state, against the state (or its policies), focused on the changing minds rather than battlefields . Naturally this comes in part in response to the significant amounts of protest actions happening right now in the United States, but the framework here is very much intended to be a general one, applicable to both armed insurgencies and non-violent protests worldwide. The world, after all, is really quite big and there are multiple major protest movements and multiple armed insurgencies happening globally at any given time. That said, much as with protracted war, a movement aiming to push against the state is naturally going to be heavily shaped by local conditions, particularly by the nature of the state against which it sets itself as well as the condition and political alignment of its people. Finally, I want to clarify how I am using terminology here at the outset. I have mostly stuck here to ‘insurgent’ to describe violent actors opposing the state and ‘protestor’ to describe non-violent ones. Obviously in mass movements, violence is not a binary but a spectrum – a single fellow kicking over a trash can does not turn a non-violent march into a riot, but equally having a ‘political wing’ does not turn an organization mounting a terror campaign ‘non-violent.’ However the strategic dichotomy is going to be useful to understanding how these groups in their ideal form tackle problems. Likewise, I am going to describe the violent movements opposing the state as ‘insurgencies’ but I want to note at the outset that I am drawing a distinction here between what I am defining as ‘insurgencies’ which lack the backing of a conventional army or the expectation of soon acquiring one, as opposed to forces in a protracted war framework who have or expect to have the backing of a conventional force, however weak (we might call the latter group guerillas, although this too is imprecise). The line between these two strategies is certainly fuzzy – many insurgencies hope to eventually transition to protracted war and the two approaches share many tactics – but there are worthwhile differences between the two. In particular, whereas the guerilla’s ca

Source: Hacker News | Original Link

Minions: Stripe’s one-shot, end-to-end coding agents—Part 2 | Stripe Dot Dev Blog Minions: Stripe’s one-shot, end-to-end coding agents—Part 2 / Metadata Date: 2026.2.19 Author: Alistair Gray Reading time: 8 min read Categories: AI Developer Productivity LLMs Engineering Share: Twitter/X LinkedIn / Article / About the author Alistair Gray Alistair is a software engineer on the Leverage team. The Leverage team builds surprisingly delightful internal products that Stripes can leverage to supercharge their productivity. Minions: Stripe’s one-shot, end-to-end coding agents / Docs Explore our guides and examples to integrate Stripe. Learn more / Social Youtube Twitter/X Discord / Resources Docs Developer Meetups © 2025 Stripe, Inc. Privacy Legal Stripe.com

Source: Hacker News | Original Link

GitHub – HailToDodongo/pyrite64: N64 Game-Engine and Editor using libdragon & tiny3d Skip to content You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert HailToDodongo / pyrite64 Public Notifications You must be signed in to change notification settings Fork 72 Star 1.9k N64 Game-Engine and Editor using libdragon & tiny3d License MIT license 1.9k stars 72 forks Branches Tags Activity Star Notifications You must be signed in to change notification settings HailToDodongo/pyrite64 main Branches Tags Go to file Code Open more actions menu Folders and files Name Name Last commit message Last commit date Latest commit History 380 Commits 380 Commits .github/ workflows .github/ workflows data data docs docs n64 n64 src src vendored vendored .editorconfig .editorconfig .gitattributes .gitattributes .gitignore .gitignore .gitmodules .gitmodules CMakeLists.txt CMakeLists.txt CMakePresets.json CMakePresets.json Changelog.md Changelog.md LICENSE LICENSE Readme.md Readme.md View all files Repository files navigation Pyrite64 N64 game-engine and editor using Libdragon and tiny3d . Note: This project does NOT use any proprietary N64 SDKs or libraries. Pyrite64 is a visual editor + runtime-engine to create 3D games that can run on a real N64 console or accurate emulators. Besides the usual editor, some extra features include: Automatic toolchain installation on Windows 3D-Model import (GLTF) from blender with fast64 material support. Support for HDR+Bloom rendering (shown here: www.youtube.com/watch?v=XP8g2ngHftY ) Support for big-texture rendering (256×256) (shown here: www.youtube.com/watch?v=rNEo0aQkGnU ) Runtime engine handling scene-management, rendering, collision, audio and more. Global asset management with automatic memory cleanup Node-Graph editor to script basic control flow Note that this project focuses on real hardware, so accurate emulation is required to run/test games on PC. Emulators that are accurate enough include Ares (v147 or newer) and gopher64 . Warning This project is still in early development, so features are going to be missing. Documentation is also still a work in progress, and breaking API changes are to be expected. Documentation Before starting, please read the FAQ ! Installation & Docs: Pyrite64 Installation Using the Editor Using the CLI Development on the editor itself: Building the Editor Showcase Cathode Quest 64 (YouTube)    |    Pyrite64 Release Video Links For anything N64 homebrew related, checkout the N64Brew discord: https://discord.gg/WqFgNWf Credits & License © 2025-2026 – Max Bebök (HailToDodongo) Pyrite64 is licensed under the MIT License, see the LICENSE file for more information. Licenses for external libraries used in the editor can be found in their respective directory under /vendored Pyrite64 does NOT force any restrictions

