Databricks has developed an integration with AWS HealthLake, allowing users to query HealthLake-managed FHIR data directly in Databricks, bringing Databricks governance, analytics, and AI capabilities to clinical datasets. This integration uses OpenSharing to securely expose data in HealthLake without requiring any export pipelines or data duplication. AI summary
Firehose
Filtered to Companies · clear filters
Browse: People · Companies · Papers · Podcasts · Hacker News · Deep dives
Browse by tag
Databricks' Feature Views unify batch and streaming features for training and serving, enabling real-time personalization with 34ms p95 latency. A real-time next-best-offer recommender for retail banking combines stable customer context with recent clickstream activity, using the same feature definitions for training and serving. AI summary
Hy4 Flash from Tencent is now available on AI Gateway with an introductory 50% discount on input, cached input, and output tokens through November 29, 2026. Hy4 Flash is the new lower-cost model in the Hy4 family. It keeps the full Hy4 feat…
Tabular foundation models (TFMs) utilize the transformer architecture to make predictions on new datasets without requiring bespoke training, accelerating predictive ML experiments. Databricks AI Runtime provides managed GPUs, MLflow tracking, and Unity Catalog integration to support experimentation and fine-tuning of TFMs. AI summary
Databricks has launched the AI Index, a publicly available resource that tracks enterprise AI model adoption, spending trends, and coding agent performance. The index combines anonymized usage data from 10,000+ companies using Unity Gateway with coding agent benchmarks based on real development tasks. AI summary
To create repeatable skills in Genie One, start by mapping the workflow for a specific task, considering factors like conditions, tools, data, and steps involved. Then, build the skill in Genie One through a conversation, describing the workflow in your own words and asking Genie One to interview you. Test the skill like Genie's first week on a new task, ensuring it delivers the desired output and correcting any issues in conversation. Govern the skill with Unity Gateway, a central platform for publishing, discovering, and governing skills, and share it with your team to make it reusable across the organization. AI summary
New tools, practical guidance, and partnerships will help prevent, detect and respond to online child sexual exploitation and abuse.
We’re expanding our ChatGPT for Academic Researchers program to offer 200,000 researchers worldwide free access to our frontier models.
Plan holiday road trips, get route tips, find local appointments, and search your saved lists with new Ask Maps features.
Explore how more than 1,800 scientists use ChatGPT in research. See findings on daily workflows, reported productivity, and expectations for discovery.
LVMH’s Media Division and OpenAI announce a partnership to support journalism with AI and make its publications more accessible through ChatGPT.
OpenAI begins embedded assessments with Apollo Research, METR, and Transluce, and Goodfire to independently evaluate model alignment, safeguards, and monitoring systems.
Introducing computer-use agents, improved clinical answers, and new healthcare plugins that work across healthcare software to get real work done.
Agents are now users of software, doing work on people's behalf. Whether you're building the agent or the software it uses (API, CLI, MCP server, SDK), evals give you a way to measure how well that work gets done. Agents probably missed you…
Hugging Face has partnered with OpenRouter to make open-weight models easy to run, allowing users to access the best provider for each request based on live measurements of price, speed, and reliability. This integration enables users to run open-weight models with the same API and token as before, while OpenRouter serves the models at scale across 80+ providers. The partnership aims to reduce costs and improve performance for open-weight models, which now cost about four times less on average compared to closed models. AI summary
The King’s Trust and OpenAI launch a £1m partnership to help up to 25,000 young people across the UK build practical AI skills for work.
Explore a candidate nonperturbative lattice definition of the Standard Model and how GPT-6 Astra helped turn the idea into a detailed proof.
The OpenAI science blog shares new findings, practical methods, and firsthand accounts from researchers using AI to advance scientific discovery.
