Firehose

Here are Google’s latest AI updates from October 2026

Vercel MCP can now show your deployments as an interactive card inside ChatGPT and other chat clients that support MCP Apps . The card shows live deployment status, build logs, and source and commit details, with a Visit button once the dep…

Flags Explorer is a tool in the Vercel Toolbar where you can see which feature flags are active on your end, and share them with your team, either by setting them for the branch or through a URL. The Explorer surface has been redesigned to …

Vercel Sandboxes that mount Drives now start faster. Median creation time dropped from 1.7s to 0.7s, and p95 dropped from 2.2s to 1.6s. The speedup applies automatically to every sandbox that uses Drives, with no SDK upgrade or configuratio…

Project Connections let applications evaluate flags defined in another Vercel project within the same team, without managing SDK keys. Previously, sharing flags between projects required copying a server SDK key into an environment variable…

OpenAI’s Grid Protectors Program will work with utilities to help future compute campuses support grid reliability and protect consumer electricity costs.

Zone setting changes used to rebuild entire DNS zones, even ones with tens of millions of records. Here is how we built a planner that picks the smallest safe build, cutting full builds by 40%, peak memory by 85%, and halving build delay.

Introducing LifeVal, a benchmark measuring how well AI helps people complete everyday tasks, from finding childcare to managing finances and government paperwork.

Vercel now checks for blocked npm packages before and during dependency installation in builds , helping prevent malicious install scripts from running. The checks use a blocklist of specific package versions selected for blocking because o…

GitHub’s annual month-long game jam is back for its 14th year! This November, make a game inspired by the theme SHIFT. Change gears, swap perspectives, bend reality, or just press a certain key. The post Game Off 2026: The theme is SHIFT ap…

Lakebase's copy-on-write branching and branch expiration (TTL) features enable fast and cheap CI/CD testing by creating isolated, writable copies of production data in seconds, scaling to zero when idle, and automatically cleaning up abandoned branches. This approach allows for 100 concurrent PR branches on a ~2.5 TB production dataset, with provisioning times under 60 seconds and storage costs comparable to traditional CI/CD methods. By leveraging Lakebase's features, teams can achieve fast and secure CI/CD testing, with security and compliance requirements met through private networking and identity management. AI summary

Databricks and SAP are expanding their partnership to provide end-to-end data and AI capabilities, allowing customers to accelerate time to value for advanced analytics and AI by leveraging SAP's deep understanding of business processes and Databricks' industry-leading data platform. The new SAP Application Data Foundation makes SAP application data available for consumption in SAP Business Data Cloud, including SAP Databricks, and provides a cloud-native way to access SAP application data, helping to lower total cost of ownership and reduce time to value. AI summary

Wealth management client onboarding requires a platform that connects case context across systems, coordinates work as conditions change, and preserves visibility and governance. Governed case context, which integrates evidence, case state, completed controls, open exceptions, agent activity, and decision history, enables supervised AI agents to coordinate the onboarding journey with governance, traceability, and human oversight. AI summary

Senior data and AI leaders are making deliberate, evidence-based decisions to balance momentum, accountability, and flexibility in order to turn AI ambition into sustained enterprise value, as the pace of AI progress continues to accelerate and expectations rise. AI summary

The article discusses the creation of an interactive Halloween projection, "MonsterHouse", using NVIDIA's Nemotron Voice Agent and various models. The system uses NVIDIA's Nemotron Speech ASR to transcribe visitor speech, Bolt E4B to generate responses, Magpie TTS Zeroshot to synthesize speech, and Audio2Face-3D to animate the character's mouth. The system is built on a single workstation and runs on NVIDIA's RTX PRO 6000 and RTX PRO 4000 GPUs. AI summary

Four, just four: that’s the key to break into any unpatched connection on the Internet that is using eMudhra certificates. Here’s how we quickly patched Cloudflare, what evidence we can see of attempted exploitation, and how much of the res…

Anthropic files S-1 with the SEC for proposed initial public offering, The sleep framework: A new way for LLMs to retain knowledge in continual learning, and many more!

Public traffic to private origins, with WAF, bot management, caching, and Workers, is now available to every customer with Cloudflare One connectivity.