Source: GitHub Trending | Original Link

The path to ubiquitous AI | Taalas The path to ubiquitous AI By Ljubisa Bajic Many believe AI is the real deal. In narrow domains, it already surpasses human performance. Used well, it is an unprecedented amplifier of human ingenuity and productivity. Its widespread adoption is hindered by two key barriers: high latency and astronomical cost. Interactions with language models lag far behind the pace of human cognition. Coding assistants can ponder for minutes, disrupting the programmer’s state of flow, and limiting effective human-AI collaboration. Meanwhile, automated agentic AI applications demand millisecond latencies, not leisurely human-paced responses. On the cost front, deploying modern models demands massive engineering and capital: room-sized supercomputers consuming hundreds of kilowatts, with liquid cooling, advanced packaging, stacked memory, complex I/O, and miles of cables. This scales to city-sized data center campuses and satellite networks, driving extreme operational expenses. Though society seems poised to build a dystopian future defined by data centers and adjacent power plants, history hints at a different direction. Past technological revolutions often started with grotesque prototypes, only to be eclipsed by breakthroughs yielding more practical outcomes. Consider ENIAC, a room-filling beast of vacuum tubes and cables. ENIAC introduced humanity to the magic of computing, but was slow, costly, and unscalable. The transistor sparked swift evolution, through workstations and PCs, to smartphones and ubiquitous computing, sparing the world from ENIAC sprawl. General-purpose computing entered the mainstream by becoming easy to build, fast, and cheap. AI needs to do the same. About Taalas Founded 2.5 years ago, Taalas developed a platform for transforming any AI model into custom silicon. From the moment a previously unseen model is received, it can be realized in hardware in only two months. The resulting Hardcore Models are an order of magnitude faster, cheaper, and lower power than software-based implementations. Taalas’ work is guided by the following core principles: 1. Total specialization Throughout the history of computation, deep specialization has been the surest path to extreme efficiency in critical workloads. AI inference is the most critical computational workload that humanity has ever faced, and the one that stands to gain the most from specialization. Its computational demands motivate total specialization: the production of optimal silicon for each individual model. 2. Merging storage and computation Modern inference hardware is constrained by an artificial divide: memory on one side, compute on the other, operating at fundamentally different speeds. This separation arises from a longstanding paradox. DRAM is far denser, and therefore cheaper, than the types of memory compatible with standard chip processes. However, accessing off-chip DRAM is thousands of times slower than on-chip memory. Conversely, compute ch