Qwen4-Max from Alibaba is now available on AI Gateway. It is the first model in the Qwen4 series: a reasoning model with native vision and a 1M-token context window, built for coding, professional work, and long-running agent tasks. Qwen4-M…
Databricks has been positioned as a Leader and highest in execution in the 2026 Gartner Magic Quadrant for Data Lakehouse Platforms, recognizing the company's pioneering work in the lakehouse market category. The lakehouse category, pioneered by Databricks, brings together data engineering, data warehousing, streaming, BI, and AI on a unified, open, governed foundation. AI summary
Apache Spark is generally considered the default choice for most teams due to its unified engine for batch, streaming, SQL, and AI, which can now achieve millisecond latency with Real-Time Mode. In a Databricks benchmark, Spark Real-Time Mode matched or beat Apache Flink on p99 latency for stateless, aggregation, and join workloads. Flink remains a better fit for specialized cases such as very large keyed state or complex event processing on an existing Flink platform. AI summary
Vercel Daemons can now send OpenTelemetry traces to Vercel Tracing , allowing you and your agents to inspect background worker operations alongside the rest of your project's traces. Vercel provides each daemon with a Unix socket for receiv…
Kitesurf, our experimental browser engine for AI agents, is now open source. Inspect the code, run it locally, or deploy your own instance on Cloudflare Workers.
AI-orchestrated penetration testing platforms like ARTEX have become increasingly powerful tools enabling effective attacks on exposed infrastructure. Here we share Cloudflare’s observations and recommended countermeasures.
Claude Fable 5.2, Inference Performance Index: changing the way we measure inference capacity, a paper on Tools Need Not Speak in Text: Latent Context Compression for Agents, and many more!
Vercel Token Pass now includes $150 of tokens per month (up from $80) for the same $30 per team per month. The credit applies to featured open-weight models served through AI Gateway , including Kimi K3, Kimi K3.1 Pro, GLM 5.3, GLM 5.3 Flas…
To build a robot pick-and-place simulation in Isaac Sim, download the UR17G arm, Robotiq 2F-140 gripper, and beverage can assets from SimReady Central, which includes simulation settings and example scenes. Validate the assets by checking joint settings, movement, and contact behavior before running a pick-and-place task. Use an agent, such as Isaac Sim's remote Python server, to interact with the simulation and generate task code for the pick-and-place task. AI summary
XYBench is a benchmark of 8K+ questions that tests how large language models respond to requests based on misconceptions, measuring their ability to identify and address the underlying goal. The benchmark evaluates models' ability to recognize and redirect users toward effective solutions, focusing on pragmatic emphasis and misconception identification. The results show that current models often prioritize the stated request over the underlying goal, highlighting a need for improvement in model evaluation and training methods. AI summary
How Meta's Reality Labs brought Cuttlefish and its QEMU backend to macOS and Apple silicon. [...] Read More... The post Cuttlefish on Mac: Running Android’s Virtual Device Framework Natively on Apple Silicon appeared first on Engineering at…
OpenAI and Databricks have partnered to enable enterprises to deploy AI models that understand the business context and can act on it. This partnership combines OpenAI's frontier intelligence models (Codex, ChatGPT, and GPT models) with Databricks' governed data, business definitions, and access rules, allowing for three key enterprise adoption patterns: accelerating work for business users with ChatGPT and the Genie plugin, coding with enterprise-level governance through Codex and Unity Gateway, and transforming business processes with GPT models and Databricks Apps. AI summary
Introducing ChatGPT Wallet, a secure way for agents in ChatGPT Work, Codex, and dots to complete purchases with user approval and existing card protections.
The llama.cpp v1.0.0 release provides a stronger foundation for local AI ecosystems, enabling multi-modal LLMs to run on a wide range of hardware. The library supports over 150 architecture variants and all mainstream speculative decoding methods, with a focus on quantization, optimized backends, and a portable C/C++ implementation. This release is part of a larger effort to make AI inference accessible at every level of the software stack. AI summary
Spend Management adds opt-in spending caps for Cloudflare Workers and Durable Objects, and more products soon. Here is how we built the metering pipeline that turns usage into runtime decisions.