An AI-powered tool called PlanAssist has been shown to speed up planning decisions by up to 58% in England, automating routine checks and allowing human officers to focus on more complex tasks. Developed by Google DeepMind, PlanAssist validates application documents, checks against heritage constraints and planning policies, and drafts assessment reports that can be edited by officers. The tool is now being rolled out to 12 additional local authorities across England, with plans for wider adoption in 2027. AI summary

Geist started as a monospace typeface for developers, built for the editors and terminals where they spend their day doing their work. From there it grew into Geist Sans for product interfaces, and then into Geist Pixel, which was inspired …

OpenAI will honor a shared robots.txt signal to help website operators communicate their AI training preferences to participating providers.

OpenAI is strengthening violence prevention in ChatGPT with expanded safeguards, earlier interventions, and a new commitment to public safety reporting.

Data Cache is exiting beta and moving to Runtime Cache . Next.js fetch caching and unstable_cache will switch over on every version without code changes. On Pro, each project and environment gets its own Runtime Cache storage allocation, wi…

Today, we’re bringing Data Cache into Runtime Cache, giving developers one product for data caching across frameworks and APIs. If you use Next.js fetch caching or unstable_cache , you won’t need to change your code. Pro and Enterprise proj…

Vercel Daemons can now call the services in their deployment through service bindings . Declare the binding on the daemon. It names the target service and the environment variable that receives the service's URL: The daemon calls that URL l…

OpenAI shares excerpts from court briefs by creators, academics, accessibility advocates, and legal experts, among others, on how fair use supports AI training and innovation.

Vercel Daemons now support configuring HTTP liveness checks. Start with an HTTP endpoint in your daemon that reports its health. Keep it lightweight so it responds quickly, and have it return an error status when the daemon is no longer hea…

Ling 3.1 Flash 1019 from InclusionAI is now available on AI Gateway , free until November 3, 2026. The updated model is designed for coding, multi-step reasoning, and agents that use tools. Its 1M-token context window supports workflows inv…

Real-Time Mode (RTM) on Spark Declarative Pipelines (SDP) is now Generally Available, enabling end-to-end processing latencies in the millisecond range for operational streaming workloads. RTM eliminates the need for separate specialized engines like Apache Flink, unifying batch, ETL, and low-latency streaming under a single framework. AI summary

Strix , the open-source AI penetration testing tool, can now run on any of the 400+ models on Vercel AI Gateway . Strix can reach models from Anthropic, OpenAI, Kimi, GLM, Deepseek, and others through a single API with no markup, removing t…

IP geolocation on Vercel now resolves visitor locations with finer network granularity, delivering city-level data for nearly all requests and postal-code data for 90% of traffic. Geolocation data is now refreshed more often, so locations k…

OpenAI’s updated Model Spec outlines five core values to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.

Over the last year we’ve been sharing our work leveraging large language models to catch bugs on the fly — a novel approach we call Catching Just-in-Time Tests (JiTTests). Now, we’re sharing the results and positive impact we’ve seen with C…

Serverless cost controls now incorporate three objectives: attribute costs by tagging serverless compute objects, control access through entitlements, and limit spend with rate limits. These controls are now available in Databricks, allowing teams to monitor and manage their serverless spend more effectively. AI summary

IBM has released Granite Switch 4.1.1, a patch update that brings incremental improvements to the modular language model architecture, including an updated Granite Speech ASR model for multilingual speech-to-text across 10 languages and an improved Decision adapter. This update is fully backwards compatible with version 4.1 and can be composed and deployed without recomputing the base model. The update demonstrates the speed and efficiency of IBM's model update process. AI summary

Cloudflare mitigated over 8,200 DDoS attacks per hour in Q3 2026 as attackers shifted to shorter, high-frequency strikes. Explore the latest threat data, including the collapse of CLDAP vectors and the impact of recent botnet takedowns.

Mistral AI has released Le Chonk, a 1-trillion-parameter multimodal model, now available with improved coding, cybersecurity, and reasoning efficiency, leading open-weight benchmarks in coding and defensive security, and offering three reasoning modes for cost-performance balance. AI summary

OpenAI uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior so responders can intervene before incidents escalate.

Ropes & Gray is developing AI for transactional diligence with OpenAI. Lawyers review grouped documents, summaries, and reports before advising clients.

