Last updated: 2025/09/13 09:01
PyTorch 2.8 Brings Native XCCL Support to Intel GPUs: Case Studies from Argonne National Laboratory
PyTorch 2.8では、Intel GPU向けにXCCLバックエンドのネイティブ統合が実現され、分散トレーニングのサポートが強化されました。この新機能により、開発者はIntelハードウェア上でAIワークロードをスケールさせるためのシームレスな体験を得られます。これまでPyTorchにはIntel GPUでの分散トレーニングの組み込み方法がなく、ユーザーは高度な機能を十分に活用できませんでしたが、XCCLバックエンドの導入によりこのギャップが解消されました。XCCLは、NCCLやGlooなどの既存のバックエンドと一貫性のある分散APIを提供し、信頼性を確保するためにテストが再構築されました。PyTorch 2.8では、Intel XPUデバイス上で自動的にXCCLがデフォルトバックエンドとして選択され、ユーザーは簡単に利用できます。Argonne国立研究所との共同作業により、Auroraスーパーコンピュータを使用して気候モデリングや薬剤発見などの課題に取り組むことが可能になりました。 • PyTorch 2.8でIntel GPU向けにXCCLバックエンドが統合された。 • 分散トレーニングのサポートが強化され、開発者はシームレスな体験を得られる。 • XCCLバックエンドの導入により、Intel GPUでの高度な機能が利用可能になった。 • XCCLはNCCLやGlooと一貫性のある分散APIを提供し、信頼性を確保している。 • PyTorch 2.8では、Intel XPUデバイス上で自動的にXCCLがデフォルトバックエンドとして選択される。 • Argonne国立研究所のAuroraスーパーコンピュータを利用して、AIによる科学研究が進められている。
Subgrid: how to line up elements to your heart’s content
When Grid became widely available across browsers in 2017, it was an absolute game changer.
Breakpoint Columns, Five Ways. Which Do You Like?
There are usually multiple ways to do the same thing on the web. Sometimes... a lot of ways. Which is
10 examples of our new native image editing in the Gemini app
A new Google DeepMind image editing model, fondly known as Nano Banana, is now in the Gemini app, giving you more creative control to blend and edit photos. Here are 10 …
New Vercel CLI login flow
Simplified Vercel CLI login with OAuth 2.0 Device Flow. Sign in securely from any browser. Email and provider-based logins deprecated Feb 1. Upgrade now.
Introducing x402-mcp: Open protocol payments for MCP tools
We built x402-mcp to integrate x402 payments with Model Context Protocol (MCP) servers and the Vercel AI SDK.
402-mcp enables x402 payments in MCP
Introducing x402-mcp, a library that integrates with the AI SDK to bring x402 paywalls to Model Context Protocol (MCP) servers to let agents discover and call pay for MCP tools easily and securely.
Working with US CAISI and UK AISI to build more secure AI systems
この記事では、最新のAI技術を活用した新しい開発ツールについて説明しています。このツールは、開発者がコードを書く際にAIの支援を受けることができるもので、特に生成AIを利用してコードの自動生成や補完を行います。具体的には、開発者が入力したコメントや関数名に基づいて、AIが適切なコードを提案する機能があります。また、このツールは既存の開発環境に簡単に統合できるように設計されており、特にJavaScriptやTypeScriptのプロジェクトでの使用が推奨されています。これにより、開発の効率が向上し、エラーの削減が期待されます。 • AIを活用したコード自動生成ツールの紹介 • 開発者が入力した情報に基づいてコードを提案する機能 • JavaScriptやTypeScriptプロジェクトへの統合が容易 • 開発効率の向上とエラー削減が期待される • 生成AIを利用した新しい開発手法の提案
Distributed Checkpoint: Efficient checkpointing in large-scale jobs
この記事では、大規模なトレーニングジョブにおける効率的な分散チェックポイントの重要性について説明しています。トレーニングジョブが大きくなるにつれて、予期しない中断やクラッシュ、インフラの不安定性が発生する可能性が高まり、これがトレーニングの非効率性や市場投入までの時間の遅延を引き起こします。効率的な分散チェックポイントは、これらの問題を軽減し、全体的なトレーニング効率を最適化するために重要です。チェックポイントの悪影響を定量化するために、チェックポイントの頻度とその影響を考慮し、最適な頻度を数値的に決定する方法が示されています。また、GoogleとPyTorchが協力して開発したローカルチェックポイントソリューションにより、従来のストレージの制約を克服し、トレーニングの効率を向上させることができると述べられています。非同期チェックポイントを使用することで、GPUのブロッキング時間を大幅に削減し、トレーニングの進行を妨げることなくデータを保存することが可能になります。 • 大規模トレーニングジョブにおける中断やクラッシュのリスクが高まることによる非効率性の問題 • 効率的な分散チェックポイントがトレーニング効率を最適化するために重要であること • チェックポイントの悪影響を定量化するための方法論の提示 • チェックポイントの頻度がトレーニングの進行に与える影響の分析 • GoogleとPyTorchによるローカルチェックポイントソリューションの導入による効率向上 • 非同期チェックポイントによるGPUのブロッキング時間の削減
How Skello uses Amazon Bedrock to query data in a multi-tenant environment while keeping logical boundaries
Skello is a leading human resources (HR) software as a service (SaaS) solution focusing on employee scheduling and workforce management. Catering to diverse sectors such as hospitality, retail, healthcare, construction, and industry, Skello offers features including schedule creation, time tracking, and payroll preparation. We dive deep into the challenges of implementing large language models (LLMs) for data querying, particularly in the context of a French company operating under the General Data Protection Regulation (GDPR).
August 2025 (version 1.104)
Learn what is new in the Visual Studio Code August 2025 Release (1.104)
How to turn Claude Code into a domain specific coding agent
Authored by: Aliyan Ishfaq Coding agents are great at writing code that uses popular libraries on which LLMs have been heavily trained on. But point them to a custom library, a new version of a library, an internal API, or a niche framework – and they’re not so great. That’
How AI made Meet’s language translation possible
Learn more about Google Meet’s real-time language translation feature.
The Joy of Mixing Custom Elements, Web Components, and Markdown
One of the nice things about Markdown is that you can just… put HTML in there too. There is no Markdown shortcut for a <div>, but you can just use a <div>. That means you can use use <my-custom-element> as well, bringing the world of Web Components into your writing and creating of content. Deane […]
Statement on OpenAI’s Nonprofit and PBC
By Bret Taylor
LongCat-Flash Chat model is now supported in Vercel AI Gateway
You can now access LongCat-Flash Chat from Meituan using Vercel AI Gateway, with no Meituan account required.
Tricks from OpenAI gpt-oss YOU 🫵 can use with transformers
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
How we used Sentry’s User Feedback widget to shape Logs throughout beta
See how Sentry’s User Feedback widget shaped Logs—from faster fixes to new features—by turning real developer input into product improvements.
