AI Pulse

<h3>Misc Changes</h3> <ul> <li>Turbopack analyzer: analyze ternaries + promises in string &amp; nullish methods: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4538796111" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/94187" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/94187/hovercard" href="https://github.com/vercel/next.js/pull/94187">#94187</a></li> <li>[turbopack] Break up <code>g

<h2>What's changed</h2> <ul> <li>Auto mode is now available on Bedrock, Vertex, and Foundry for Opus 4.7 and Opus 4.8. Opt in by setting <code>CLAUDE_CODE_ENABLE_AUTO_MODE=1</code></li> </ul>

<p>v16.3.0-canary.35</p>

<h3>Misc Changes</h3> <ul> <li>Update <code>shortcircuit_if_known</code> to <code>eval_shortcircuit()</code>: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4534827873" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/94163" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/94163/hovercard" href="https://github.com/vercel/next.js/pull/94163">#94163</a></li> <li>Remove dead code from <code>JsV

<h2>What's changed</h2> <ul> <li>Plugins in <code>.claude/skills</code> directories are now automatically loaded, no marketplace required</li> <li>Added <code>claude plugin init <name></code> to scaffold a new plugin in <code>.claude/skills</code></li> <li>Added autocomplete for <code>/plugin</code> arguments: subcommands, installed plugin names, and plugins from known marketplaces</li> <li><code>claude agents</code>: the <code>agent</code> field in <code>settings.json</code> is now honored for

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/IOQuiz2026_social.max-600x600.format-webp.webp">We used Google AI Studio to vibe code a quiz about our top I/O 2026 announcements.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Gemini_Omni_and_Gemini_3.5_hero.max-600x600.format-webp.webp">Watch 11 videos showing the capabilities of Gemini Omni and Gemini 3.5, announced at Google I/O 2026.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/FutureLabs_social.max-600x600.format-webp.webp">University of Waterloo students develop AI prototypes like sign language tutors to reshape the future of education and work.

Function invocations are moving from package-based to per-unit pricing for Pro and new Enterprise customers. You’ll continue paying the same effective rate until the end of your current billing cycle. Starting with your next billing cycle you’ll be billed per unit to align costs directly with your usage. The new rate is $0.0000006 per invocation (previously $0.60 per 1M invocations) for Pro customers. Per‑unit billing scales more smoothly across team sizes and usage patterns. It also helps teams

HTTP requests are inexpensive. Vercel charges ~$2/million, a fraction of a cent per call. But a single prompt to an agent on a frontier model can cost $2, making AI a million times more expensive, and inference theft one of the highest-margin businesses an attacker can run. We have seen this type of attack on our own APIs. If you have AI endpoints exposed to the internet, the risk of abuse is high and can easily run up bills in the tens of thousands of dollars or more. Protecting those endpoints

<h2>What's changed</h2> <ul> <li>Fixed an issue when using Opus 4.8 where thinking blocks were modified, leading to API errors.</li> </ul>

<h2>0.3.1 (2026-05-29)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/aws-sdk-v0.3.0...aws-sdk-v0.3.1">aws-sdk-v0.3.0...aws-sdk-v0.3.1</a></p> <h3>Chores</h3> <ul> <li><strong>client:</strong> update lockfiles to have proper dependencies on standardwebhooks (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/5e9b5237e53aefda65637bd11eb0d07f50bbc952">5e9b523</a>)</li> </ul>

<h2>0.2.4 (2026-05-29)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/foundry-sdk-v0.2.3...foundry-sdk-v0.2.4">foundry-sdk-v0.2.3...foundry-sdk-v0.2.4</a></p> <h3>Chores</h3> <ul> <li><strong>client:</strong> update lockfiles to have proper dependencies on standardwebhooks (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/5e9b5237e53aefda65637bd11eb0d07f50bbc952">5e9b523</a>)</li> </ul>

<h2>0.100.1 (2026-05-29)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.100.0...sdk-v0.100.1">sdk-v0.100.0...sdk-v0.100.1</a></p> <h3>Bug Fixes</h3> <ul> <li><strong>streaming:</strong> carry encrypted_content on beta compaction blocks (<a href="https://github.com/anthropics/anthropic-sdk-typescript/issues/1025" data-hovercard-type="pull_request" data-hovercard-url="/anthropics/anthropic-sdk-typescript/pull/1025/hovercard">#1025</a>) (<a hr