On Vercel, the Vercel Flags SDK now initializes faster on cold starts, and usage reporting no longer keeps functions running after the response. Faster cold starts Two changes speed up SDK initialization on cold starts: The SDK loads and pa…

Procore has implemented OpenSharing on Databricks to power its construction analytics solution, doubling the catalog of extractable data for its customers across all projects and regions. This allows customers to pull data for their downstream systems and business processes hourly, rather than daily. AI summary

Procore, a construction management software provider, replaced its previous analytics product with OpenSharing on Databricks, doubling the catalog of extractable data for its customers. This change lifted the once-a-day refresh ceiling, enabling customers to pull data hourly, and doubled data availability, accelerating new product experiences and giving real estate facility owners, general contractors, and specialty contractors a single trusted view of every job. AI summary

Node.js 26 LTS is now Generally Available for builds and Vercel Functions . This release includes: HTTP/2 by default for built-in fetch ( Undici 8 ) lets multiple requests share a connection with servers that support it. This can reduce lat…

Credit Saison’s ChatGPT Enterprise pilot delivered 954% ROI. See how leaders and employees helped turn the results into companywide adoption.

Mistral Large 4 from Mistral now uses updated weights on AI Gateway , with lower pricing for input, cached input, and output tokens. According to Mistral’s release announcement , the updated weights improve agentic coding and defensive cybe…

The Vercel Plugin is now available in the Google Antigravity marketplace. Use the Vercel Plugin to give your Antigravity agent guidance on Vercel APIs and recommended patterns, with skills for products such as AI SDK , Vercel Functions , an…

You can now log or block requests from AI crawlers and AI assistants separately with the AI Bots managed ruleset in Vercel Firewall . AI crawlers: bots that collect content for indexing and model training AI assistants: bots that fetch page…

AnySpeechRecognizer is an open model for speech recognition on Apple platforms, providing a choice of recognition engines, including NVIDIA's Parakeet and Alibaba's Qwen3-ASR, with speaker diarization capabilities. The package mirrors Apple's Speech framework and uses Core ML models on the Hugging Face Hub. AnyDecisionModel is an open model for decision-making, using a trained decision head like Cloudflare's Clef, and supports multiple backends, including MLX, Core AI, and llama.cpp. AI summary

Mistral Large 4, a multimodal, multilingual, Mixture-of-Experts (MoE) model, takes text and images as input and generates text, with 1T total parameters, 49B active parameters, and a context window of 1M tokens. AI summary

Project ProtoStar aims to make publicly funded scientific research more usable by releasing datasets, models, and code in AI-ready formats, supported by open infrastructure and practical publishing guidance, with the goal of increasing the scientific value of the work released. The initiative is hosted by Hugging Face and brings together universities, research organizations, industry, and the open science community to support the release of scientific artifacts in machine-actionable formats. AI summary

You can now run CI for your Turborepo with one line of configuration. In Vercel's internal repositories, our wait time from git push to "CI passed" is 2-6x shorter, dramatically improving our CI's ability to keep up with our coding agents. …

Box Mount lets agents running in Vercel Sandbox read and write Box files through standard filesystem operations. It syncs a Box folder with a directory inside the sandbox. Unmount before you stop the sandbox, and everything the agent wrote …

Databricks has developed a platform that enables businesses to build voice agents that can understand customer context, adapt to individual needs, and make fast AI decisions in real-time, creating value through personalized customer interactions. AI summary

Google's Gemini 4 Carbon is its strongest model yet, featuring state-of-the-art cybersecurity and long horizon capabilities, and has achieved significant gains across various coding and agentic tasks, as well as key cybersecurity defense capabilities. AI summary

Today, we laid out our vision for GitHub as the most comprehensive, end-to-end software delivery system in the world. The post GitHub Universe 2026: Evolving every layer of the stack appeared first on The GitHub Blog .

See how a record year of growth reached every industry, every field, and every kind of builder in the 2026 Octoverse report. The post Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents appeared first on The …

This post describes the next generation of Meta Andromeda, Meta’s personalized ads retrieval engine, and how co-designing the model, system, and GPU hardware unlocks a new efficiency frontier for ads retrieval. In production, the system sco…

NVIDIA Nemotron 3.5 Transcribe is an open-weight, promptable model that transcribes 40 languages, translates 31 of them, and attributes speakers in 13, with top leaderboard results for transcription, speaker attribution, and translation. The model can be fine-tuned to adapt to domain vocabulary and accents, with a case study demonstrating a 45.8% relative reduction in word error rate on held-out speakers. The model can be fine-tuned using the NeMo Speech fine-tuning tools and Granary v2 data pipeline. AI summary

OpenAI and ServiceNow expand their partnership to bring OpenAI models and agents into ServiceNow and ServiceNow workflows into ChatGPT and Codex.