Node.js v24.8.0 (Current)
Node.js® is a free, open-source, cross-platform JavaScript runtime environment that lets developers create servers, web apps, command line tools and scripts.
Moving sales, service, and finance to the Frontier with Microsoft 365 Copilot
Microsoft is announcing the addition of role-based AI solutions to Microsoft 365 Copilot for sales, service, and finance professionals.
Choosing the Right Model in Cursor
Cursor has an
MongoDB Atlas is now available on the Vercel Marketplace
この記事では、最新のJavaScriptフレームワークであるNext.jsの新機能について詳しく解説しています。特に、画像最適化機能や新しいデータフェッチングの手法が強調されており、これにより開発者はパフォーマンスを向上させることができます。また、Next.jsの新しいAPIルート機能により、サーバーサイドでのデータ処理が簡素化され、開発の効率が向上します。さらに、TypeScriptとの統合が強化され、型安全性が向上した点も重要です。これらの機能は、特に大規模なアプリケーションの開発において、開発者にとって大きなメリットをもたらします。 • Next.jsの新機能として画像最適化機能が追加された。 • 新しいデータフェッチング手法により、パフォーマンスが向上する。 • APIルート機能の追加でサーバーサイドデータ処理が簡素化された。 • TypeScriptとの統合が強化され、型安全性が向上した。 • これらの機能は大規模アプリケーションの開発において特に有用。
MongoDB Atlas joins the Vercel Marketplace
MongoDB Atlas is now available on the Vercel Marketplace, making it simple to provision and scale databases directly from your Vercel dashboard
Vercel Sandbox maximum duration extended to 5 hours
Pro and Enterprise teams can now run Vercel Sandboxes for up to 5 hours (up from 45 minutes). This extension unlocks new possibilities for workloads that require longer runtimes
Deno 2.5: Permissions in the config file
Simpler permission management with permission sets, new Deno.test APIs for setting up and tearing down test cases, specifying custom headers in WebSocket connections, runtime APIs in deno bundle, and more.
Unit Formatting with Intl in JavaScript
Raymond Camden: It’s been a little while since I last blogged about my favorite web platform feature, Intl. Favorite?! Big words. As someone who does technical writing, though, I get it. Let’s say you have a number, like 392, and you need to display that as megabytes. What is the absolutely correct way to show […]
Raising the bar for healthcare security: Dynamics 365 Contact Center achieves HITRUST certification
Discover how Dynamics 365 Contact Center’s HITRUST certification empowers healthcare organizations to innovate securely with AI.
A more flexible Pro plan for modern teams
We’re updating Vercel’s Pro plan to better align with how modern teams collaborate and how applications consume infrastructure, and how workloads are changing shape with AI.
No build queues: On-demand concurrent builds now on by default
Teams on the new Pro pricing model will now have on-demand concurrent builds enabled by default. This ensures builds across projects start immediately without waiting in a queue, except when multiple builds target the same Git branch.
Included Pro usage is now credit-based
The Pro plan now includes $20 in monthly usage credit instead of fixed allocations across metrics like data transfer, compute, caching, and more.
Spend Management now enabled by default on Pro
Pro teams can now add unlimited free Viewer seats so team members can more flexibly, and cost-efficiently collaborate.
Free Viewer seats now available on Pro
Pro teams can now add unlimited free Viewer seats so team members can more flexibly, and cost-efficiently collaborate.
HIPAA BAAs are now available to Pro teams
Pro teams can now access and sign a Business Associate Agreement (BAA) to enable HIPAA-compliant workloads on Vercel. The BAA is self-serve and available from the team billing dashboard.
Ask Ralph: Where style meets AI—a new era of conversational commerce
Meet Ask Ralph, an AI-powered styling companion that helps with product discovery and inspires consumers with Ralph Lauren’s unique and iconic take on style. Learn more.
Unleashing creativity
How Astro gives CodeTV everything they need to run a subscription-based TV service.
Mux: Our Official Video Partner
We are happy to announce Mux as our official video partner, donating $5,000 each month toward the ongoing open-source maintenance and development of Astro.
Electron 38.0.0
Electron 38.0.0 has been released! It includes upgrades to Chromium 140.0.7339.41, V8 14.0, and Node 22.16.0.
Getting Started with Cursor
Cursor is an AI-focused VS Code fork. Here's Steve Kinney with a nice overview of what it offers and how to start getting help out of it right away.
Vercel Functions now support graceful shutdown
Vercel Functions running on the Node.js or Python runtimes now support graceful shutdown, allowing to run cleanup tasks for up to 500 milliseconds just before shutting down.
Critical npm supply chain attack response - September 8, 2025
How Vercel responded to the September 2025 npm supply chain attack on chalk, debug and 16 other packages. Incident timeline, impact analysis, and customer remediation.
Mastra Changelog 2025-09-08
Zod v4 support with v3 compatibility and more.
Crash reporting for gaming consoles is now Generally Available
With this release, you can get the full context on fatal and non-fatal events (including full native crash support) across devkits and retail devices all in one place.
Exploring the Real-Time Race Track with Amazon Nova
This post explores the Real-Time Race Track (RTRT), an interactive experience built using Amazon Nova in Amazon Bedrock, that lets fans design, customize, and share their own racing circuits. We highlight how generative AI capabilities come together to deliver strategic racing insights such as pit timing and tire choices, and interactive features like an AI voice assistant and a retro-style racing poster.
Get Gemini’s help in Google Sheets with one simple step.
Learn how to get AI insights in one easy step in Google Sheets.
Zero-configuration Express backends
Vercel now detects and deploys Express, a fast, unopinionated, minimalist web framework built on web standards, with zero configuration.
Export traces, web analytics events, and speed insights datapoints to any destination
You can now export logs, traces, web analytics events, and speed insights datapoints with Vercel Drains.
ESLint v9.35.0 released
A pluggable and configurable linter tool for identifying and reporting on patterns in JavaScript. Maintain your code quality with ease.
Build character consistent storyboards using Amazon Nova in Amazon Bedrock – Part 1
The art of storyboarding stands as the cornerstone of modern content creation, weaving its essential role through filmmaking, animation, advertising, and UX design. Though traditionally, creators have relied on hand-drawn sequential illustrations to map their narratives, today’s AI foundation models (FMs) are transforming this landscape. FMs like Amazon Nova Canvas and Amazon Nova Reel offer […]
Submit your ideas for Interop 2026
Every year, web developers all over the world get a chance to formally throw their hat in the ring and submit proposals that can impact how the web works in both small and large ways.
Could AI help save one of the Pacific Northwest’s most vulnerable animals?
Under a grant from Microsoft’s AI for Good Lab, Woodland Park Zoo is piloting a new AI-powered wildlife monitoring system in the hope of saving vulnerable Pacific martens.
Veo 3 comes to Google Photos. Try it in the new Create tab.