Vercel Sandbox now supports installing and running Docker inside a sandbox. An agent can build containers, install system packages, and modify files without touching your host system. Install Docker, start the daemon, and serve a containerized application: Docker in a Sandbox is useful for running containerized services like Redis or Postgres as test dependencies, validating container images before deploying, or previewing applications served from a container. Combined with persistent sandboxes

Vercel Sandboxes now allow opening and binding to port 8080 as an ingress domain . This port was used as a controller port, and that has now moved to port 23456. Learn more about Sandbox in the documentation . Read more

<p>v16.3.0-canary.33</p>

<h2>What's changed</h2> <ul> <li>Opus 4.8 is here! Now defaults to high effort · /effort xhigh for your hardest tasks</li> <li>Introducing dynamic workflows: ask Claude to create a workflow and it orchestrates work across tens to hundreds of agents in the background, so you can take on larger, more complex tasks. Run <code>/workflows</code> to view your runs</li> <li>Fast mode on Opus 4.8 is now available at a fraction of its previous cost: 2x the standard rate for 2.5x the speed</li> <li>The le

<h2>0.100.0 (2026-05-28)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.99.0...sdk-v0.100.0">sdk-v0.99.0...sdk-v0.100.0</a></p> <h3>Features</h3> <ul> <li><strong>api:</strong> Add support for claude-opus-4-8, mid-conversation system blocks, and usage.output_tokens_details (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/bb0bf278ea0afdf383b6ea711cadf3ab788ea84a">bb0bf27</a>)</li> </ul> <h3>Documentation</h3> <ul> <li>

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.6.0...v2.7.0">2.7.0</a> (2026-05-27)</h2> <h3>Features</h3> <ul> <li>Additional computer_use field support for vertex. (<a href="https://github.com/googleapis/python-genai/commit/b4828fa5b085d2dbb47a0adc9c10e4b35d60ad64">b4828fa</a>)</li> <li><strong>interaction-api:</strong> Allow "text/csv" as a supported document mime type for Interaction API. (<a href="https://github.com/googleapis/python-genai/commit/543137b78d392c7d81a287f3

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/KW_KNH.max-600x600.format-webp.webp">Here are 12 of the biggest Google I/O 2026 keynote moments, including news about Gemini Omni, Gemini 3.5 Flash and more.

<h3>Misc Changes</h3> <ul> <li>Fix short circuit evaluation of AND / OR in the Turbopack analyzer for string &amp; nullish-related methods: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4534168282" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/94159" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/94159/hovercard" href="https://github.com/vercel/next.js/pull/94159">#94159</a></li> <li>T

Claude Opus 4.8 is now available on Vercel AI Gateway . Claude Opus 4.8 is built for long-horizon agentic execution and handles complex, multi-step coding tasks like refactors that previously required human correction mid-task. The model also produces clearer, less hedgy prose for knowledge work like drafting documents, analyzing data, and building presentations. To use Opus 4.8, set model to anthropic/claude-opus-4.8 in the AI SDK . AI Gateway provides a unified API for calling models, tracking

<h2>What's changed</h2> <ul> <li>Added <code>skipLfs</code> option to <code>github</code>/<code>git</code> plugin marketplace sources to skip Git LFS downloads during clone and update</li> <li>Claude Code now shows a one-time notice when your npm global install can't auto-update; <code>/doctor</code> lists the fixes</li> <li>Status line commands now receive <code>COLUMNS</code> and <code>LINES</code> environment variables so scripts can size output to the terminal width</li> <li><code>claude age

AI Gateway now supports a team-wide provider allowlist. Teams can restrict which providers can serve requests, so traffic only routes to approved providers. The allowlist applies to every request through AI Gateway, including Bring Your Own Key (BYOK) traffic. Regulated teams typically vet AI providers across multiple dimensions with security and legal sign-off, ending up with a vendor set that reflects the specific requirements of their org. The allowlist turns that approved-vendor list into a

Amazon OpenSearch Serverless is now available in the Vercel Marketplace , with guided setup and automatic project configuration built into the Vercel dashboard. Built for agentic infrastructure By bringing Amazon OpenSearch Serverless into the Vercel Marketplace, teams benefit from a unified workflow: In-dashboard provisioning: Spin up OpenSearch collections directly from the Vercel dashboard Automatic configuration: Environment variables are automatically injected into your projects on provisio