Source: Hacker News | Original Link

GitHub – blackboardsh/electrobun: Build ultra fast, tiny, and cross-platform desktop apps with Typescript. Skip to content You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert blackboardsh / electrobun Public Uh oh! There was an error while loading. Please reload this page . Notifications You must be signed in to change notification settings Fork 94 Star 5.6k Build ultra fast, tiny, and cross-platform desktop apps with Typescript. blackboard.sh/electrobun License MIT license 5.6k stars 94 forks Branches Tags Activity Star Notifications You must be signed in to change notification settings blackboardsh/electrobun main Branches Tags Go to file Code Open more actions menu Folders and files Name Name Last commit message Last commit date Latest commit History 1,663 Commits 1,663 Commits .github .github kitchen kitchen package package scripts scripts templates templates .gitignore .gitignore .gitmodules .gitmodules BETA_RELEASE.md BETA_RELEASE.md BUILD.md BUILD.md CEF.md CEF.md CLAUDE.md CLAUDE.md LICENSE LICENSE README.md README.md debug.js debug.js View all files Repository files navigation Electrobun Get started with a template npx electrobun init What is Electrobun? Electrobun aims to be a complete solution-in-a-box for building, updating, and shipping ultra fast, tiny, and cross-platform desktop applications written in Typescript. Under the hood it uses bun to execute the main process and to bundle webview typescript, and has native bindings written in zig . Visit https://blackboard.sh/electrobun/ to see api documentation, guides, and more. Project Goals Write typescript for the main process and webviews without having to think about it. Isolation between main and webview processes with fast, typed, easy to implement RPC between them. Small self-extracting app bundles ~12MB (when using system webview, most of this is the bun runtime) Even smaller app updates as small as 14KB (using bsdiff it only downloads tiny patches between versions) Provide everything you need in one tightly integrated workflow to start writing code in 5 minutes and distribute in 10. Apps Built with Electrobun Audio TTS – desktop text-to-speech app using Qwen3-TTS for voice design, cloning, and generation Co(lab) – a hybrid web browser + code editor for deep work Video Demos Star History Contributing Ways to get involved: Follow us on X for updates @BlackboardTech or @yoav.codes Join the conversation on Discord Create and participate in Github issues and discussions Let me know what you’re building with Electrobun Development Setup Building apps with Electrobun is as easy as updating your package.json dependencies with npm add electrobun or try one of our templates via npx electrobun init . This section is for building Electrobun from source locally in order to contribute fixes to it. Prerequ

Source: GitHub Trending | Original Link

GitHub – vxcontrol/pentagi: ✨ Fully autonomous AI Agents system capable of performing complex penetration testing tasks Skip to content You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert vxcontrol / pentagi Public Notifications You must be signed in to change notification settings Fork 429 Star 2.4k ✨ Fully autonomous AI Agents system capable of performing complex penetration testing tasks pentagi.com License MIT license 2.4k stars 429 forks Branches Tags Activity Star Notifications You must be signed in to change notification settings vxcontrol/pentagi master Branches Tags Go to file Code Open more actions menu Folders and files Name Name Last commit message Last commit date Latest commit History 207 Commits 207 Commits .github .github .vscode .vscode backend backend build build examples examples frontend frontend observability observability .dockerignore .dockerignore .env.example .env.example .gitignore .gitignore Dockerfile Dockerfile EULA.md EULA.md LICENSE LICENSE NOTICE NOTICE README.md README.md docker-compose-graphiti.yml docker-compose-graphiti.yml docker-compose-langfuse.yml docker-compose-langfuse.yml docker-compose-observability.yml docker-compose-observability.yml docker-compose.yml docker-compose.yml entrypoint.sh entrypoint.sh View all files Repository files navigation PentAGI P enetration testing A rtificial G eneral I ntelligence 🚀 Join the Community! Connect with security researchers, AI enthusiasts, and fellow ethical hackers. Get support, share insights, and stay updated with the latest PentAGI developments. ⠀ 📖 Table of Contents Overview Features Quick Start Advanced Setup Development Testing LLM Agents Embedding Configuration and Testing Function Testing with ftester Building Credits License 🎯 Overview PentAGI is an innovative tool for automated security testing that leverages cutting-edge artificial intelligence technologies. The project is designed for information security professionals, researchers, and enthusiasts who need a powerful and flexible solution for conducting penetration tests. You can watch the video PentAGI overview : ✨ Features 🛡️ Secure & Isolated. All operations are performed in a sandboxed Docker environment with complete isolation. 🤖 Fully Autonomous. AI-powered agent that automatically determines and executes penetration testing steps. 🔬 Professional Pentesting Tools. Built-in suite of 20+ professional security tools including nmap, metasploit, sqlmap, and more. 🧠 Smart Memory System. Long-term storage of research results and successful approaches for future use. 📚 Knowledge Graph Integration. Graphiti-powered knowledge graph using Neo4j for semantic relationship tracking and advanced context understanding. 🔍 Web Intelligence. Built-in browser via scraper for gathering latest information from web sources. 🔎 Externa