Shield uses OpenAI to help IT technicians hand off work in legacy portals. Its Forge agents navigate systems so technicians can focus on issues needing their judgment.
Databricks Real-Time Mode is a Spark Structured Streaming trigger that processes records continuously as they arrive, reducing latency to milliseconds. It's enabled by changing one trigger line and supports stateful operators, stateful aggregation, deduplication, and transformWithState. Real-Time Mode is generally available on Databricks and rolling out across open source Apache Spark releases, offering a faster alternative to Apache Flink for most workloads. AI summary
In October, we experienced eight incidents that resulted in degraded performance across GitHub services. The post GitHub availability report: October 2026 appeared first on The GitHub Blog .
You can now scan code changes for vulnerabilities with Vercel Security Review in GitHub repositories that have not yet been linked to a Vercel project. By default, Security Review covers all unlinked repositories the Vercel GitHub App can a…
Parallel Decisions Fast from Parallel is now available on AI Gateway and is free through November 23, 2026. It searches the live web to answer typed questions about companies, products, policies, and other changing information, returning pr…
The build cache size limit on all Vercel build machines has been increased to 5 GB. This allows projects to save more installed dependencies and framework files for reuse between deployments, which can reduce install and build times. Build …
Vercel Sandbox now initializes up to 4x faster outside the Washington, D.C. ( iad1 ) region when reused image data is served from a regional cache. Initialization now reads cached image data close to where the sandbox runs, avoiding round t…
To build strong safety evaluations for AI models, it's essential to assemble a team with diverse skills, including subject-matter and technical experts, to ensure evals cover realistic and diverse test cases. Evaluations should be run safely in a sandbox environment with real-time monitoring to prevent unintended harm. AI summary
Agent skills capture repeatable engineering work in files teams can review and reuse, while tools, project context, and permissions remain specific to the host that runs them. The post Build once, run on any agent: A practical guide to agen…
Lakebase's LTAP design stores data in a single, durable copy in the lake, using a hierarchy of caches and object storage to serve both operational data and analytics, eliminating the need for a costly CDC pipeline. AI summary
Token Pass now includes Kimi K3.1 Pro , Moonshot AI’s most capable model for coding and agentic workflows, with vision, tool use, and a 1M context window. Token Pass is an Early Access subscription for Pro teams that includes Kimi K3, GLM 5…
Built primarily on OpenAI models, Amplitude’s proactive product agent, Wave, identifies what to build, helps teams ship changes with coding agents, and measures the impact.
Constellation is building a governed data foundation on Databricks using Unity Catalog for lineage, access control, and shared business context, enabling unified governance and semantics across the enterprise. AI summary
Eclipse Aerospace simplified flight-data ingestion with Databricks by leveraging PySpark custom data sources, which enabled them to bring proprietary flight telemetry into Databricks without a separate ingestion platform. This approach reduced latency, cost, and operational overhead, while improving observability and engineering efficiency. AI summary
Researchers have developed LLMON, a markup language for LLMs to explicitly communicate structure and semantics, addressing the brittleness and vulnerability to prompt injection attacks of current LLMs. The LLMON models, derived from Granite 4.1 base models, maintain general capability while eliminating prompt injection attacks by 47% and 95% compared to non-LLMON models on benchmarks. LLMON enables explicit instruction/data separation, indirect prompt injections are mostly ignored, and unrelated passages do not get pulled into the answer. AI summary
Google has introduced Gemini 4 Hydrogen, a new AI model designed for everyday coding and agentic workflows, offering a balance of speed, intelligence, and cost across four thinking levels. This model is intended to be used in conjunction with Gemini 4 Carbon, a deeper, more complex model, and is available to everyone today, including free users in the Gemini app and Google Search. AI summary
We are bringing Gemini 4 Hydrogen to Google Search globally to provide you with more helpful, visual, and intuitive responses.