<p>v16.3.0-canary.31</p>

Conductor on Vercel Multiple parallel coding agents running at scale in the cloud Model-agnostic, works with Claude Code, Codex, and other coding agents Used by engineering teams at Notion, Linear, Ramp, and Life360 Interacting with a single coding agent feels natural, and Conductor makes directing a fleet of agents just as intuitive. It's a GUI for spinning up multiple coding agents in parallel, each working on an isolated branch of your codebase at once. You review their work, merge what works

<h2>What's changed</h2> <ul> <li><code>/code-review --fix</code> now applies review findings to your working tree after the review, surfacing reuse, simplification, and efficiency suggestions; <code>/simplify</code> now invokes <code>/code-review --fix</code></li> <li>Skills and slash commands can now set <code>disallowed-tools</code> in frontmatter to remove tools from the model while the skill is active</li> <li>Added <code>/reload-skills</code> command to re-scan skill directories without res

<h2>0.99.0 (2026-05-27)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.98.1...sdk-v0.99.0">sdk-v0.98.1...sdk-v0.99.0</a></p> <h3>Features</h3> <ul> <li>support custom file size caps (<a href="https://github.com/anthropics/anthropic-sdk-typescript/issues/1029" data-hovercard-type="pull_request" data-hovercard-url="/anthropics/anthropic-sdk-typescript/pull/1029/hovercard">#1029</a>) (<a href="https://github.com/anthropics/anthropic-sdk-typesc

The Vercel CLI now ships an optional experimental native binary that starts faster, is even more secure, and requires no Node.js runtime dependency. Binaries are code-signed, allowing your OS to verify that they came from Vercel and haven't been modified. Additionally, on macOS, credentials are stored in the system Keychain scoped to the binary, so other processes cannot access them without explicit permission. You can opt in by installing the experimental package: The -f flag is required becaus

The deployments list has been redesigned with a denser layout, so you can see more deployments at once. Environments are now grouped with statuses, and the updated layout makes branches and commits easier to scan. The mobile experience has also been improved, making it easier to scan deployment activity on busy projects. Open your dashboard to see the updated design. Read more

<p>v16.3.0-canary.30</p>

<h2>0.98.1 (2026-05-26)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.98.0...sdk-v0.98.1">sdk-v0.98.0...sdk-v0.98.1</a></p> <h3>Bug Fixes</h3> <ul> <li><strong>client:</strong> preserve directory prefix in skills.versions.create uploads (<a href="https://github.com/anthropics/anthropic-sdk-typescript/issues/1024" data-hovercard-type="pull_request" data-hovercard-url="/anthropics/anthropic-sdk-typescript/pull/1024/hovercard">#1024</a>) (<a

Vercel Sandboxes now automatically save and restore filesystem state between sessions. Persistence is on by default, meaning no snapshots to manage or state to track manually. Each sandbox has a durable, customizable name that acts as a unique reference in your project. You can create, retrieve, or resume a sandbox by name. Vercel spins sessions up and down automatically, without interrupting your workflow. Create a persistent sandbox When you call Sandbox.create() , persistence is enabled by de

Firecrawl is now available on the Vercel Marketplace , allowing Vercel teams to power AI agents and applications with structured web data without managing crawling infrastructure. This integration helps developers scrape websites into LLM-ready formats, search and retrieve full page content, and interact with dynamic pages for retrieval and agent workflows. Key capabilities include: Scrape pages into markdown, HTML, structured data, or screenshots Search the web and retrieve full page content in

Vercel Domains now supports price sorting and availability filtering. Price sorting shows lower cost domains first, so you can quickly find a domain that fits your budget. Availability filtering moves unavailable domains to the bottom of the search results, so you can focus on domains that you can actually purchase. Try it at vercel.com/domains . Read more

This week we are gradually rolling out an update to Vercel Microfrontends routing for aliases and branch-assigned domains. Aliases inherit Microfrontends routing Aliasing a Microfrontends URL with vc alias now preserves the full microfrontends routing config from the source deployment. Previously, the new alias only inherited the deploymentId . Update to the latest Vercel CLI to pick up the change. Branch domains route across all projects A project domain assigned to a git branch now routes to t

<p>v16.3.0-canary.29</p>

Endeavour Energy on Vercel Sub-1s page loads during peak storm traffic Five-minute data sync cycle across all upstream systems 38% faster deployments compared to their previous platform When a summer storm rolls through New South Wales in Australia, hundreds of thousands of people pull up the same page: the Endeavour Energy outage map . During the worst storms, people check it from the dark, on a phone with patchy reception, while a tree is still down on the line outside. They want one thing: ac