Source: GitHub Trending | Original Link

“Made in EU” – it was harder than I thought. When I decided to build my startup on European infrastructure, I thought it would be a straightforward swap. Ditch AWS, pick some EU providers, done. How hard could it be? Turns out: harder than expected. Not impossible, I did it, but nobody talks about the weird friction points you hit along the way. This is that post. Why bother? Data sovereignty, GDPR simplicity, not having your entire business dependent on three American hyperscalers, and honestly, a bit of stubbornness. I wanted to prove it could be done. The EU has real infrastructure companies building serious products. They deserve the traffic. The stack Here’s what I landed on after a lot of trial, error, and migration headaches. Hetzner handles the core compute. Load balancers, VMs, and S3-compatible object storage. The pricing is almost absurdly good compared to AWS, and the performance is solid. If you’ve never spun up a Hetzner box, you’re overpaying for cloud compute. Scaleway fills the gaps Hetzner doesn’t cover. I use their Transactional Email (TEM) service, Container Registry, a second S3 bucket for specific workloads, their observability stack, and even their domain registrar. One provider, multiple services, it simplifies billing if nothing else. Bunny.net is the unsung hero of this stack. CDN with distributed storage, DNS, image optimization, WAF, and DDoS protection, all from a company headquartered in Slovenia. Their edge network is genuinely impressive and their dashboard is a joy to use. Coming from Cloudflare, I felt at home rather quickly. Nebius powers our AI inference. If you need GPU compute in Europe without sending requests to us-east-1 , they’re one of the few real options. Hanko handles authentication and identity. A German provider that gives you passkeys, social logins, and user management without reaching for Auth0 or Clerk. More on this in the “can’t avoid” section — it doesn’t eliminate American dependencies entirely, but it keeps the auth layer European. Self-hosting: Rancher, my beloved This is where things get fun… and time-consuming. I self-host a surprising amount: Gitea for source control Plausible for privacy-friendly analytics Twenty CRM for customer management Infisical for secrets management Bugsink for error tracking All running on Kubernetes, with Rancher as the glue keeping the whole cluster sane. Is self-hosting more work than SaaS? Obviously. But it means my data stays exactly where I put it, and I’m not at the mercy of a provider’s pricing changes or acquisition drama. For email, Tutanota keeps things encrypted and European. UptimeRobot watches the monitors so I can sleep. The parts that were extra hard Transactional email with competitive pricing. This one surprised me. Sendgrid, Postmark, Mailgun, they all make it trivially easy and reasonably cheap. The EU options exist, but finding one that matches on deliverability, pricing, and developer experience took real effort. Scaleway’s TEM works, but