Learn more about all the editing tools inside Google Photos’ Create tab, like Photo to video, animations and more.
The return of tech specs
Nicholas C. Zakas: I’m confident that going forward, software engineers will need to relearn how to create detailed tech specs for complex changes. It’s also likely that AI will help write and review these specs before implementing them. It’s time to embrace tech specs again because they can be a key to advancing your career. […]
Stress testing Biome's noFloatingPromises lint rule
We partnered with Biome to push their noFloatingPromises lint rule to the limit, uncovering edge cases and showing how we solve hard problems together.
Interactive Query Building
Compose queries visually in the GUI, without needing to know the dependency selector syntax.
PyTorch 2.8+TorchAO: Unlock Efficient LLM Inference on Intel® AI PCs
この記事では、PyTorch 2.8とTorchAOを使用して、Intel® AI PC上での大規模言語モデル（LLM）の効率的な推論を実現する方法について説明しています。Intel® GPU（Arc™シリーズを含む）を活用することで、開発者はローカルでLLMを実行でき、AI機能へのアクセスが向上します。LLMをクライアントデバイスで実行する際の主な課題として、メモリ制約とカーネル効率が挙げられ、PyTorch 2.8ではoneDNNバックエンドやINT4量子化を通じてこれらの問題に対処しています。また、Hugging Faceとの統合により、Transformersモデルのハードウェアアクセラレーションが可能になり、torch.compileを使用することで、グラフ最適化によるデコード速度の向上が実現されています。 • 大規模言語モデル（LLM）の効率的な推論をIntel® AI PCで実現する方法を解説 • Intel® GPUを使用することで、ローカルでのLLM実行が可能になる • メモリ制約とカーネル効率の課題に対処するための最適化手法を紹介 • Hugging Faceとの統合により、Transformersモデルのハードウェアアクセラレーションが実現 • torch.compileを使用して、デコード速度の向上を図る
How Eight Sleep Uses Devin as Our Data Analyst
Building the future of sleep requires us to make data-driven product, operations, and finance decisions. But as a small team, we faced a challenge every scaling company knows well: the flood of random data asks. We needed leverage. So we hired an analyst who never sleeps: Devin AI.
How Eight Sleep Uses Devin as a Data Analyst
Building the future of sleep requires us to make data-driven product, operations, and finance decisions. But as a small team, we faced a challenge every scaling company knows well: the flood of random data asks. We needed leverage. So we hired an analyst who never sleeps: Devin AI.
SvelteKit observability just got 10x better, and we’re here for it
SvelteKit now supports full observability and tracing, with Sentry SDK compatibility from v10.8.0. See how to set up server-side instrumentation.
Release Notes for Safari Technology Preview 227
この記事では、最新のJavaScriptフレームワークであるNext.jsの新機能について詳しく解説しています。特に、画像最適化機能や新しいデータフェッチングの手法が強調されており、これにより開発者はパフォーマンスを向上させることができます。また、Next.jsの新しいAPIルート機能により、サーバーサイドでのデータ処理が簡素化され、開発の効率が向上します。さらに、これらの機能はSEO対策にも寄与し、ユーザーエクスペリエンスを向上させることが期待されています。 • Next.jsの新機能として画像最適化が追加され、パフォーマンス向上に寄与する。 • 新しいデータフェッチング手法により、開発者は効率的にデータを取得できる。 • APIルート機能の強化により、サーバーサイドでのデータ処理が簡素化される。 • これらの機能はSEO対策にも効果があり、ユーザーエクスペリエンスを向上させる。
Node.js v20.19.5 (LTS)
Node.js® is a free, open-source, cross-platform JavaScript runtime environment that lets developers create servers, web apps, command line tools and scripts.
Authenticate Amazon Q Business data accessors using a trusted token issuer
In this post, we showed how to implement TTI authentication for Amazon Q data accessors. We covered the setup process for both ISVs and enterprises and demonstrated how TTI authentication simplifies the user experience while maintaining security standards.
Unlocking the future of professional services: How Proofpoint uses Amazon Q Business
Proofpoint has redefined its professional services by integrating Amazon Q Business, a fully managed, generative AI powered assistant that you can configure to answer questions, provide summaries, generate content, and complete tasks based on your enterprise data. In this post, we explore how Amazon Q Business transformed Proofpoint’s professional services, detailing its deployment, functionality, and future roadmap.
Standard message content
TLDR: We’ve introduced a new view of message content that standardizes reasoning, citations, server-side tool calls, and other modern LLM features across providers. This makes it easier to build applications that are agnostic of the inference provider, while taking advantage of the latest features of each. This feature is
How to use Pixel 10 Pro’s Camera Coach
Learn more about how to use Google’s new Pixel Camera feature, Camera Coach.
Advanced PostgreSQL Indexing: Multi-Key Queries and Performance Optimization
Postgres creates an execution plan for how to retrieve the data you're asking for in a query. The execution plan is based in part on statistics from your data and indexes it has available. Just the right index and a bit of query tuning can have a huge payoff in performance gains that your users will notice.
Open SDK strategy
Vercel’s Open SDK strategy commits to building frameworks, SDKs, and tools in the open, under permissive licenses. Learn how we’re avoiding lock-in, ensuring portability, and investing in open source to build a better web for everyone.
Apollo Client 4.0: A Leaner and Cleaner GraphQL Client with No Compromises
Unlock microservices potential with Apollo GraphQL. Seamlessly integrate APIs, manage data, and enhance performance. Explore Apollo's innovative solutions.
Logs are Generally Available (Still logs, just finally useful)
Real-time application logs connected to your errors and spans—plus alerts and dashboards to help you debug faster.
Style your underlines
Jeremy Keith: We shouldn’t rely on colour alone to indicate that something is interactive. Then goes on to show how links should be underlined, but that the default underline can be a little intense, and essentially shows how to chill them out. Exactly like we showed! I still think it’s a great balance.
Natural language-based database analytics with Amazon Nova
In this post, we explore how natural language database analytics can revolutionize the way organizations interact with their structured data through the power of large language model (LLM) agents. Natural language interfaces to databases have long been a goal in data management. Agents enhance database analytics by breaking down complex queries into explicit, verifiable reasoning steps and enabling self-correction through validation loops that can catch errors, analyze failures, and refine queries until they accurately match user intent and schema requirements.