<h3>Misc Changes</h3> <ul> <li>[Skills] Mark backport-pr and create-pr as internal: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4509476516" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/94076" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/94076/hovercard" href="https://github.com/vercel/next.js/pull/94076">#94076</a></li> </ul> <h3>Credits</h3> <p>Huge thanks to <a class="user-menti

<h2>What's changed</h2> <ul> <li>Internal infrastructure improvements (no user-facing changes)</li> </ul>

<p>v16.3.0-canary.27</p>

<h2>What's changed</h2> <ul> <li><code>/usage</code> now shows a per-category breakdown of what's driving your limits usage — skills, subagents, plugins, and per-MCP-server cost</li> <li><code>/diff</code> detail view can now be scrolled with the keyboard (arrows, <code>j</code>/<code>k</code>, <code>PgUp</code>/<code>PgDn</code>, <code>Space</code>, <code>Home</code>/<code>End</code>)</li> <li>Markdown output now renders GFM task list checkboxes (<code>- [ ] todo</code> / <code>- [x] done</code

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/IO26_Dialogues_3z680sK.max-600x600.format-webp.webp">A recap of the 2026 I/O Dialogues, where leaders discuss the future of AI, quantum computing, robotics and creativity.

<h2>What's changed</h2> <ul> <li>Fixed the Bash tool returning exit code 127 on every command for some users (a regression introduced in 2.1.147)</li> </ul>

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.5.0...v2.6.0">2.6.0</a> (2026-05-21)</h2> <h3>Features</h3> <ul> <li>Add <code>enable_prompt_injection_detection</code> for Computer Use feature for the Gemini API. (<a href="https://github.com/googleapis/python-genai/commit/b1f632d5d617a8e24fd4cb87556a4eb180b6d29b">b1f632d</a>)</li> <li>Add budget_exceeded status (<a href="https://github.com/googleapis/python-genai/commit/15443c0211446c41a9ab2705dd2282d324581eb7">15443c0</a>)</l

<h3>Misc Changes</h3> <ul> <li>[ci] Add an <code>TURBO_TASKS_AVAILABLE_PARALLELISM</code> environment variable that overrides <code>std::thread::available_parallelism</code> callsites: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4490118742" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/93995" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/93995/hovercard" href="https://github.com/ver

<h2>What's changed</h2> <ul> <li>Pinned background sessions (<code>Ctrl+T</code> in <code>claude agents</code>) now stay alive when idle, are restarted in place to apply Claude Code updates, and are shed under memory pressure only after non-pinned sessions</li> <li>Renamed <code>/simplify</code> to <code>/code-review</code>. It now reports correctness bugs at a chosen effort level (e.g., <code>/code-review high</code>); pass <code>--comment</code> to post findings as inline GitHub PR comments. T

<h2>0.98.0 (2026-05-21)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.97.1...sdk-v0.98.0">sdk-v0.97.1...sdk-v0.98.0</a></p> <h3>Features</h3> <ul> <li><strong>api:</strong> Add support for thinking-token-count beta for estimated tokens in thinking block deltas when streaming (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/0528d47b0a390ee7f88d72f85ffe6079ae1c6c00">0528d47</a>)</li> </ul>

You can now access anomaly alerts and their details directly through the Vercel CLI . With the vercel alerts command, you can list all alerts for a team or given project. For each alert, you can view the start time, the type of alert, and whether or not the alert is still active. With the --ai option, the AI investigation results appear alongside each alert. You and your agent can act on alerts without leaving the terminal. Available on Observability Plus . Learn more about vercel alerts in the

Qwen 3.7 Max from Alibaba is now available on Vercel AI Gateway . The model is designed as an agent foundation, with capabilities spanning coding, office workflow automation, and long-horizon autonomous execution. Qwen 3.7 Max shows improvements in frontend prototyping and complex multi-file engineering. The model supports office and productivity tasks through multi-agent orchestration and sustains coherent reasoning across long-horizon tool-calling sessions. To use Qwen 3.7 Max, set model to al