Source: Hacker News | Original Link

Sina Visitor System

Source: Weibo Hot Search | Original Link

华尔街见闻

Source: Wall Street CN | Original Link

华尔街见闻

Source: Wall Street CN | Original Link

Defer available in gcc and clang – Jens Gustedt’s Blog February 15, 2026 Defer available in gcc and clang About a year ago I posted about defer and that it would be available for everyone using gcc and/or clang soon. So it is probably time for an update. Two things have happened in the mean time: A technical specification (TS 25755) edited by JeanHeyd Meneide is now complete and moves through ISO’s complicated publication procedures. Both gcc and clang communities have worked on integrating this feature into their C implementations. I have not yet got my hands on the gcc implementation (but this is also less urgent, see below), but I have been able to use clang’s which is available starting with clang-22. I think that with this in mind everybody developing in C could and should now seriously consider switching to defer for their cleanup handling: no more resource leakage or blocked mutexes on rarely used code paths, no more spaghetti code just to cover all possibilities for preliminary exits from functions. I am not sure if the compiler people are also planning back ports of these features, but with some simple work around and slightly reduced grammar for the defer feature this works for me from gcc-9 onward and for clang-22 onward: #if __has_include() # include # if defined(__clang__) # if __is_identifier(_Defer) # error “clang may need the option -fdefer-ts for the _Defer feature” # endif # endif #elif __GNUC__ > 8 # define defer _Defer # define _Defer _Defer_A(__COUNTER__) # define _Defer_A(N) _Defer_B(N) # define _Defer_B(N) _Defer_C(_Defer_func_ ## N, _Defer_var_ ## N) # define _Defer_C(F, V) \ auto void F(int*); \ __attribute__((__cleanup__(F), __deprecated__, __unused__)) \ int V; \ __attribute__((__always_inline__, __deprecated__, __unused__)) \ inline auto void F(__attribute__((__unused__)) int*V) #else # error “The _Defer feature seems not available” #endif So this is already a large panel of compilers. Obviously it depends on your admissible compile platforms whether or not these are sufficient for you. In any case, with these you may compile for a very wide set of installs since defer does not need any specific software infrastructure or library once the code is compiled. As already discussed many times, the gcc fallback uses the so-called “nested function” feature which is always subject of intense debate and even flame wars. Don’t worry, the implementation as presented here, even when compiled with no optimization at all, does not produce any hidden function in the executable, and in particular there is no “trampoline” or whatever that would put your execution at risk of a stack exploit. You may also notice that there is no fallback for older clang version. This is because their so-called “blocks” extension cannot easily be used as a drop-in to replace nested function: their semantics to access variables from the surrounding scope are different and not compatible with the defer feature as defined by TS 25755

Source: Hacker News | Original Link

“Made in EU” – it was harder than I thought. When I decided to build my startup on European infrastructure, I thought it would be a straightforward swap. Ditch AWS, pick some EU providers, done. How hard could it be? Turns out: harder than expected. Not impossible, I did it, but nobody talks about the weird friction points you hit along the way. This is that post. Why bother? Data sovereignty, GDPR simplicity, not having your entire business dependent on three American hyperscalers, and honestly, a bit of stubbornness. I wanted to prove it could be done. The EU has real infrastructure companies building serious products. They deserve the traffic. The stack Here’s what I landed on after a lot of trial, error, and migration headaches. Hetzner handles the core compute. Load balancers, VMs, and S3-compatible object storage. The pricing is almost absurdly good compared to AWS, and the performance is solid. If you’ve never spun up a Hetzner box, you’re overpaying for cloud compute. Scaleway fills the gaps Hetzner doesn’t cover. I use their Transactional Email (TEM) service, Container Registry, a second S3 bucket for specific workloads, their observability stack, and even their domain registrar. One provider, multiple services, it simplifies billing if nothing else. Bunny.net is the unsung hero of this stack. CDN with distributed storage, DNS, image optimization, WAF, and DDoS protection, all from a company headquartered in Slovenia. Their edge network is genuinely impressive and their dashboard is a joy to use. Coming from Cloudflare, I felt at home rather quickly. Nebius powers our AI inference. If you need GPU compute in Europe without sending requests to us-east-1 , they’re one of the few real options. Hanko handles authentication and identity. A German provider that gives you passkeys, social logins, and user management without reaching for Auth0 or Clerk. More on this in the “can’t avoid” section — it doesn’t eliminate American dependencies entirely, but it keeps the auth layer European. Self-hosting: Rancher, my beloved This is where things get fun… and time-consuming. I self-host a surprising amount: Gitea for source control Plausible for privacy-friendly analytics Twenty CRM for customer management Infisical for secrets management Bugsink for error tracking All running on Kubernetes, with Rancher as the glue keeping the whole cluster sane. Is self-hosting more work than SaaS? Obviously. But it means my data stays exactly where I put it, and I’m not at the mercy of a provider’s pricing changes or acquisition drama. For email, Tutanota keeps things encrypted and European. UptimeRobot watches the monitors so I can sleep. The parts that were extra hard Transactional email with competitive pricing. This one surprised me. Sendgrid, Postmark, Mailgun, they all make it trivially easy and reasonably cheap. The EU options exist, but finding one that matches on deliverability, pricing, and developer experience took real effort. Scaleway’s TEM works, but