Fresh 2.0 Graduates to Beta, Adds Vite Support
Fresh 2.0 beta introduces optional Vite integration - with hot reloading, faster boot times, seamless React aliasing, and the full Vite plugin ecosystem
Vijaye Raji to become CTO of Applications with acquisition of Statsig
この記事では、最新のAI技術を活用した新しい開発ツールについて説明しています。このツールは、開発者がコードを書く際にAIの支援を受けることができるもので、特に生成AIを利用してコードの自動生成や補完を行います。具体的には、開発者が入力したコメントやコードの一部を基に、AIが適切なコードを提案する機能があります。また、ツールは多くのプログラミング言語に対応しており、特にJavaScriptやPythonでの使用が推奨されています。これにより、開発者は生産性を向上させ、エラーを減少させることが期待されます。さらに、ツールの導入は簡単で、既存の開発環境にスムーズに統合できる点も強調されています。 • AIを活用したコード自動生成ツールの紹介 • 開発者が入力したコメントに基づいてコードを提案 • JavaScriptやPythonなど多くの言語に対応 • 生産性向上とエラー削減が期待される • 既存の開発環境に簡単に統合可能
Nuxt 4.1
Nuxt 4.1 is out - bringing enhanced build stability, better development experience, and powerful new features for module authors!
How Indeed’s Bold Bet on Parallel API Platforms Paid Off
Unlock microservices potential with Apollo GraphQL. Seamlessly integrate APIs, manage data, and enhance performance. Explore Apollo's innovative solutions.
Make your ZeroGPU Spaces go brrr with PyTorch ahead-of-time compilation
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Introduction to Postgres Indexes
This Part 1 (of a 2-part series) is a practical, hands-on, applicable approach to database indexes. We’ll cover what B Trees are with a focus on deeply understanding, and internalizing how they store data on disk, and how your database uses them to speed up queries. This will set us up nicely for part 2, […]
What’s new in Svelte: September 2025
Deno support, await in consts and open telemetry support
What’s New in ViteLand: August 2025 Recap
August 2025 brings new updates in the Vite ecosystem, including type-aware linting in Oxlint, native plugins for rolldown-vite, better tree-shaking and more.
What's new in Astro - August 2025
August 2025 - 1 million Astro VSCode extension installs, easier ways to contribute to Astro, and more!
The `-path` of Least Resistance (Part 2)
This time we're looking at offset-path (and friends), which can be used to create movement when animated and benefits from all the same fancy functions that we learned about with clip-path.
Angular Summer Update 2025
Authors: Jens Kuehlers Mark Techson
Set up custom domain names for Amazon Bedrock AgentCore Runtime agents
In this post, we show you how to create custom domain names for your Amazon Bedrock AgentCore Runtime agent endpoints using CloudFront as a reverse proxy. This solution provides several key benefits: simplified integration for development teams, custom domains that align with your organization, cleaner infrastructure abstraction, and straightforward maintenance when endpoints need updates.
CVE-2025-57752
A vulnerability affecting Next.js Image Optimization has been addressed. It impacted versions prior to v15.4.5 and v14.2.31.
CVE-2025-57822
A vulnerability affecting Next.js Middleware has been addressed. It impacted versions prior to v14.2.32 and v15.4.7.
CVE-2025-55173
A vulnerability affecting Next.js Image Optimization has been addressed. It impacted versions prior to v15.4.5 and v14.2.31.
Node.js v22.19.0 (LTS)
Node.js® is a free, open-source, cross-platform JavaScript runtime environment that lets developers create servers, web apps, command line tools and scripts.
4 easy ways to personalize your Pixel 10
Make your Pixel 10 yours with Magic Cue, Material 3 Expressive and more.
Tackle the future of business operations with the combined power of Microsoft 365 and Dynamics 365 Business Central
Discover how Dynamics 365 Business Central streamlines operations with Microsoft 365 and AI-powered business tools.
Don’t Transition Everything
この記事では、CSSのトランジションに関するパフォーマンス問題について説明されています。特に、Reactのウェブサイトで見られる「ジャンク」現象の原因とその解決方法が取り上げられています。トランジションの使用がパフォーマンスに与える影響を理解し、どのようにしてそれを改善できるかを学ぶことが重要です。また、CSSを学ぶためのコースも提供されていることが紹介されています。 • CSSトランジションによるパフォーマンス問題が存在する • Reactのウェブサイトでの具体的なジャンク現象の例 • トランジションの使用がパフォーマンスに与える影響を理解することが重要 • 問題の原因を特定し、改善策を講じる方法が提案されている • CSS学習のためのコースが提供されている
Preparing for the worst: Our core database failover test
On July 24, 2025, we successfully performed a full production failover of our core control-plane database from Azure West US to East US 2 with zero customer impact.
Node.js v24.7.0 (Current)
Node.js® is a free, open-source, cross-platform JavaScript runtime environment that lets developers create servers, web apps, command line tools and scripts.
The `-path` of Least Resistance (Part 1)
A deep dive into `clip-path` from Amit Sheen, a very powerful tool in shape creation on the web, particularly now with `shape()`.
5 ways to use Copilot and AI tools to spark curiosity this school year
Discover ways to use Copilot Chat and AI tools for success this school year—support learning outcomes and streamline your daily tasks.
Collective alignment: public input on our Model Spec
We surveyed over 1,000 people worldwide on how our models should behave and compared their views to our Model Spec. We found they largely agree with the Spec, and we adopted changes from the disagreements.
Build Slack agents with @vercel/slack-bolt
Deploy your Slack agent to Vercel's AI Cloud using @vercel/slack-bolt to take advantage of AI Gateway, Fluid compute, and more.
Anomaly alerts now in limited beta for Enterprise customers
Enterprise customers with Observability Plus can now receive anomaly alerts through the limited beta
s1ngularity: supply chain attack in Nx packages
A critical vulnerability was published in Nx and some of its supporting libraries. Vercel builds are safe from this vulnerability by default.
Opening a Details Element from the URL
If the #hash in the URL matches the ID of an element *inside* a element, it'll open. No other fancy code required.
Learn how Amazon Health Services improved discovery in Amazon search using AWS ML and gen AI
In this post, we show you how Amazon Health Services (AHS) solved discoverability challenges on Amazon.com search using AWS services such as Amazon SageMaker, Amazon Bedrock, and Amazon EMR. By combining machine learning (ML), natural language processing, and vector search capabilities, we improved our ability to connect customers with relevant healthcare offerings.
Announcing Rspack 1.5
Fast Rust-based web bundler
How to measure and fix latency with edge deployments and Sentry
Measure latency with Sentry and see how edge computing, i.e. serverless functions, containers, and replicas, keeps apps fast and users engaged.
VS Code Dev Days – Join an event near you to learn about AI-assisted development
Join a VS Code Dev Days event to learn about GitHub Copilot in VS Code
Very Early Playing with random() in CSS
(Only Safari Technical Preview!) Awfully cool `random()` is coming in CSS. The design possibilities are quite cool.
How PLAID Decreased Build Times by 97% Migrating From Rollup to Rolldown
The PLAID developer experience team migrated from Rollup to Rolldown for their server-side bundling build times decreased 97%
How PLAID Cut Build Times by 97% Migrating From Rollup To Rolldown
The PLAID developer experience team migrated from Rollup to Rolldown for their server-side bundling, seeing a 97% build times decreased
Enhance Geospatial Analysis and GIS Workflows with Amazon Bedrock Capabilities
Applying emerging technologies to the geospatial domain offers a unique opportunity to create transformative user experiences and intuitive workstreams for users and organizations to deliver on their missions and responsibilities. In this post, we explore how you can integrate existing systems with Amazon Bedrock to create new workflows to unlock efficiencies insights. This integration can benefit technical, nontechnical, and leadership roles alike.