How AI might enter the org chart—beneath, beside, and between us.

Vercel Container Registry now supports up to 2 GB per compressed image layer, up from 500 MB, allowing larger layers to be pushed without splitting them. The total image size limit remains 15 GB. The higher layer limit is useful for images …

See how Legora uses OpenAI models, ChatGPT, and Codex to accelerate legal workflows, expand contract analysis, and free lawyers to focus on higher-value work.

Today, we are introducing Virtual Models on AI Gateway . Virtual Models are an abstraction of the model string and configurations your apps and developers use. With a single virtual model ID, you can change and update models, configure fall…

We’re highlighting 20 scientific foundations, institutions, and research groups around the world that we’ve partnered with in an expansion of our AI for Science program.

You can now query AI traffic in Vercel Observability to understand how AI bots are accessing your site. AI traffic insights show: Requests by verified AI bot: how many requests each bot makes Most-crawled pages: pages most frequently reques…

You can now use fx , our open-source coding agent harness, in T3 Code on web, desktop, and mobile. T3 Code is an open-source control plane for coding agents, allowing you to add agents like fx as providers for your coding sessions. The buil…

Vital Skills, an open-source Agent Skills catalog, equips Genie Code with healthcare and life sciences (HLS) domain expertise, enabling it to apply specialized tools and best practices. This integration improves Genie Code's performance on in-scope tasks, resulting in increased consistency, accuracy, and rigor, with fewer steps and checks missed. AI summary

Improving document extraction accuracy with ai_extract and MLflow Prompt Registry involves using schema descriptions for field-level rules, global instructions for document-level logic, precision mode for structural complexity, and citations to diagnose errors. A repeatable workflow can be built using Genie Code, which automates the core of this workflow, scoring extraction field by field against labeled data, proposing schema and instruction fixes as reviewable diffs, and re-running the eval to show before-and-after results. AI summary

Databricks Lakehouse Federation allows querying on-premises databases directly from Databricks with standard SQL, without moving data, and enables subscription to third-party data from the Databricks Marketplace, also without data movement. This approach lets users create federated connections to external databases, query them directly, and combine the data with third-party data in a single SQL query, without persisted intermediate objects. AI summary

LFM3 is a new generation of Liquid Foundation Models (LFMs) that combines text, vision, audio, omni, and decision models onto one shared backbone, allowing for specialized intelligence in various modalities. LFM3 models can be fine-tuned on a single GPU for specific tasks, enabling deployment on devices, private infrastructure, or the cloud. Six specialized models (text, vision, audio, omni, and decision) are built on top of the shared LFM3 backbone, each with its own unique architecture and training recipe. AI summary

Large language models (LLMs) can produce highly variable judgments of novel research ideas, with small changes in the evaluation setup resulting in dramatically different verdicts. This instability is particularly pronounced when evaluating AI-generated ideas, where the choice of prompt, reasoning budget, and model can significantly impact the outcome, and more search or reasoning does not necessarily produce more consistent judgments. AI summary

Genie 4, a new world simulation model, generates realistic and interactive environments, including photorealistic visuals, fluid movement, and native audio. The model simulates complex events, including human speech and reasoning, and is built on top of the Gemini family of models, which provide advanced reasoning capabilities. Genie 4 introduces a new standard of realism to real-time interactive simulation, enabling safe and infinitely variable simulations that can be used for various applications, including training environments for robots and agents. AI summary

This latest deep dive into Meta’s Privacy Aware Infrastructure (PAI) takes a look at PrivacyLib. A privacy requirement must hold as data moves through Meta’s diverse data and processing systems. Historically, supporting a new requirement me…

The 2026 Python Typing Survey received 1,794 responses. Type hint use changed little across three years, but checker selection changed sharply. Among respondents who reported which checkers they use, half selected more than one, and missing…

GPT-6 Astra tops Retell’s voice agent benchmark, while Codex helps it reach a $1.6M annual revenue run rate per employee.

An examination of global Internet disruptions observed by Cloudflare Radar during the third quarter of 2026. This review analyzes how power failures, natural disasters, and government shutdowns impacted network connectivity and traffic.