Source: Hacker News | Original Link

华尔街见闻

Source: Wall Street CN | Original Link

恒生科技指数跌近3%，智谱及MiniMax市值双双逆势突破3000亿港元-36氪 账号设置 我的关注 我的收藏 申请的报道 退出登录 登录 搜索 36氪Auto 数字时氪 未来消费 智能涌现 未来城市 启动Power on 36氪出海 36氪研究院 潮生TIDE 36氪企服点评 36氪财经 职场bonus 36碳 后浪研究所 暗涌Waves 硬氪 氪睿研究院 媒体品牌 企业号 企服点评 36Kr研究院 36Kr创新咨询 企业服务 核心服务 城市之窗 政府服务 创投发布 LP源计划 VClub VClub投资机构库 投资机构职位推介 投资人认证 投资人服务 寻求报道 36氪Pro 创投氪堂 企业入驻 创业者服务 创投平台 AI测评网 首页 快讯 资讯 推荐 财经 AI 自助报道 城市 最新 创投 汽车 科技 专精特新 直播 视频 专题 活动 搜索 寻求报道 我要入驻 城市合作 02 月 20 恒生科技指数跌近3%，智谱及MiniMax市值双双逆势突破3000亿港元 2026-02-20 16:14 分享至 打开微信“扫一扫”，打开网页后点击屏幕右上角分享按钮 2月20日，香港恒生指数收跌1.10%，恒生科技指数收跌2.91%。互联网科技股普跌，百度跌超6%，阿里巴巴跌近5%。半导体股走弱，华虹半导体跌近6%，中芯国际跌超3%。国产AI大模型板块逆势收涨，智谱涨近43%，MiniMax涨超14%，两者市值双双冲破3000亿港元。（财联社） 原文链接 下一篇 美元/日元延续涨势，涨0.3%至155.48日元 2月20日，美元/日元延续涨势，涨0.3%至155.48日元。（每日经济新闻） 12分钟前 24小时热榜 查看更多榜单 谷歌Gemini 3.1 Pro新王登场，一口气手搓Win11操作系统，造出模拟城市app，SVG效果绝了 遇到艰难决策时，一定要用“三选一”法则 谷歌三巨头同台：未来5年，AI 最大变量是什么？ 6小时前 美团怎么了？ 2小时前 《镖人》、袁和平与渐远的江湖 5小时前 Gemini 3.1 Pro 曝光，能力翻倍价格不变，谷歌想重新定义 AI 竞争规则 5小时前 互称“队友”十年，他们却说自己的婚姻很幸福 6小时前 这些 iPhone 小技巧，让你过年出好片睡好觉 2小时前 关于36氪 城市合作 寻求报道 我要入驻 投资者关系 商务合作 关于我们 联系我们 加入我们 36氪欧洲站 36氪欧洲站 36氪欧洲站 Ai产品日报 网络谣言信息举报入口 热门推荐 热门资讯 热门产品 文章标签 快讯标签 合作伙伴 36氪APP下载 iOS & Android 本站由 阿里云 提供计算与安全服务 违法和不良信息、未成年人保护举报电话：010-89650707 举报邮箱：[email protected] 网上有害信息举报 © 2011~ 2026 北京多氪信息科技有限公司 | 京ICP备12031756号-6 | 京ICP证150143号 | 京公网安备11010502057322号 意见反馈 36氪APP 让一部分人先看到未来 36氪 鲸准 氪空间 推送和解读前沿、有料的科技创投资讯 一级市场金融信息和系统服务提供商 聚焦全球优秀创业者，项目融资率接近97%，领跑行业