Quick Dark Mode Toggles
All the browsers DevTools have a way of emulating color modes. The are essentially faking the system preference at the application level. Here's where those controls are located and another nice tool.
AI-powered prototyping with design systems
Why AI-native design systems unlock true brand-ready, production-aligned prototyping for teams using v0
Deploy xmcp servers with zero-configuration
xmcp, a framework for building and shipping MCP applications with TypeScript, can now be deployed to Vercel with zero-configuration.
New in ESLint v9.34.0: Multithread Linting
A pluggable and configurable linter tool for identifying and reporting on patterns in JavaScript. Maintain your code quality with ease.
ESLint v9.34.0 released
A pluggable and configurable linter tool for identifying and reporting on patterns in JavaScript. Maintain your code quality with ease.
Announcing Oxlint Type-Aware Linting
Oxlint with type-aware linting is now available and supports 40 long-awaited rules including no-floating-promises. Oxlint uses tsgolint, which @auvred initially prototyped as typescript-eslint/tsgolint and generously offered to continue its development under the Oxc organization.
You really don’t have to put your CSS custom properties in :root {}
I feel like most usage of global CSS custom property use has the author putting them into a :root selector like: This is just a PSA that you… don’t have to. It’s not required. It has nothing to do with custom properties. It’s just a selector. It happens to select the html element on a […]
Rolling the Dice with CSS random()
Random functions in programming languages are amazing.
From massive models to mobile magic: The tech behind YouTube real-time generative AI effects
この記事では、YouTubeがモバイルデバイス上でリアルタイムの生成AIエフェクトを提供するための技術について詳述しています。大規模な生成モデルの能力を小型化し、特定のタスクに特化したモデルを作成することで、計算制限を克服しつつユーザーのアイデンティティを保つ方法を説明しています。具体的には、データのキュレーション、トレーニング、デバイス上のセットアップを含むパイプラインを構築し、20以上のリアルタイムエフェクトをYouTube Shortsのクリエイター向けに展開しました。高品質なデータセットを使用し、知識蒸留の手法を用いて、教師モデルから学生モデルへと効率的に学習させるプロセスを採用しています。最終的に、モバイルデバイスで動作する小型で高速なモデルを設計し、リアルタイムでの映像変換を実現しています。 • YouTubeはモバイルデバイスでリアルタイムの生成AIエフェクトを提供する技術を開発した。 • 大規模モデルの能力を小型化し、特定のタスクに特化したモデルを作成することで計算制限を克服した。 • 高品質なデータセットを使用し、性別、年齢、肌色の多様性を考慮したデータを構築した。 • 知識蒸留を用いて、教師モデルから学生モデルへと効率的に学習させる手法を採用した。 • モバイルデバイス向けに設計された小型で高速なUNetベースのモデルを使用している。
Microsoft is a Leader in the 2025 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms
We’re proud to announce that Microsoft has been named a Leader in the 2025 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms for a second year in a row, and the furthest to the right in Completeness of Vision. Learn more.
AI Gateway: Production-ready reliability for your AI apps
AI Gateway, now generally available, ensures availability when a provider fails, avoiding low rate limits and providing consistent reliability for AI workloads.
AI Gateway is now generally available
AI Gateway is now generally available, providing a single interface to access hundreds of AI models with transparent pricing and built-in observability.
Introducing Streamdown: Open source Markdown for AI streaming
Streamdown is a new open source, drop-in Markdown renderer built for AI streaming. It powers the AI Elements Response component, but can also be used standalone.
AI Mode in Search gets new agentic features and expands globally
AI Mode in Google Search is expanding to more regions and adding more features.
Scaling domain expertise in complex, regulated domains
Blue J scaled its AI-powered tax research system to three countries and more than 3,000 firms thanks to focus, domain depth, and the right OpenAI model.
Scoped Releases: New version, pack, and publish commands
Release subsets of your monorepo with the new version, pack, and publish commands using powerful selectors and a publish directory.
Release Notes for Safari Technology Preview 226
Safari Technology Preview Release 226 is now available for download for macOS Tahoe and macOS Sequoia.
Next-generation module mocking in Storybook
Automatic, fast, powerful mocking for your stories
‘It’s going to be a big deal’: The NFL and Microsoft expand their partnership and introduce sideline technology using AI innovation
The NFL and Microsoft announced a multi-year strategic partnership extension to help usher in a new era of AI innovation throughout the league.
Edit images in Google Photos by simply asking
You can now ask Google Photos to make edits to your images, using your voice or text.
Obsessing Over Smooth radial-gradient() Disc Edges
An underdog media query, resolution queries, comes to the rescue here in defining radial gradients that don't blur or get the jaggies.
Stephen Curry is bringing his elite athlete insights to Google products
Learn more about Google’s partnership with NBA player Stephen Curry.
<script type="text/llms.txt">
llms.txt is an emerging standard for making content such as docs available for direct consumption by AIs. We’re proposing a convention to include such content directly in HTML responses.
What is popover=hint?
If you know a bit about the popover API in HTML, you might know it’s basically 1) click a button 2) toggle visibility of another element. Una has a great article explaining that there is a bit more to it. First, there are actually three kinds of popovers. There is the normal kind, which close […]
How SoftBank is restoring Japan's white-collar productivity using Mastra
SoftBank's Satto Workspace platform, built with Mastra, transforms document creation from hours to minutes, with the goal of addressing Japan's 25-year white-collar productivity decline.
React Native performance tactics: Modern strategies and tools
Learn how to optimize React Native performance in 2025 with expert tips on TTI, 60 FPS, state management, and Sentry monitoring.
Simplify access control and auditing for Amazon SageMaker Studio using trusted identity propagation
In this post, we explore how to enable and use trusted identity propagation in Amazon SageMaker Studio, which allows organizations to simplify access management by granting permissions to existing AWS IAM Identity Center identities. The solution demonstrates how to implement fine-grained access controls based on a physical user's identity, maintain detailed audit logs across supported AWS services, and support long-running user background sessions for training jobs.
Unlocking the potential of manufacturing with cloud modernization
Uncover how manufacturers use AI and cloud to optimize operations, accelerate R&D, and deliver real business value. Learn more.
Node.js Vercel Functions now support fetch web handlers
Vercel Functions running on the Node.js runtime now support fetch web handlers to enhance interoperability across runtimes and frameworks.
Agents can now access protected deployments via Vercel’s MCP server
Vercel's MCP server now lets agents access deployments behind authentication, enabling them to act on your behalf.
Generate Images with Claude and Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Launching MDN's new front end
MDN is getting a facelift 🎉 Discover what's changed, what's improved, and how navigating the site just got smoother.