<h2>What's changed</h2> <ul> <li>Renamed <code>/simplify</code> to <code>/code-review</code> with an optional effort level (e.g. <code>/code-review high</code>)</li> <li>Auto mode no longer suppresses <code>AskUserQuestion</code> when the user or a skill explicitly relies on it</li> <li>Fixed Windows PowerShell tool failing with "command line is invalid" when <code>pwsh</code> is installed via winget or the Microsoft Store (regression in v2.1.124)</li> <li>Fixed MCP <code>resources/list</code>,

You can now configure weighted traffic splits for Vercel Flags with the new vercel flags split command in the Vercel CLI. This allows you to send a percentage of traffic to one variant and the rest to another. Run the command interactively, or pass the environment, bucketing attribute, and variant weights as flags: Update to the latest version of the Vercel CLI and read the documentation to get started. Read more

<p>v16.3.0-canary.25</p>

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/MissouriSocial.max-600x600.format-webp.webp">We’re helping build the state’s next-generation workforce and investing in energy programs.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/100_things_Social.max-600x600.format-webp.webp">This year at Google I/O 2026, we announced Gemini Omni, Google Antigravity, Universal Cart and so much more. Here are the highlights.

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.4.0...v2.5.0">2.5.0</a> (2026-05-20)</h2> <h3>Features</h3> <ul> <li>Add Gemini 3.5 Flash model to options (<a href="https://github.com/googleapis/python-genai/commit/706fd021fb0e0119e0867be2267592e7709cd274">706fd02</a>)</li> </ul>

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Screenshot_2026-05-15_at_4.21.2.max-600x600.format-webp.webp">See and hear your colleagues in true-to-life size and sound, making hybrid meetings feel more inclusive and connected.

Grok Build 0.1 is now available on Vercel AI Gateway . This is a beta coding model trained for agentic coding, currently in early access, and powers the Grok Build CLI app. Reasoning effort is not configurable, and there is no non-reasoning mode. To use Grok Build 0.1, set model to xai/grok-build-0.1 in the AI SDK . AI Gateway provides a unified API for calling models, tracking usage and cost, and configuring retries, failover, and performance optimizations for higher-than-provider uptime. It in

The Vercel AI Gateway plugin gives any WordPress site access to hundreds of models from 40+ providers through a single API key. Providers include Anthropic, Google, OpenAI, xAI, DeepSeek, MiniMax, Moonshot AI, and more. The plugin is implemented as a connector for the new WordPress AI Client , which requires WordPress 7.0, released today. What this enables Any AI-powered plugin works automatically. Plugins built on the WordPress AI Client pick up the connector without their own provider integrat

You can now pause a Workflow run on a Chat SDK card and resume it when someone clicks a button. The same flow works for form submissions. Buttons and modals accept a new callbackUrl prop, and the event payload is sent to that endpoint. To build a card like this, create a workflow webhook and pass its URL to each button's callbackUrl prop inside your <Card> component: For the <Modal> component, the form data is in the payload. callbackUrl works for buttons on most platforms with an official adapt

You can now read the parent issue or pull request when your bot is mentioned in a Linear or GitHub comment. message.subject resolves to that parent with title, status, URL, and the full typed payload. message.subject is cached per message, so repeated access only hits the API once. It resolves to null on Slack and other chat platforms, where there's no parent resource. Direct access to platform SDKs The GitHub, Linear, and Slack adapters now expose their underlying platform SDKs. Use them to ext

Chat SDK now ships a built-in AI SDK toolset through the new chat/ai subpath. One createChatTools(chat) call wires Chat SDK's read and write actions into your agent. Approval by default: write tools are gated by a requireApproval option. Presets: reader , messenger , and moderator scope the toolset. Lazy loading: only the tools your preset allows are constructed. toAiMessages and its supporting types have moved to chat/ai . The previous chat re-exports are flagged @deprecated . Read the document

<p>v16.3.0-canary.24</p>

<h2>What's changed</h2> <ul> <li>Added <code>claude agents --json</code> to list live Claude sessions as JSON for scripting (tmux-resurrect, status bars, session pickers)</li> <li>Added <code>agent_id</code> and <code>parent_agent_id</code> attributes to <code>claude_code.tool</code> OTEL spans, and fixed trace parenting so background subagent spans nest under the dispatching Agent tool span</li> <li>Status line JSON input now includes GitHub repo and PR information when detected</li> <li><code>

