Firehose

A team shifted their open-source project's development workflow to Buzz, a platform that allows compute to be pooled and shared within a community, powered by MeshLLM. The shift resulted in a 56% increase in PRs merged, a doubling of PRs opened, and a 66% increase in merged line churn, indicating improved code quality and a reduction in tedious work. This setup, facilitated by volunteers and existing agents, demonstrated a significant boost in throughput without requiring significant process changes. AI summary

Vercel Connect now includes a managed connector for Microsoft Teams . Creating one gives your organization a Teams bot that your apps and agents run. People can @mention it in channels or message it directly, and your code receives the mess…

We’re open-sourcing Rebalancer, the assignment-problem solver that has been used to solve resource allocation problems throughout Meta for over nine years. Rebalancer separates several related concerns: how to specify an assignment problem,…

Deployments now show their billable duration and CPU minutes in the dashboard, vc inspect , and the REST API . Use it to understand how each build contributes to your usage. Billable duration is the build and post-build time combined, round…

The authors reformulate block removal from large language models as a constrained binary optimization (CBO) problem, equivalent to finding low-energy states of an Ising glass, a disordered spin system with all-to-all interactions and a fixed number of "up" spins. This approach allows for efficient pruning of models by computing the energy of candidate configurations and using classical and quantum-inspired solvers to find good solutions. AI summary

Python Workers allow developers to run Python web frameworks and AI orchestration libraries natively in the Cloudflare Workers runtime. You can seamlessly integrate with Cloudflare's ecosystem including D1, R2, and Workers AI without writin…

Petal, the next step in Meta’s subsea innovation, will be the first subsea cable to deliver petabit capacity at transoceanic distances, connecting France and the United States over approximately 7,000 km (4,300 mi). Expected to enter servic…

OpenAI is working with an independent Advisory Group on Mathematics and Artificial Intelligence to guide the review and communication of emerging AI results.

OpenAI outlines a path to shared global AI standards, calling for coordinated evaluation, reporting, and governance to improve safety.

Explore new OpenAI Academy learning paths for employees, developers, leaders, educators, and students to build and demonstrate practical AI skills.

Grok 4.7 from SpaceXAI is now available on AI Gateway and 40% off through September 27. The discount applies automatically when you call spacexai/grok-4.7 . Grok 4.7 has a 500K token context window and supports low, medium, high, and xhigh …

The Hugging Face team has released a new version of the tokenizers library (v1) with improved performance, including a hand-written splitter instead of a regex engine, a cache that answers repeated words without merging, and a merge loop that never touches the allocator. The library now scales across cores and processes multiple pre-token spans in one call. The new version is faster than the previous version (v0.23) by 3-30 times, depending on the model and input size. AI summary

Using GPT-5.6, V7 turns scattered company files into context agents can use to complete complex, source-linked work.

RADAR is a four-stage system for catching gray failures, which are partial, silent outages that can cost customers and revenue without being detected by traditional monitoring systems. The four stages are: (1) reliability metrics, (2) anomaly detection, (3) alerting, and (4) root-cause analysis, which can be applied to any metric, including billing, conversion, or model performance. RADAR can be built on Databricks using native components and an AI-agent scaffold, allowing developers to catch gray failures in minutes at over 90% precision and 95% faster discovery. AI summary

Enterprise teams on Flexible Commitment plans can now use Spend Management , already available on Pro , at no additional cost. You can set a budget at any time in Spend Management settings . Set a budget per billing cycle, and when your tea…

mcp-handler now has experimental support for WebMCP , the proposed web standard for exposing tools to in-browser agents. Add a single script tag to your site, and your existing MCP tools become available there too. Opt tools in by adding th…

Databricks enforces corporate data security on personal devices through a four-layered strategy: device management, identity and access, zero trust, and application management. This approach ensures that company data remains secure while respecting user privacy, and it's exemplified in the deployment of the Genie mobile app, which is built and secured using Databricks' own infrastructure. AI summary

Cloudflare's global network is immense but not limitless. As we look for small ways to trim our resource usage, we sometimes get lucky and we can cut significantly more. Here’s how we reduced one of our Pingora-based service's RAM usage wit…

v0 now installs private packages from npm and custom registries using credentials stored as shared environment variables on Vercel. This makes it easier for teams to build with their existing design systems, component libraries, and interna…

We Must Pace the Frontier, ToolGrad: Efficient tool-use dataset generation with textual "gradients", a paper on The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement, and many more!