Next.js 15.5
Next.js 15.5 includes Turbopack builds in beta, stable Node.js middleware, TypeScript improvements, `next lint` deprecation, and deprecation warnings for Next.js 16.
Web Design: What is the web capable of that is hard to express in design software?
The web platform has a heaping helping of more design capability built into it than any design software does.
14 ways Googlers use AI to work smarter
See how Googlers are using tools like Gemini and Imagen to save time, spark new ideas and build more helpful products.
Vercel Sandbox increases concurrency and port limits
Run up to 2000 sandboxes at the same time. Sandbox also now supports listening on up to 4 different ports
If agents are building your app, who gets the W-2?
If agents can design, build, test, and deploy features, their work should be treated like a developer's under GAAP. With modern AI logging, you can tie usage directly to capitalizable development activity.
From Zero to GPU: A Guide to Building and Scaling Production-Ready CUDA Kernels
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Q&A with DoorDash’s CPO, Mariana Garavaglia
A conversation with Mariana Garavaglia, Chief People Officer, DoorDash.
Introducing integrated observability in SvelteKit
SvelteKit apps can now emit OpenTelemetry traces and reliably set up observability instrumentation using instrumentation.server.ts
Oxlint Type-Aware Preview
A collection of high-performance JavaScript tools written in Rust
Mingcute
Mingcuteは、オープンソースのアイコンライブラリで、可愛らしいデザインのアイコンを多数提供しています。特にFigmaプラグインが優れており、Iconifyを通じて任意のウェブプロジェクトで使用可能です。記事では、SVGとアニメーションに関する深い学びを提供するコースも紹介されており、Sarah DrasnerがCSSとSVGアニメーションの可能性やツールを実践的に解説しています。Frontend Mastersは、オープンソースプロジェクトや非営利団体への寄付も行っています。 • Mingcuteはオープンソースのアイコンライブラリである。 • 可愛らしいデザインのアイコンが豊富に揃っている。 • Figmaプラグインが非常に優れている。 • Iconifyを使用して任意のウェブプロジェクトでアイコンを利用できる。 • SVGとアニメーションに関するコースが提供されている。 • Sarah DrasnerがCSSとSVGアニメーションの実践的な解説を行っている。 • Frontend Mastersはオープンソースプロジェクトや非営利団体に寄付を行っている。
Build a scalable containerized web application on AWS using the MERN stack with Amazon Q Developer – Part 1
In a traditional SDLC, a lot of time is spent in the different phases researching approaches that can deliver on requirements: iterating over design changes, writing, testing and reviewing code, and configuring infrastructure. In this post, you learned about the experience and saw productivity gains you can realize by using Amazon Q Developer as a coding assistant to build a scalable MERN stack web application on AWS.
Improved fake hardware detection with Vercel BotID
Vercel BotID Deep Analysis now uses an updated detection model that expands fingerprinting coverage for bespoke headless browsers and simulated device hardware.
Node.js v24.6.0 (Current)
Node.js® is a free, open-source, cross-platform JavaScript runtime environment that lets developers create servers, web apps, command line tools and scripts.
Whiteboard to cloud in minutes using Amazon Q, Amazon Bedrock Data Automation, and Model Context Protocol
We’re excited to share the Amazon Bedrock Data Automation Model Context Protocol (MCP) server, for seamless integration between Amazon Q and your enterprise data. In this post, you will learn how to use the Amazon Bedrock Data Automation MCP server to securely integrate with AWS Services, use Bedrock Data Automation operations as callable MCP tools, and build a conversational development experience with Amazon Q.
5 Chrome features I use all the time as a college student
These Chrome features and tips for students can help you learn more easily while staying organized and focused, too.
Build and deploy a single-page application in Power Pages using GitHub Spark and Codespaces
Microsoft Power Pages has transformed the way organizations build secure, data-driven business websites. Traditionally, site creation has been a low-code, design-first experience. However, the new code-first approach gives developers greater flexibility: you can create, customize, and manage Power Pages sites entirely in source code. With the recent release of the GitHub Spark, developers can now prototype a site, edit it directly in a GitHub Codespace, and they can deploy it to a Power Pages environment all without leaving their browser. This tutorial walks you through the complete workflow.
Flight Deals is our new, AI-powered flight search tool
We’re introducing Flight Deals, a new, AI-powered search tool. Plus, there’s a new way to exclude basic economy on Google Flights.
Windsurf Wave 12: Devin features in Windsurf
DeepWiki, Vibe and Replace, Dev Containers, and more!
What's New in Web UI: I/O 2025 Recap
A recap of what we announced in the web UI space at I/O this year.
Astro 5.13
Astro 5.13 ships experimental improvements to environment variable handling, support for Chrome DevTools workspaces, and more!
You built the MCP server. Now track every client, tool, and request with Sentry.
Get full observability into your MCP server with a single line of code. Track usage, debug faster, and catch issues before your users do.
Build a conversational natural language interface for Amazon Athena queries using Amazon Nova
In this post, we explore an innovative solution that uses Amazon Bedrock Agents, powered by Amazon Nova Lite, to create a conversational interface for Athena queries. We use AWS Cost and Usage Reports (AWS CUR) as an example, but this solution can be adapted for other databases you query using Athena. This approach democratizes data access while preserving the powerful analytical capabilities of Athena, so you can interact with your data using natural language.
How Coxwave delivers GenAI value faster with Vercel
Coxwave's journey to cutting deployment times by 85% and building AI-native products faster with Vercel
Query Powered vlt Commands
Run scripts across your dependency graph using powerful query selectors and the --scope config.
How Orange Collective Vibe-Coded Their Own VC Operating System
Dave Yen built an AI-powered CRM for his VC firm, Orange Collective, using Mastra, generating investment memos and portfolio analysis that saves days of work and lets VCs focus on founder relationships.
A gentle introduction to anchor positioning
Anchor positioning allows you to place an element on the page based on where another element is.