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/I_O_26_Header_1_AYohnR4.max-600x600.format-webp.webp">Introducing a $100 AI Ultra plan — plus, new features and benefits for Google AI Plus, Pro and Ultra subscribers.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/Search_AI_and_search_engine_v46.max-600x600.format-webp.webp">We shared the next step in our journey to bring together the best of a search engine with the best of AI.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/gemini-3-5__keyword__blog-heade.max-600x600.format-webp.webp">At Google I/O we released Gemini 3.5, our latest series of models combining frontier intelligence with action.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/SundarKeynote-hero.max-600x600.format-webp.webp">The latest from Google I/O: See how we’re helping you get more done with Gemini.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/GoogleWorkspace-IO.max-600x600.format-webp.webp">Announcing new voice capabilities in Gmail, Docs and Keep, a new design tool called Google Pics and updates to AI Inbox.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/images/AI_Mode_US.max-600x600.format-webp.webp">One year after launch, see how AI Mode’s users are shifting from keywords to natural language queries.

<img src="https://storage.googleapis.com/gweb-uniblog-publish-prod/original_images/Collection-Hero.gif">At Google I/O 2026, we shared how we’re making AI more helpful for everyone. See everything we announced.

<h2>0.97.1 (2026-05-19)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.97.0...sdk-v0.97.1">sdk-v0.97.0...sdk-v0.97.1</a></p> <h3>Bug Fixes</h3> <ul> <li><strong>runner:</strong> skip tool calls SessionToolRunner does not own (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/9987379abeaf28b17edb5518ac229d2a6caa4bf6">9987379</a>)</li> </ul>

The Nuxt MCP Toolkit now supports MCP apps . Your agent tools can return interactive HTML responses that MCP clients like Claude and ChatGPT render inline, rather than plain-text responses. Declare a tool with the defineMcpApp macro, then read pre-hydrated data, trigger follow-up prompts, or call other tools from inside the UI with the useMcpApp composable. The toolkit bundles each Vue SFC into a self-contained HTML file at build time and serves it from your MCP endpoint. Read the Nuxt MCP Toolk

<h2>0.29.2 (2026-05-19)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/bedrock-sdk-v0.29.1...bedrock-sdk-v0.29.2">bedrock-sdk-v0.29.1...bedrock-sdk-v0.29.2</a></p> <h3>Bug Fixes</h3> <ul> <li>align @types/node in sub-packages to fix CI build (<a href="https://github.com/anthropics/anthropic-sdk-typescript/issues/1017" data-hovercard-type="pull_request" data-hovercard-url="/anthropics/anthropic-sdk-typescript/pull/1017/hovercard">#1017</a>) (<a hre

<h2>0.16.1 (2026-05-19)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/vertex-sdk-v0.16.0...vertex-sdk-v0.16.1">vertex-sdk-v0.16.0...vertex-sdk-v0.16.1</a></p> <h3>Bug Fixes</h3> <ul> <li>align @types/node in sub-packages to fix CI build (<a href="https://github.com/anthropics/anthropic-sdk-typescript/issues/1017" data-hovercard-type="pull_request" data-hovercard-url="/anthropics/anthropic-sdk-typescript/pull/1017/hovercard">#1017</a>) (<a href="h

<h2>0.97.0 (2026-05-19)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.96.0...sdk-v0.97.0">sdk-v0.96.0...sdk-v0.97.0</a></p> <h3>Features</h3> <ul> <li><strong>client:</strong> Add support for self-hosted sandboxes in CMA with sandbox helpers (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/659a343c820e316229715466b64e420428ee762b">659a343</a>)</li> </ul> <h3>Bug Fixes</h3> <ul> <li><strong>typescript:</strong> upgrad

Gemini 3.5 Flash is now available on Vercel AI Gateway . This model has improved coding proficiency and parallel agentic execution loops versus previous Flash versions. It also brings improvements to core reasoning, instruction following, and multi-turn coherence, with stronger performance on complex tasks and higher-quality reasoning traces in thinking mode. 3.5 Flash defaults to the medium thinking level, balancing response quality with faster, more cost-efficient generation. To use Gemini 3.5

<h2>What's changed</h2> <ul> <li>Added <code>/resume</code> support for background sessions — sessions started via <code>claude --bg</code> or agent view now appear alongside interactive ones, marked with <code>bg</code></li> <li>Added elapsed duration to background subagent completion notifications (e.g. "Agent completed · 3h 2m 5s")</li> <li>The <code>/plugin</code> browse and discover panes now show when a plugin was last updated</li> <li><code>/model</code> now changes the model for the curr