H6.ai uses Lakebase Postgres to build AI-powered financial automation by providing every customer with a dedicated database, isolating their financial data, and enabling frequent regression testing at scale through copy-on-write branches, disposable schema-migrated branches, and a panel of AI judges. AI summary
Lakehouse//RT scales real-time analytics without increasing costs by supporting high concurrency, faster query execution, autoscaling in small increments, and eliminating duplicate infrastructure. This approach enables customers to deliver real-time performance with sub-100ms latency and thousands of concurrent queries, while reducing costs by 6x to 10x compared to separate real-time serving layers. AI summary
Databricks Genie is now generally available in Slack, allowing users to get governed answers grounded in their business context without leaving the Slack workspace. Genie can update documents and tickets through MCP connectors after user approval, analyze files, and run analyses on a schedule. Governed by design, every question runs with the asker's permissions, and responses are visible to everyone in the channel by default, with the option to set responses to be private to the person who asked. AI summary
Databricks' Unity Gateway powers collaborative AI agents with Block's Buzz by providing centralized access to frontier and open-weight AI models, allowing employees to experiment with different combinations of models and agent harnesses to find what works best for each task. This flexibility enables teams to work with multiple agents in one place, without having to switch between different AI tools. By integrating Unity Gateway with Buzz, Block can efficiently manage and optimize AI costs and performance. AI summary
When we first launched Workflow SDK, the goal was simple: write regular async JavaScript, and let the SDK keep it running through crashes, deployments, and long waits. Last year, we shipped just that , with the world-class developer experie…
Gondola helps creatives get credit and get hired, with Codex raising its lead engineer’s output by 50% and ChatGPT simplifying daily work.
With Codex, Holiday Extras is turning customer expertise into better apps, higher booking conversion and new possibilities for growth.
Agents are now users of software, doing work on people's behalf. Whether you're building the agent or the software it uses (API, CLI, MCP server, SDK), evals give you a way to measure how well that work gets done. Agents probably missed you…
Vercel Sandbox can now install your team's private npm packages using credentials stored in team Shared Environment Variables . Custom registries such as GitHub Packages and JFrog Artifactory are also supported. Pass credentials: { npm: tru…
Inkling 2 Small from Thinking Machines is now available on AI Gateway. The multimodal reasoning model supports coding agents, visual understanding, and professional knowledge work, with text and image inputs and text output. Inkling 2 Small…
Python 3.15 is now available as a Vercel Managed Image for Vercel Sandbox . To run a sandbox with Python 3.15, select vercel/sandbox/python:3.15 using the image option in either SDK or the --image flag in the CLI. The default universal imag…
Gemini 4 Hydrogen from Google is now available on AI Gateway. Gemini 4 Hydrogen succeeds Gemini 3.8 Flash, improving agentic coding, knowledge work, and tool use while staying in the same latency class. It handles longer engineering tasks a…
This article provides a fine-tuning guide for 3D generation models, specifically TRELLIS.2, using low-rank adaptation (LoRA). The model is fine-tuned on a subset of the Google Scanned Objects (GSO) dataset, which has 188 footwear objects. The fine-tuning process involves adding LoRA adapters to the sparse-structure flow DiT and the shape flow DiT, which control the object's shape and geometry. The adapters are trained using a rectified flow matching objective, and the original model weights are kept frozen. The fine-tuning script and containers are available for local and Hugging Face Jobs deployment. AI summary
OpenAI commits $2 million to research on AI and children’s safety. The work supports shared tools to measure risks and opportunities for young people.
A conversation with Sam Nickless, Partner and Chief Executive Officer at Gilbert + Tobin.
