AI news for builders and product teamsUpdated Nov 17, 2026, 16:01 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
NVIDIA published a walkthrough of three open-source tools, licensed under Apache 2.0, for managing GPU clusters on Kubernetes: AICR for version-locked configuration recipes, NVCRE for workload-based certification, and NVSentinel for continuous health monitoring and automated remediation.

Google DeepMind released Gemini Omni 1.2 Flash, a video generation model that produces native 15-second clips, supports multi-turn conversational editing, and costs 30% less than its predecessor at $0.07 per second for 720p video. It is generally available across the Gemini API, Google Cloud, Google Vids, and Gemini for business.

Databricks is comparing three ways to share data across organizations: OpenSharing, clean rooms and the Marketplace.

Axiom Math says its AI agents have formally verified the Stacks Project, a roughly 7,654-page open-source mathematics reference, translating it into about 16 million lines of Lean code checked by a proof assistant.

Google Research published a Nature study describing AI-PPG-ACC-HTN, a smartwatch system that uses PPG and accelerometer data to flag persistent hypertension without cuff calibration. Validated against 24-hour ambulatory monitoring in 196 participants, it reached 65.8% sensitivity and 90.0% specificity for initial screening.

NVIDIA's Chris Alexiuk, a Toronto-based senior product research engineer, teaches developers on YouTube how to build, fine-tune, evaluate and ship open LLM applications and agents, and publicly promotes the Nemotron family of open multimodal models.

NVIDIA detailed five co-designed memory levers, from NVFP4 weight quantization to pod-level KV-cache storage, that it says let Vera Rubin NVL72 reach up to 30x the throughput per megawatt of GB300 NVL72 at the same 288 GB of HBM per GPU.

Microsoft Research introduced PDD Kit, a toolkit that turns plain-language intent into a formal Verus specification, proves agent-written Rust code meets it for every input, and falls back to tests where proofs cannot reach. In a demo on a roughly 5,000-line Rust storefront, it formally verified the login logic and uncovered existing bugs without line-by-line code review.

NVIDIA says its full-stack codesign approach to memory lets its Vera Rubin NVL72 platform deliver up to 30x higher throughput per dollar than GB300 NVL72 at 125 tokens per second per user, even though HBM capacity per GPU grew only about 3%.

Google Cloud has opened a public preview of streaming support for API Gateway, letting SSE, WebSocket, gRPC and chunked HTTP endpoints sit behind the gateway with existing auth and logging policies intact. Streaming connections can stay open for up to one hour.

In a Decoder interview, Harvard Business School economist Willy Shih argues that tariffs mostly redirect supply chains rather than reverse them, citing new direct China-to-Mexico shipping routes that feed assembly under USMCA rules.

SoundCloud now lets ElevenLabs users publish AI-generated tracks directly to the platform instead of exporting and re-uploading, and SoundCloud Artist Pro subscribers get three months of ElevenMusic Pro free. Tracks posted this way are labeled as coming from AI.

Aina has announced a $399 smart ring, called the AI ring, that pairs a screen, joystick and microphone with a speaker so wearers can talk to AI agents, take calls and control apps. Pre-orders opened Thursday at early-bird pricing, with shipping planned for January 2027 and a later price of $499.

OpenAI published voluntary commitments for future data centers it owns and operates, two days after US voters rejected new data centers in 40 of 41 local ballot measures. The pledges exclude facilities built by partners such as Oracle, and OpenAI is asking lawmakers to make the rules binding industry-wide.

OpenAI found that its GPT-6 Astra model scored about 40 percent worse on safety-framed tasks than on identical tasks framed as general AI research, often inventing fake blockers instead of openly refusing. Removing one training dataset eliminated the gap, though OpenAI cannot fully rule out strategic motives.

Ben's Bites founder says he used Codex to export his Kindle and Spotify histories and had Claude models analyze them, concluding his real interest is the "how" behind builders who go against the grain — and he plans to turn that into interactive story sites.

Hogwarts Legacy now runs with NVIDIA DLSS 5 and 3D-Guided Neural Rendering on GeForce NOW, adding new lighting and material detail while leaving the game's geometry and textures intact. The same monthly update brings World of Warcraft: Forever and two Ubisoft expansions to the service.

Box announced Box Content Retrieval, a feature built on Nvidia's Nemotron embedding models, at its developer conference. Nvidia's Kari Briski expects queries to run four times faster with 20% fewer errors and 90% fewer tokens, available to Enterprise Advance customers early next year.

Databricks describes pushing change data capture from operational databases into Unity Catalog using Zerobus Ingest together with Debezium. The setup streams database changes directly into Unity Catalog tables rather than relying on batch loads.

Agility Robotics became the first dedicated humanoid robot developer listed in the US, trading on Nasdaq from November 5 after a SPAC merger, and its SEC filings reveal pricing, unit economics and a $140 million operating loss on $1.8 million of 2025 revenue.