Flat Rate CDN is now available in Limited Beta for Pro teams. It replaces usage-based CDN pricing with a fixed monthly fee. Viral posts, unfiltered bots, or misconfigured routes can turn a normal month into a surprise bill. Flat Rate CDN makes your cost predictable. Fixed monthly price : Pay a fixed amount for Vercel CDN each month, regardless of traffic and usage. Covers spikes across Vercel CDN : Edge Requests, Fast Data Transfer, and other products are no longer metered. No overages : Your bi

<p>v16.3.0-canary.23</p>

You can now run Claude Managed Agents with Vercel Sandbox . Claude Managed Agents handles the model, harness, tools, and session state. Self-hosting lets you bring the execution environment, so an agent's tool calls run on your existing Vercel infrastructure with your private APIs, internal services, and customer data. Each agent session runs in its own isolated Firecracker microVM, using the same infrastructure that powers 1B+ Vercel deployments with enterprise-grade security, availability, and

Monorepos can now opt in to a single, consolidated commit status on pull requests instead of one commit status per project. For repos with many projects, teams can configure GitHub branch protection once, then manage which Vercel projects are required for merge in each project’s settings. Read more about Github Commit statuses and and enable Consolidated Commit Status from your project settings . Read more

Vercel Firewall now waives CDN Requests and Fast Data Transfer for any traffic denied, challenged, or rate‑limited by Web Application Firewall (WAF). Vercel has always provided unlimited DDoS mitigation at no cost. Vercel WAF, included in CDN cost, gives you custom rules, managed rules, and rate limiting for bad traffic that isn't DDoS. With this change, you don't pay for requests or bandwidth that WAF denies, challenges, or rate‑limits. That means no surprise bill when a scraper hammers your pr

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.3.0...v2.4.0">2.4.0</a> (2026-05-17)</h2> <h3>Features</h3> <ul> <li>Support Agent and Environment APIs. (<a href="https://github.com/googleapis/python-genai/commit/ef20d6ba54a183280c046a4346c125168d5f8a95">ef20d6b</a>)</li> </ul> <h3>Bug Fixes</h3> <ul> <li>Output_text for turns that don't end with text. (<a href="https://github.com/googleapis/python-genai/commit/2afdefff90f3fb4f98075a18c99d44b9e722dde6">2afdeff</a>)</li> <li>Pa

<h3>Misc Changes</h3> <ul> <li>turbo-tasks: prioritize recomputed tasks with Recomputation priority: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4455190483" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/93876" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/93876/hovercard" href="https://github.com/vercel/next.js/pull/93876">#93876</a></li> </ul> <h3>Credits</h3> <p>Huge thanks to <a

<h3>Core Changes</h3> <ul> <li>Redesign dev overlay: cleaner shell + instant fix-card guidance: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4419730771" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/93755" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/93755/hovercard" href="https://github.com/vercel/next.js/pull/93755">#93755</a></li> </ul> <h3>Misc Changes</h3> <ul> <li>Trace middle

<h2>What's changed</h2> <ul> <li>Added plugin dependency enforcement: <code>claude plugin disable</code> now refuses when another enabled plugin depends on the target (with a copy-pasteable disable-chain hint), and <code>claude plugin enable</code> force-enables transitive dependencies</li> <li>Added projected context cost (per-turn and per-invocation token estimates) to the <code>/plugin</code> marketplace browse pane</li> <li>Added <code>worktree.bgIsolation: "none"</code> setting to let backg

<h3>Core Changes</h3> <ul> <li>[turbopack] fix feature usage telemetry: <a class="issue-link js-issue-link" data-error-text="Failed to load title" data-id="4304590681" data-permission-text="Title is private" data-url="https://github.com/vercel/next.js/issues/93100" data-hovercard-type="pull_request" data-hovercard-url="/vercel/next.js/pull/93100/hovercard" href="https://github.com/vercel/next.js/pull/93100">#93100</a></li> <li>Add an info panel for the existing "Cache disabled" indicator: <a cla

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.2.0...v2.3.0">2.3.0</a> (2026-05-15)</h2> <h3>Features</h3> <ul> <li>Add content union to UserInputStep (<a href="https://github.com/googleapis/python-genai/commit/a5059a82dc596f9555dd3221aa6e7414d50df24a">a5059a8</a>)</li> <li>Interaction.{output_text,output_image,output_audio,output_video} (<a href="https://github.com/googleapis/python-genai/commit/975d16a3cea49282137dd2a901b219820a641b64">975d16a</a>)</li> <li>Migrate Agent En