Source: 36Kr | Original Link

智谱 、MINIMAX市值突破3000亿港元，超越快手、携程-36氪 账号设置 我的关注 我的收藏 申请的报道 退出登录 登录 搜索 36氪Auto 数字时氪 未来消费 智能涌现 未来城市 启动Power on 36氪出海 36氪研究院 潮生TIDE 36氪企服点评 36氪财经 职场bonus 36碳 后浪研究所 暗涌Waves 硬氪 氪睿研究院 媒体品牌 企业号 企服点评 36Kr研究院 36Kr创新咨询 企业服务 核心服务 城市之窗 政府服务 创投发布 LP源计划 VClub VClub投资机构库 投资机构职位推介 投资人认证 投资人服务 寻求报道 36氪Pro 创投氪堂 企业入驻 创业者服务 创投平台 AI测评网 首页 快讯 资讯 推荐 财经 AI 自助报道 四川 最新 创投 汽车 科技 专精特新 直播 视频 专题 活动 搜索 寻求报道 我要入驻 城市合作 02 月 20 智谱 、MINIMAX市值突破3000亿港元，超越快手、携程 2026-02-20 16:17 分享至 打开微信“扫一扫”，打开网页后点击屏幕右上角分享按钮 2月20日，港股开盘后，大模型公司智谱和MINIMAX股价均大幅上涨，创上市以来新高。截至发稿，智谱股价上涨36%至 691港元，MiniMax股价上涨12%至957港元，两家大模型企业市值均突破3000亿，接连超越携程、快手和京东市值，逼近泡泡玛特（3273亿港元）与百度（3500亿港元）。（第一财经） 原文链接 下一篇 恒生科技指数跌近3%，智谱及MiniMax市值双双逆势突破3000亿港元 2月20日，香港恒生指数收跌1.10%，恒生科技指数收跌2.91%。互联网科技股普跌，百度跌超6%，阿里巴巴跌近5%。半导体股走弱，华虹半导体跌近6%，中芯国际跌超3%。国产AI大模型板块逆势收涨，智谱涨近43%，MiniMax涨超14%，两者市值双双冲破3000亿港元。（财联社） 5分钟前 24小时热榜 查看更多榜单 谷歌Gemini 3.1 Pro新王登场，一口气手搓Win11操作系统，造出模拟城市app，SVG效果绝了 遇到艰难决策时，一定要用“三选一”法则 谷歌三巨头同台：未来5年，AI 最大变量是什么？ 6小时前 美团怎么了？ 2小时前 《镖人》、袁和平与渐远的江湖 5小时前 Gemini 3.1 Pro 曝光，能力翻倍价格不变，谷歌想重新定义 AI 竞争规则 5小时前 互称“队友”十年，他们却说自己的婚姻很幸福 6小时前 这些 iPhone 小技巧，让你过年出好片睡好觉 2小时前 关于36氪 城市合作 寻求报道 我要入驻 投资者关系 商务合作 关于我们 联系我们 加入我们 36氪欧洲站 36氪欧洲站 36氪欧洲站 Ai产品日报 网络谣言信息举报入口 热门推荐 热门资讯 热门产品 文章标签 快讯标签 合作伙伴 36氪APP下载 iOS & Android 本站由 阿里云 提供计算与安全服务 违法和不良信息、未成年人保护举报电话：010-89650707 举报邮箱：[email protected] 网上有害信息举报 © 2011~ 2026 北京多氪信息科技有限公司 | 京ICP备12031756号-6 | 京ICP证150143号 | 京公网安备11010502057322号 意见反馈 36氪APP 让一部分人先看到未来 36氪 鲸准 氪空间 推送和解读前沿、有料的科技创投资讯 一级市场金融信息和系统服务提供商 聚焦全球优秀创业者，项目融资率接近97%，领跑行业

Source: 36Kr | Original Link