We dive into these questions and other AI hot takes on the latest episode of the GitHub Podcast. The post Should you read the code, is RAG dead, and did Skills kill MCP? appeared first on The GitHub Blog .

We are expanding our AI & Economy team with world-class academic advisors, fellows, and core internal researchers.

Google worked side-by-side with designers Jane Wade and Sergio Hudson to custom-design Google Flow tools to prep for NYFW.

OpenAI introduces the Australian Youth Safety Blueprint, a six-pillar roadmap for safer AI experiences that protect and empower young people.

Within 24 hours of launching on AI Gateway, Jev from TypeSafe AI reached more than twice as many paid teams as any previous model launch, making it the fastest-adopted model in gateway history. Jev passed every other comparison model in its…

In August 2026, Hacktron reported what looked like a remote code execution (RCE) vulnerability in Next.js image optimization. Their investigation found that the vulnerable code was not in Next.js itself, but upstream in libheif, an AVIF ima…

Anthropic is partnering with Accenture on embedded evaluation of its frontier AI models, which will include evaluating and red-teaming models, conducting alignment assessments, and testing model safeguards. The partnership aims to embed evaluators within Anthropic, providing them with access comparable to an employee's to assess how the company operates and verify its safety commitments. The partnership is non-exclusive, and Anthropic plans to work with other evaluators under different funding arrangements. AI summary

GLM 5.3 FlashX is now available on AI Gateway. GLM 5.3 FlashX is a high-speed serving option for Z.ai's multimodal coding model, delivering inference at ~200 tokens per second for faster streamed responses. The higher serving speed is usefu…

You and your agents can now deploy static artifacts to Vercel in under one second through Vercel CLI. Run vercel deploy to share a prototype, publish an HTML report, or preview a page created by your coding agent. Vercel automatically detec…

You can now opt into Turbo build machines on any individual deployment. This is useful when you need to increase resources temporarily without changing project settings. You can do this in three ways: Include #VERCEL_BUILD_MACHINE=TURBO in …

Google and the UN system have launched the UN System Data Commons, a new open platform making global statistics accessible and easy to search.

You can now run Harbor evals on Vercel Sandbox. Harbor is the open-source harness behind Terminal-Bench , whose registry includes many other benchmarks such as SWE-bench, tau3-bench and OSWorld. Pass --env vercel to harbor run and each tria…

[email protected] adds Notion skills databases as an install source for agent skills . Notion skills are reusable agent skills written as Notion pages. Teams author, review, and update them in the workspace they already use, then install them in…

AI agents require a database that supports continuous, concurrent reads and writes across multiple memory types, not the one-request-at-a-time pattern traditional apps use. Five evaluation criteria define a production-ready agent database: branch isolation, serverless scaling, hybrid search, ACID guarantees, and unified platform access. AI summary

Developers, a new layer called Omnigent in Databricks enables engineers to define an agent once, including the model, tools, policies, and limits, and run it across any harness, reducing the need to rebuild and manage multiple instances. Omnigent integrates with the Foundation Model APIs for unified cost and governance tracking. Additionally, a new web search component called Nimble, which can adapt to specific use cases and self-learn the best retrieval methods, can be integrated to improve the accuracy and efficiency of web search. AI summary

AIOps (Artificial Intelligence for IT Operations) combines AI and machine learning with observability to automate IT operations, detecting anomalies, correlating events, identifying root causes, and automating incident response, ultimately reducing downtime risk, cutting alert fatigue, and accelerating decision speed. AI summary

Trade-lifecycle modernization is now a priority due to cumulative pressure from growing data volumes, higher expectations for real-time insight, AI initiatives moving toward production, and shorter settlement cycles. Firms must connect research, trading, risk, operations, and compliance on governed data to unlock repeatable value from AI. AI summary

Cooley built GO Public with ChatGPT Work to bring intelligence to the IPO process, helping lawyers surface issues earlier and focus judgment where it matters most.