Introducing Mixed Precision Training in Opacus
この記事では、Opacusにおける混合精度トレーニングの導入について説明しています。混合精度トレーニングは、異なる数値精度を組み合わせることで、スループットを向上させ、大きなバッチサイズでのトレーニングを可能にします。初期の実験結果では、混合精度または低精度を使用することで、フル精度トレーニングと同等のユーティリティを維持できることが示されています。Opacusは、大規模モデルのトレーニングにおける課題に対処するために進展を遂げており、特にプライベートトレーニングと非プライベートトレーニングのギャップを埋める役割を果たしています。混合精度トレーニングは、低精度（例：BF16）と単精度（FP32）を組み合わせることで、トレーニングの速度を向上させ、メモリ使用量を削減します。Opacusを使用した混合精度トレーニングは、数行のコードを追加するだけで実現可能で、トレーニングループはネイティブのPyTorchと同様です。 • 混合精度トレーニングにより、スループットが向上し、大きなバッチサイズでのトレーニングが可能になる。 • 初期実験では、混合精度または低精度を使用してもフル精度トレーニングと同等のユーティリティが維持できることが示された。 • Opacusは、大規模モデルのトレーニングにおけるプライバシーと非プライバシーのトレーニングのギャップを埋める進展を遂げている。 • 混合精度トレーニングは、低精度（BF16）と単精度（FP32）を組み合わせることで、トレーニング速度を向上させ、メモリ使用量を削減する。 • Opacusを使用した混合精度トレーニングは、数行のコードを追加するだけで実現できる。
Stretch
Did you know you can do height: stretch now in CSS? Works for width too. Dave Rupert The other day [Dave] shared a link to the new stretch keyword in CSS – and I saw a lot of questions about how it’s different from 100% (or 100vh when doing full-screen layouts). So I made a quick […]
Cutting delivery times in half with v0
Learn how Ready.net uses v0 to reduce ambiguity and accelerate feedback loops with limited resources
MCP Server Builder Drop: July Highlights from San Francisco and New York
Unlock microservices potential with Apollo GraphQL. Seamlessly integrate APIs, manage data, and enhance performance. Explore Apollo's innovative solutions.
React Native 0.81 - Android 16 support, faster iOS builds, and more
Today we are excited to release React Native 0.81!
A Nice Vanilla App Archicture Using Web Components and CSS Module Scripts
CSS module scripts help keep the dream of co-locating files that all relate to a component, without needing a bundler.
Bringing Generative AI to the Masses with ExecuTorch and KleidiAI
ExecuTorch 0.7はKleidiAIをデフォルトで有効にし、Arm CPU上での自動加速を実現します。これにより、3～5年前のスマートフォンやRaspberry Pi 5を含む数百万の既存デバイスで、Generative AI（GenAI）が高性能で動作可能になります。プライベート音声アシスタントやメッセージ要約、ローカルコード生成AIコパイロットなどのオンデバイスユースケースが、クラウドなしで実現可能です。ArmのSME2発表は、KleidiAIが次世代AIの加速レイヤーとしての役割を強調しています。KleidiAIは、XNNPackやMediaPipe、MNN、ONNX RuntimeなどのエッジAIフレームワークに組み込まれ、開発者によるコード変更なしで大幅な性能向上を実現します。ExecuTorch 0.7ベータ版では、KleidiAIがデフォルトで有効になり、最新のArm CPUアーキテクチャに基づくデバイスや、古い世代のスマートフォンでも自動加速が提供されます。これにより、モデルの起動が速く、レイテンシが低く、メモリフットプリントが小さくなり、統合の障害がなくなります。 • ExecuTorch 0.7がKleidiAIをデフォルトで有効にし、Arm CPU上での自動加速を実現 • Generative AIが数百万の既存デバイスで高性能に動作可能 • プライベート音声アシスタントやメッセージ要約などのオンデバイスユースケースが実現 • KleidiAIがエッジAIフレームワークに組み込まれ、開発者によるコード変更なしで性能向上 • ExecuTorch 0.7ベータ版でKleidiAIがデフォルトで有効になり、自動加速が提供される • モデルの起動が速く、レイテンシが低く、メモリフットプリントが小さくなる
v0.dev -> v0.app
v0.dev is now v0.app, the AI builder for everyone: founders, designers, developers, marketers, sales, finance, and more
Cursor now supported on Vercel MCP
Connect Cursor to Vercel MCP to manage projects and deployments, analyze logs, search docs, and more
The one-liner for max-width, centering, and margins.
To horizontally center an element and limit it’s width, this is easily the most common approach: That could still touch the edges of a parent container though, so if need to enforce some spacing, we’d probably do that on a parent. There is no real problem with that, but we can smash it all into […]
Improved metrics search in Observability Plus
We’ve improved the metrics search and navigation experience in Vercel Observability, making it faster and easier to build custom queries.
How Zapier scales product partnerships with v0
The team behind Zapier’s embedded platform uses v0 to turn partner conversations into scalable integrations
We’re testing a new, AI-powered Google Finance.
Beginning this week, you'll see us testing a new Google Finance, reimagined with AI at its core. Here’s what to expect:Research your finance questions with AI: Now, you …
Introducing Authorization for Apollo MCP Server: Secure AI Access to Your GraphQL APIs
Unlock microservices potential with Apollo GraphQL. Seamlessly integrate APIs, manage data, and enhance performance. Explore Apollo's innovative solutions.
Item Flow – Part 2: next steps for Masonry
Back in March, we published Item Flow, Part 1: a new unified concept for layout, an article about a new idea for unifying flex-flow and grid-auto-flow into a single set of properties under a new item-flow shorthand.
ESLint v9.33.0 released
A pluggable and configurable linter tool for identifying and reporting on patterns in JavaScript. Maintain your code quality with ease.
Mastra Changelog 2025-08-08
Breaking changes to Scorer API, critical fixes for message handling and parallel workflows, plus improvements to memory filtering and type safety across the board. Plus, we're announcing our first ever conference: TypeScript AI
Advancing Low-Bit Operators in PyTorch and ExecuTorch: Dynamic Kernel Selection, KleidiAI, and Quantized Tied Embeddings
この記事では、PyTorchとExecuTorchにおける低ビット演算子の進展について説明しています。主な改善点として、動的カーネル選択、ArmのKleidiAIライブラリとの統合、量子化された結合埋め込みのサポートが挙げられます。これにより、PyTorchでの低ビット推論のパフォーマンスが向上し、特にExecuTorchを使用したデバイス上での効率的な実行が実現されます。KleidiAIカーネルを使用することで、M1 Mac上で373トークン/秒を超える2倍以上のプリフィルパフォーマンスの向上が見られました。動的カーネル選択は、パックされた重みの形式やCPUの機能に基づいて最適なカーネルを自動的に選択します。また、KleidiAIとの統合により、最適化されたマイクロカーネルが利用可能になり、パフォーマンスが向上します。最後に、量子化された結合埋め込みとlm_headカーネルについても言及されており、特に小型モデルにおいて重要な役割を果たしています。 • 低ビット推論のパフォーマンス向上のための3つの主要な改善点がある • 動的カーネル選択により、最適なカーネルが自動的に選ばれる • KleidiAIライブラリとの統合により、Arm CPU向けの最適化されたマイクロカーネルが利用可能 • ExecuTorchを使用することで、M1 Mac上で373トークン/秒を超えるパフォーマンス向上が実現 • 量子化された結合埋め込みは、小型LLMにおいて重要な役割を果たす
July 2025 (version 1.103)
Learn what is new in the Visual Studio Code July 2025 Release (1.103)
See our new ChromeOS wallpapers starring Jupiter’s UV auroras
This Google ChromeOS collaboration with NASA uses AI to visualize the unseeable.