You can now sort the providers behind a model by cost, time to first token (TTFT), or throughput (TPS) in AI Gateway . The default provider order blends provider reliability, quality of model output, cost, and speed of response. You can now use sort for explicit control over ranking criteria. For models with many providers and noticeable cost or speed variation, you can use sort to optimize on your dimension of choice. Ranking is computed at request time, so newly added providers, price changes,

You can now use native curl syntax with the Vercel CLI. The vercel curl command accepts full URLs, bare hostnames, and the --url flag, and uses your Vercel auth to bypass Deployment Protection . If you've linked a project, you can also pass just a path: Update to the latest Vercel CLI version and run vercel curl to get started. Learn more in the Vercel CLI documentation . Read more

<h2>What's changed</h2> <ul> <li>Added new <code>claude agents</code> flags: <code>--add-dir</code>, <code>--settings</code>, <code>--mcp-config</code>, <code>--plugin-dir</code>, <code>--permission-mode</code>, <code>--model</code>, <code>--effort</code>, and <code>--dangerously-skip-permissions</code> to configure dispatched background sessions</li> <li>Fast mode now uses Opus 4.7 by default (previously Opus 4.6). Set <code>CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE=1</code> to pin fast mode to O

You can now restrict access to production source maps with Protected Source Maps , which puts browser .map files behind Vercel Authentication . Your team can fetch them; everyone else gets a 404. Source maps are how you debug minified production code. They give you readable stack traces and your original source code, with the real filenames and line numbers intact. New projects have Protected Source Maps enabled by default. Existing projects can opt in from Settings → Deployment Protection, with

<h2>What's changed</h2> <ul> <li>Added <code>terminalSequence</code> field to hook JSON output so hooks can emit desktop notifications, window titles, and bells without a controlling terminal</li> <li>Added <code>CLAUDE_CODE_PLUGIN_PREFER_HTTPS</code> to clone GitHub plugin sources over HTTPS instead of SSH, for environments without a GitHub SSH key</li> <li>Added <code>ANTHROPIC_WORKSPACE_ID</code> environment variable for workload identity federation — scopes the minted token to a specific wor

<h2>0.96.0 (2026-05-13)</h2> <p>Full Changelog: <a href="https://github.com/anthropics/anthropic-sdk-typescript/compare/sdk-v0.95.2...sdk-v0.96.0">sdk-v0.95.2...sdk-v0.96.0</a></p> <h3>Features</h3> <ul> <li><strong>api:</strong> Add BetaManagedAgentsSearchResultBlock types (<a href="https://github.com/anthropics/anthropic-sdk-typescript/commit/08f02f3d0c34a9563b17af40b227acf3b105f8a3">08f02f3</a>)</li> <li><strong>api:</strong> Add support for cache diagnostics beta (<a href="https://github.com

Trusted Sources lets protected deployments accept short-lived identity tokens (OIDC) from Vercel projects and external services you authorize, so you no longer have to share a long-lived Protection Bypass for Automation secret. Trusted Sources is the recommended approach, but Protection Bypass for Automation continues to work Callers attach an OIDC token in the x-vercel-trusted-oidc-idp-token header. Vercel then verifies the signature, checks the claims you configured, and confirms the environme

<h2><a href="https://github.com/googleapis/python-genai/compare/v2.1.0...v2.2.0">2.2.0</a> (2026-05-12)</h2> <h3>Features</h3> <ul> <li>Added missing FunctionCallResultDelta type and <code>arguments</code> field to the ArgumentDelta type (<a href="https://github.com/googleapis/python-genai/commit/b0131f80c2da97926bcef00b4abef9334272ee7a">b0131f8</a>)</li> </ul>

<h2>What's changed</h2> <ul> <li>Improved Agent tool <code>subagent_type</code> matching to accept case- and separator-insensitive values (e.g. <code>"Code Reviewer"</code> resolves to <code>code-reviewer</code>)</li> <li>Updated agent color palette</li> <li>Fixed <code>/goal</code> silently hanging when <code>disableAllHooks</code> or <code>allowManagedHooksOnly</code> is set — now shows a clear message instead of an indicator that never resolves</li> <li>Fixed a regression in settings hot-relo