Here's a summary of the article for a developer/AI-ML audience: Detecting credential-driven attacks on Databricks involves identifying suspicious activity patterns that deviate from normal user behavior. A six-stage model outlines the typical progression of an attack: (1) gaining access with valid credentials, (2) mapping the environment, (3) harvesting more credentials, (4) finding a workable execution path, (5) establishing persistence, and (6) collecting data and cleaning up access. To detect these attacks, look for non-linear event sequencing, unusual activity patterns, and changes in user behavior. Utilize Databricks' verbose audit logs and the Detection App's Genie agent to spot these patterns. AI summary
A Lead Solutions Architect at Databricks spends their typical week balancing building tech, coaching colleagues, and partnering with executive management on business cases, often switching between coding, scaling applications, and collaborating with global teams on complex customer challenges. AI summary
Databricks has expanded its AI research and opened a new 235,000-square-foot headquarters in San Francisco, solidifying its commitment to the city as its home base for data and AI innovation. AI summary
The Gemini Decisions API is a fast, calibrated multimodal decision-making tool that can evaluate up to 1,000 yes/no, choice, and rubric-scoring questions in a single pass across text, images, audio, video, and PDFs. It achieves 150-250 ms end-to-end latency at $0.10 per 1M input tokens with zero output tokens billed. AI summary
Ads ranking models are growing toward LLM scale faster than the hardware to run them can keep up. An extra unit of compute has more impact on some ad requests, models, and chips than on others. The challenge is deciding where to allocate it…
Bright Machines captures PostgreSQL changes at the factory edge with Debezium, writes them as Apache Iceberg, sends them out over a single auditable outbound path, and reads them in place in Databricks through Unity Catalog federation. This setup enables customers to see live, governed dashboards of their factory data without opening any inbound access to the air-gapped plant. AI summary
Databricks introduces Genie, an autonomous AI coworker that handles long-running enterprise tasks on behalf of users, operating within established guardrails. Genie Ontology provides trusted business context, while Genie Search finds information across various systems to deliver accurate and complete answers. Unity Gateway controls tool access and sets spending limits, ensuring autonomy within set boundaries. AI summary
Tabletop is a new Google Labs experiment for group collaboration. Build interactive itineraries, polls, and brackets with your team. Join the waitlist.
Google DeepMind has released AlphaGenome 2, a next-generation DNA sequence model that predicts how genetic variants affect specific human cell types, offering a powerful tool for researchers to uncover the cellular drivers of disease. AlphaGenome 2 improves upon its predecessor, AlphaGenome, by zooming in on individual cell types and cell states, and redesigning the model's architecture to better capture subtle differences between closely related cell types. The model can predict variant effects and gene regulation for new cell states, such as those that occur from disease or vaccination. AI summary
Genie One is an AI cowork home that brings together search, organizational directories, and artifacts to empower business users to find insights, automate work, and scale confidently with Unity Gateway. It integrates with various tools and platforms, including Databricks, and offers features like autonomous action, MCP connections, and shareable skills to automate workflows and reduce ad hoc requests to the data team. AI summary
Genie with Genie Ontology achieves 91% accuracy on a new enterprise benchmark, outperforming two leading frontier agents by 46% and 33% respectively, while using less than half the tokens and finishing three to four times faster. AI summary
Databricks has introduced Genie plugins that enable AI agents like ChatGPT, Claude, Microsoft Copilot Cowork, and Grok Bot to access governed business context from Databricks, improving the accuracy and relevance of their responses. The Genie plugins work natively in these agents, allowing them to draw on Genie Ontology for certified definitions and relationships, and write and run SQL queries on relevant data. This integration enables agents to provide more trusted insights and better support business users. AI summary
From enterprise context to action: Genie Agents now analyze more data sources to transform your work
Genie Agents now analyze unstructured data from Unity Catalog volumes, enterprise tools, and documents, enabling autonomous work and actionable insights. They can reason across governed tables, thousands of documents, and enterprise tools, and take action in the tools teams already use, without manual intervention. AI summary
Databricks has released Genie Ontology, a Generally Available agentic ontology that combines learned and modeled knowledge, ranking it by authority and delivering the most trusted context to agents while respecting existing user permissions. This release introduces Genie Search, Ontology Snippets, external knowledge sources, and organizational directories, expanding Genie Ontology's capabilities to connect to a wide variety of sources and provide a unified understanding of the business. AI summary
Cloudflare is building an experimental consensus service using a never-before-deployed consensus algorithm, QuePaxa. In this post we explain how QuePaxa works and why we chose it.