We are launching a PHP Capture the Flag challenge integrated with our Bug Bounty program to test our WAF protections. Security researchers are invited to bypass our defenses and capture the flag to help make the Internet safer.

Migrating from Google BigQuery to Databricks Lakehouse requires a strategic framework, encompassing a phased approach with governance, technology, and people elements. This framework enables organizations to unify BI, ETL, and multi-model AI workloads into one environment, reducing operational overhead and improving predictability. AI summary

Vercel and ElevenLabs are partnering to bring 14 ElevenLabs models to AI Gateway . To celebrate the launch, all ElevenLabs models are 50% off on AI Gateway for the first 14 days, through November 9th. AI Gateway supports six different types…

OpenAI and Oracle are expanding their partnership to bring business applications and AI agents into ChatGPT and power new Oracle offerings.

The fx adapter for the AI SDK’s HarnessAgent now runs the harness and model requests in the same process as your application. Built-in filesystem and shell operations remain isolated in the sandbox. The adapter uses libfx , the TypeScript l…

You can now deploy a plain root Dockerfile or Containerfile on Vercel's Container runtime when no other framework is detected. Vercel selects the Container preset, builds that file, and routes all traffic to the resulting image. To get star…

The Vercel Plugin is now available in OpenCode . The plugin provides on-demand guidance on Vercel APIs and recommended patterns, with skills for AI SDK, Vercel Functions, Workflow, and more. It also includes specialist agents for deployment…

With OpenAI Daybreak, Check Point speeds up the research behind new security protections, enabling its analysts to turn vulnerability intelligence into customer defenses faster.

We are expanding our partnership with the General Services Administration (GSA) through a new agreement that gives government partners a 40% discount on Claude Enterprise, Claude for Government, and the Claude API.

We're at the start of a major effort to speed up drug development with Claude. Over the past several months, our scientists have engaged in deep research partnerships with six organizations developing new medicines.

The AI SDK HarnessAgent layer now accepts image inputs across every harness adapter. Pass images alongside text in the standard AI SDK message format: a content array mixing { type: 'text' } and { type: 'file' } parts. Image input is proces…

Vercel Agent sessions in the dashboard and Slack now cost 67% less on average. We rebuilt how Agent manages context, reducing the tokens needed per session and improving prompt-cache reuse. These improvements are live and require no action.…

Lakeflow Connect unifies productivity and AI data from various sources, including Atlassian, Microsoft, Google Workspace, GitHub, and major AI providers, into a single governed foundation in Databricks. This allows for seamless ingestion of unstructured and structured data, enabling AI agents to provide grounded answers and powering analytics and BI dashboards. AI summary

A configuration mismatch in GLM 5.3 on Workers AI triggered an SGLang bug that exposed KV cache pages across requests. Here is how we addressed the issue.

How Greptile uses GPT-6 Astra to test user workflows, Daybreak to strengthen security, and Codex to reduce rework.

You and your agents can now browse and search more than 1000 services available in Vercel Connect without leaving your terminal. vercel connect discover searches by name, alias, keyword, and description. type: embedded-entry-inline id: H1F0…

Databricks Free Edition is a perpetually free version of the Databricks Data + AI Platform, designed specifically for learning and non-commercial use, allowing students and educators to use the same tools professionals use in production. It runs in the browser with no cloud contract or credit card required, supporting an entire data + AI curriculum, from introductory data science and statistics through data engineering, BI, machine learning, AI, and capstone projects. AI summary

AgentEnv , Scale AI's framework for testing agents against tasks and scoring their results, now runs on Vercel Sandbox . The Vercel provider ships in AgentEnv 0.9.1334 and later, and runs the agent and its task environment in separate Sandb…

Databricks AI Search now natively supports Streaming Tables, Materialized Views, and Managed tables using Iceberg v3+, enabling real-time intelligence, personalized experiences, and cross-engine AI applications without the need for separate synchronization pipelines. AI summary

At the 2026 Open Compute Project (OCP) Global Summit, Meta shared the networking innovations that are directly supporting our latest AI models, including Muse Wonder and Muse Spark, and our gigawatt-scale Prometheus cluster. Highlights IceP…

Learn AI literacy via Google’s latest AI Quests lesson on neuroscience, “Polar Peaks.”