AI Gateway Production Index — September 2026 Every month, AI Gateway routes tens of trillions of tokens between production applications and AI labs. That traffic gives us a view of what AI usage actually looks like in today's enterprise, an…

A rewrite this size wasn't affordable before agents. Here's what porting the Copilot agent runtime to 800,000 lines of production Rust actually took. The post Migrating the GitHub Copilot runtime to Rust, using Copilot appeared first on The…

GPT-Live 1 from OpenAI is now available on AI Gateway. GPT-Live 1 is a full-duplex voice model and can listen and speak at the same time. Many voice models use turn detection to respond. Full duplex removes that boundary, so a user can paus…

OpenAI for Law brings frontier intelligence for law, custom firm workflows, connected legal data sources, and legal-grade controls for confidential client work.

You can now connect native Marketplace resources to custom environments . Previously, resource connections could only target production, preview, and development environments. Choose custom environments when connecting a resource from the V…

Anthropic has introduced the Life Sciences Verification Program (LSVP), a beta program offering refined safeguards for biology-related work, allowing life science professionals to access Anthropic's Mythos, Opus, and Sonnet models. LSVP grants are available for teams and institutions, with a verification process reviewing research credentials, security standards, and ethical research oversight. The program's safeguards aim to protect against access compromise, insider threats, and agent misuse, with monitoring usage against intended use cases and data retention for 30 days to identify potential misuse. AI summary

A modern storefront can look healthy while malicious JavaScript quietly siphons revenue, hijacks clicks, or rewrites analytics. See how Cloudflare's machine learning models surface evasive client-side attacks for analyst investigation.

Hobby projects now retain fewer deployments past the 30-day retention window. Hobby teams get 10GB of Deployment Storage . Every deployment you keep uses some of it, and going over the limit can block you from deploying until you free some …

MATCH_RECOGNIZE is a new SQL operator available in Public Preview that allows detecting patterns and sequences from event data using regex-like pattern-matching, simplifying pattern detection and sequence analysis across various industries. AI summary

OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.

Mem0 is now available as a native integration on the Vercel Marketplace , giving your AI agents and apps long-term memory. Mem0 remembers user preferences, facts, and context across sessions, so your app stops starting from scratch. Install…

Builds using Secure Compute or Static IPs now start 64% faster, with the average time from deployment creation to build start dropping from 6.7 seconds to 2.4 seconds. Previously, each build waited for a new build container to boot with its…

OpenAI and AARP are bringing free, hands-on ChatGPT workshops to 1,000 older adults across 10 U.S. cities to build practical AI skills safely.

Explore new AI-powered advertising experiences from OpenAI, including Sponsored Agents, tools for marketers, and integrations with HubSpot and Shopify.

GPT-6 Astra helps Hex’s data agents turn answers into interactive visualizations that employees are proud to share.

Learn how ChatGPT Work and Codex analytics help teams understand AI usage and spend, identify training needs, and connect adoption to business outcomes.

Open, private and multilingual AI is coming to your web browser. Mistral and Mozilla team up to put powerful, trustworthy AI where you already browse.

New OpenAI Economic Research shows how workers use AI beyond traditional roles and which new activities become recurring parts of their work.