We’ve spent a lot of time over the past two years talking about ways to improve MCP, add new features and capabilities, and get on a more predictable release cadence. Clearly, there has been no shortage of ideas both from the maintainer and…
A new online archive of Esther Mahlangu’s life and art is now on Google Arts & Culture.
More than 20,000 veterans and transitioning service members have claimed a free year of ChatGPT Plus. See how they use it to prepare for civilian careers.
BotPsy, a 0.8B-parameter agentic language model, achieves state-of-the-art (SOTA) tool-use and agentic performance for its size, outperforming models up to 1.2B parameters. It excels in instruction following, multi-turn conversations, and multi-step tasks, but struggles with open-ended search and complex information retrieval. BotPsy is designed for secure on-device workflows, providing exceptional instruction-following reliability and efficiency while keeping user data private. AI summary
Gemini Decision 1 from Google is now available on AI Gateway. The multimodal decision model works with text, images, audio, video, and PDF documents, returning Boolean probabilities, choices, and scores. Use it to route requests, check poli…
Token pooling-aware training for multi-vector embedding models improves retrieval quality by teaching the encoder to preserve information when its vectors are merged, helping to retrieve better from the same storage budget. By applying a loss wrapper to a multi-vector retrieval loss, pooling-aware training can be used with existing models, resulting in a 97.8% retention of unpooled NanoBEIR nDCG@10 with a 4x reduction in document vectors. AI summary
JioStar’s Vijay Seshadri explains AI discovery at 500M-user scale. Read his lessons on evals, guardrails, and taking AI into production.
Gemini Intelligence is rolling out to more devices. See how expanded capabilities across these devices help you do more.
Unity Catalog Metric Views are now available in Power BI, enabling a unified, governed semantic layer for reporting and AI. This allows for consistent results across Power BI, AI/BI dashboards, and Genie, and keeps metric views and semantic models in sync with DirectQuery and Power BI tasks. AI summary
Git is releasing updates addressing six security vulnerabilities affecting recursive clones, reference handling, merges, binary patches, and imports. Here is what users need to know. The post Git security vulnerabilities announced appeared …
Your algorithm is a suite of AI-powered tools that makes Facebook recommendations more transparent and steerable, with a single editable view of the interests that shape Feed and Reels. We’re sharing how Your algorithm turns people’s natura…
Triton Low-Level Extensions (TLX), is an extension to Triton that helps engineers optimize GPU performance while preserving Triton’s productive programming model. TLX introduces explicit warp-group and cluster-level orchestration, letting d…
Gemini 4 Carbon, a new frontier model, is now available to developers, enterprises, scientists, and consumers, offering higher-precision knowledge work, improved agentic coding, and stronger cybersecurity capabilities. It pairs frontier accuracy with practical production economics, ranking #1 on independent benchmarks and being more than 2X lower in cost per task compared to the next best external model. AI summary
Google and Adobe are integrating creative and productivity tools. Edit PDFs in Drive and use Adobe AI in Gemini. See what’s new in this partnership.
Google Cloud launches Gemini for healthcare to unify clinical data and automate workflows. Reduce administrative burnout. See how it works.
Google DeepMind researchers developed AlphaProof Code, a novel agentic framework that combines human expertise with AI agents to formally verify complex software at unprecedented depth, successfully verifying the code of Signal's cryptographic protocol, a widely deployed cryptographic protocol protecting billions of private conversations. AI summary