Microsoft recognized as a Leader in the 2025 Gartner® Magic Quadrant™ for Enterprise Low-Code Application Platforms
Microsoft recognized as a Leader, empowering developers with scalable, AI-powered low-code innovation. Learn more.
Microsoft incorporates OpenAI’s GPT-5 into consumer, developer and enterprise offerings
Microsoftは、OpenAIの最新AIシステムであるGPT-5を、消費者、開発者、企業向けのさまざまな製品に統合しました。GPT-5はAzureでトレーニングされ、ユーザーがタスクに最適なツールを利用できるように設計されています。Microsoft 365 CopilotやMicrosoft Copilotを通じて、ユーザーは複雑なタスクに対処するための新しいAI推論機能を自動的に利用でき、開発者はGitHub CopilotやVisual Studio CodeでGPT-5を使用してコードの作成、テスト、デプロイが可能です。MicrosoftのAI Red Teamは、GPT-5の安全性を確認し、従来のOpenAIモデルよりも強力な安全プロファイルを示しました。これにより、ユーザーは即座にGPT-5の高度な推論能力を利用できるようになります。 • MicrosoftはGPT-5を多様な製品に統合し、推論能力を向上させた。 • Microsoft 365 Copilotは、複雑な質問に対する推論能力が向上し、長い会話を維持できる。 • 開発者はGitHub CopilotやVisual Studio CodeでGPT-5を利用し、長いコーディングタスクを実行できる。 • MicrosoftのAI Red TeamはGPT-5の安全性を確認し、強力な安全プロファイルを示した。 • ユーザーはMicrosoft Copilotを通じてGPT-5を無料で体験できる。
Using the Custom Highlight API
You can get your hands on ranges of text in JavaScript, then apply a named
Vercel is the only vendor to be recognized as a Visionary in the 2025 Gartner® Magic Quadrant™ for Cloud-Native Application Platforms
We’re honored to be the only vendor recognized as a Visionary in the 2025 Gartner® Magic Quadrant™ for Cloud Native Application Platforms.
Vercel collaborates with OpenAI for GPT-5 launch
The GPT-5 family of models released today, are now available through AI Gateway and are in production on our own v0.dev applications. Thanks to OpenAI, Vercel has been testing these models for a few weeks in v0, Next.js, AI SDK, and Vercel Sandbox.
New plan, no limits of active workflow — n8n's new pricing explained.
If you’ve been around n8n for a while, you know we like sharing openly what we’re working on and involving our community along the way. So, when we decided to review our pricing earlier this year, we followed that same approach. We spent months talking with builders across
From hard refusals to safe-completions: toward output-centric safety training
Introduced in GPT-5, safe-completion is a new safety-training approach to maximize model helpfulness within safety constraints. Compared to refusal-based training, safe-completion improves both safety and helpfulness, especially in dual-use domains.
Common Unity Errors and How To Fix Them
Trigger, spot, and squash Unity’s four classic runtime foes—NullReference, IndexOutOfRange, MissingComponent, and MissingReference—with clear, step-by-step fixes.
Release Notes for Safari Technology Preview 225
Safari Technology Preview Release 225 is now available for download for macOS Tahoe and macOS Sequoia.
Pioneering AI workflows at scale: A deep dive into Asana AI Studio and Amazon Q index collaboration
Today, we’re excited to announce the integration of Asana AI Studio with Amazon Q index, bringing generative AI directly into your daily workflows. In this post, we explore how Asana AI Studio and Amazon Q index transform enterprise efficiency through intelligent workflow automation and enhanced data accessibility.
Insulin resistance prediction from wearables and routine blood biomarkers
この記事では、ウェアラブルデータと日常的な血液検査を活用して、インスリン抵抗性（IR）を効果的に予測する新しい手法を提案しています。この手法は、2型糖尿病のリスクスクリーニングを早期に行うためのスケーラブルでアクセス可能なアプローチを提供します。2型糖尿病は世界中で数億人に影響を及ぼしており、その前兆としてインスリン抵抗性が重要です。従来のIR測定方法は侵襲的で高価なため、早期発見が困難でした。そこで、ウェアラブルデバイスからのデータ（安静時心拍数、歩数、睡眠パターン）と日常的な血液検査（空腹時血糖、脂質パネル）を用いてIRリスクを推定する機械学習モデルを開発しました。このアプローチは、特に肥満や運動不足の高リスク個人において強いパフォーマンスを示しました。また、インスリン抵抗性を理解するためのエージェントも紹介されており、個別の推奨を安全に行う手助けをします。 • インスリン抵抗性（IR）の早期発見が重要であること • 従来のIR測定方法は侵襲的で高価であるため、早期発見が困難 • ウェアラブルデータと日常的な血液検査を用いた新しい予測手法を提案 • 機械学習モデルがIRリスクを推定する能力を持つ • 特に肥満や運動不足の高リスク個人において強いパフォーマンスを示す • インスリン抵抗性を理解するためのエージェントを導入 • この研究は情報提供と研究目的のために設計されている
Process multi-page documents with human review using Amazon Bedrock Data Automation and Amazon SageMaker AI
In this post, we show how to process multi-page documents with a human review loop using Amazon Bedrock Data Automation and Amazon SageMaker AI.
How Fulton County Schools use Copilot Chat to empower student innovation
Fulton County Schools uses Copilot Chat in the classroom to support learning, reduce tasks, and build career readiness. Learn more.
Get the number of auto-fit/auto-fill columns in CSS
この記事では、CSSのグリッドレイアウトにおけるauto-fitおよびauto-fillカラムの数を取得する方法について説明しています。特に、特定の行や列を強調表示したり、レスポンシブな非矩形グリッドを作成したりする際に役立ちます。CSSのコンテナクエリユニット、CSS変数、数学関数を使用して、ブレークポイントなしでこれを実現する方法が紹介されています。具体的には、.gridクラスを持つ要素に対して、auto-fitまたはauto-fillを使用してカラムを設定し、Safariのバグに対処するための代替手段も提案されています。最終的に、カラム数を計算するためのCSSコードも示されており、サポートされていないブラウザでの動作についても言及されています。 • CSSのグリッドレイアウトでauto-fit/auto-fillカラムの数を取得する方法を解説 • 特定の行や列を強調表示するための技術 • レスポンシブな非矩形グリッドの作成が可能 • CSSのコンテナクエリユニットと数学関数を使用 • Safariのバグに対処するための代替手段を提案 • カラム数を計算するための具体的なCSSコードを示す • サポートされていないブラウザでの動作についての注意点
Introducing Vercel MCP: Connect Vercel to your AI tools
Vercel now has an official hosted MCP server (aka Vercel MCP), which you can use to connect your favorite AI tools, such as Claude or VS Code, directly to Vercel.
Introducing AI Elements: build AI interfaces faster
Focus on your AI’s intelligence, not the UI scaffolding. AI Elements is now available as a new Vercel product to help frontend engineers build AI-driven interfaces in a fraction of the time.