Jev from TypeSafe AI is now available on AI Gateway . Jev is a probabilistic decision model for software: state goes in, typed Choice, Score, and Boolean answers come out. Regular language models generate text one token at a time, which the…

Is Agentic reports now let you view your checks through one of four site types: Docs & content, Business, App, or Commerce. For example, the Commerce view highlights payment and checkout standards like x402, UCP, and ACP, while the App view…

Google has launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, its most advanced live dialogue models yet, offering more intuitive and intelligent conversations. These models handle complex reasoning, real-time visual context, and background task execution without interrupting the conversation. AI summary

Databricks' Genie and AI business processes can operationalize ML insights into governed business operations for energy theft detection, enabling teams to accelerate the loop from flagged meter to recovered revenue to a safer household. A governed workflow connects interpretation, investigation prioritization, dispatch-ready reporting, and recovery workflows, with Lakebase maintaining live case state and recovery totals. Trusted answers and metrics are provided through Genie One, Unity Catalog, and Unity Gateway, enabling leaders to ask questions in plain English and receive answers grounded in governed data. AI summary

Researchers at IBM developed a method to improve the consistency of large language models (LLMs) like GPT-4.1, which can significantly impact their reliability in mission-critical applications. By analyzing an agent's past trajectories and identifying "flat" decisions, where the model is uncertain, they created a new type of guideline that helps stabilize these decisions. This approach, called consistency guidelines, can improve the Pass^5 metric, which measures the fraction of tasks an agent succeeds on all runs, by up to 22.9 percentage points. AI summary

We’re moving beyond traditional text translation to build models that understand the world’s rich, living languages exactly as they are expressed.

The true measure of AI is who it helps. Here’s how it’s impacting lives today. We're focused on key areas where advanced technology can help make extraordinary progress …

Explore this collection to see how experts and local leaders are using AI breakthroughs to ensure everyone can share the opportunity of AI.

You can now scope access to individual Workers and assign narrower Developer Platform roles, so teammates, CI tokens, and agents get only the access they need to debug, deploy, or monitor safely.

Cloudflare is giving site owners a way to stay discoverable while disallowing AI training. New controls and an Accountable designation establish a shared model with Apple, Google, and Microsoft.

We’ve translated ATLAS’s millions of global data points into an interactive, open-access experience.

Delphi on Vercel 10 engineers with no dedicated infrastructure role Everyone ships code, including product and design 100+ production deploys a day behind feature flags Delphi builds digital minds. They capture what someone has written, rec…

Databricks' marketing team uses Genie, an AI analytics assistant, 3x more often in decision-making, with over 85% adoption across the marketing organization. They achieved this by building a governed Marketing Lakehouse, documenting data and business context, encoding verified answers and examples, teaching Genie the language of their business, and continuously evaluating and improving the system through user feedback. AI summary

Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking from Google are now available on AI Gateway. Both models support real-time spoken interactions for voice assistants, conversational experiences, and applications that respond through aud…

Managed Postgres should take routine database operations such as patching, scaling, failover, and backups off the database team's plate. Lakebase automates these operations on serverless infrastructure with automatic scaling, scale-to-zero, point-in-time recovery, branching, pgvector, and PostGIS. AI summary

The AI SDK harness layer now supports authenticating harnesses through their native subscriptions, where the underlying harness supports them. The harness layer runs different coding agents through the same HarnessAgent interface, so you ca…

Databricks now supports on-demand state repartitioning for Apache Spark Structured Streaming, allowing users to resize partitions without rebuilding checkpoint state, enabling more flexible tuning and scaling of stateful streaming queries. This feature is available in Databricks Runtime 18 and above with the RocksDB state store provider. AI summary

Christina Koch sits down with James Manyika, Google’s Senior Vice President of Research, Labs, Technology & Society.

DevFest 2026 is back and here’s how you can connect with one of the more than 800 global events to build, secure, and scale in the agentic AI era.

Fyxer uses OpenAI models, fine-tuning, memory, and real user feedback to organize inboxes and draft emails in each user’s voice.

Perplexity uses Astra to write communications, change software, and monitor production systems, and checks in much less frequently than with earlier models.