AI news for builders and product teamsUpdated Nov 17, 2026, 18:00 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Google DeepMind released Genie 4 in research preview on Tuesday; it generates photorealistic interactive worlds with real-time audio, supports prompt-based scene edits mid-simulation, and holds a consistent scene for up to two minutes, twice Genie 3's limit.

VicOne LAB R7 describes tests in which manipulated inputs, including a poster read as an instruction by a robot dog running Gemma 4 E4B and crafted inaudible audio in a hospital-robot simulation, changed robot behavior without altering assigned tasks. It argues safety evidence must cover deliberately falsified sensor and message data, not just accidental faults.

Mozilla is adding AI form filling and webpage monitoring to Smart Window, Firefox's AI browsing experience, with form filling pulling details from open tabs such as a LinkedIn page to complete job applications.

Ai2's study finds LLM novelty judges are highly unstable: a small prompt wording change moved GPT-5.4's accuracy at picking the more novel of two ideas from 93.5% to 40.9%, and purpose-built evaluators underperformed a simple Claude Opus 4.6 prompt.

Apple's new iPad Mini is the first iPad with an IP67 dust and water-resistance rating, alongside an OLED display, A20 Pro chip, and a thinner, lighter body. It starts at $699, a $200 increase over the 2024 model, with preorders opening October 28 and shipping November 6.

Google DeepMind announced Genie 4, a natively multimodal real-time world model that generates photorealistic interactive simulations with sound and dialogue, now in a research preview for trusted testers. Consumer and developer access is planned to start with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio.

Google DeepMind announced Genie 4, a world model that improves on Genie 3 with more realistic physics and human motion, native audio including interactive speech, and steering from Gemini 3.8 Flash. Access will expand to developers via the Gemini API and AI Studio, but pricing hasn't been set.

NVIDIA published an open post-training workflow for adapting Nemotron models to local rules and culture, releasing locale-grounded persona datasets for France and Belgium and a French public-service example built on Nemotron 3.5 Lightning.

NVIDIA's Factory Operations Blueprint coordinates multiple AI agents to investigate production problems such as failed circuit boards or rush orders, pulling evidence from separate factory systems. A Siemens report cited by NVIDIA estimates unplanned downtime costs an average large plant in the surveyed sectors $253 million a year.

Apple has announced a new M6 iMac, starting at $1,499, with pre-orders opening October 28 and retail availability on Friday, November 6. Apple claims up to five times the AI performance of the M4 model, alongside Wi-Fi 7, Bluetooth 6 and faster SSD storage.

Surfshark's analysis of its web content blocker found a single counterfeit TikTok Shop generated more than 320,000 phishing blocks between May and September 2026, or 19% of all global phishing domain blocks in that period, and the domain remained active as of October.

C-BRAIN, a consortium of 18 institutions launched in November 2025 and led by Dr. Randall Bateman, is working with NVIDIA to apply federated AI to Alzheimer's research data, including unpublished and negative results it calls "dark data." Its open source OpenScientist tool completed analyses in minutes that scientists estimated would take weeks to months, at under $10 per run.

NVIDIA detailed three optimizations for mixture-of-experts inference on DGX Spark that raised decode throughput by 42.5% and cut mean end-to-end latency by 24.5% on GSM8K, with no measurable accuracy change. The post also warns that combining the techniques requires a specific LM-head sharing interface, or performance drops below baseline.

Rohit Patel has left Meta after eight years, most recently as a director at Meta Superintelligence Labs, to launch Ndoc, an unfunded three-person startup pitching a unified AI-native workspace as a replacement for Microsoft Office and Google Workspace. An alpha is opening to waitlisted customers, with a beta planned for the coming weeks and a full launch next year.

A Thomson Reuters principal engineer ran the open-source coding agent OpenCode against Qwen3.5 9B served locally by Ollama on a MacBook with Wi-Fi switched off, and the agent fixed a one-line bug in an expense calculator so all five existing tests passed.

Thore Graepel, a former DeepMind researcher who worked on AlphaGo, has founded Metis Reasoning to apply AlphaGo's search-and-verify approach to general reasoning via an external "knowledge ledger" rather than knowledge stored in model weights. The company has published a white paper only, with no working system, benchmarks or comparisons to current reasoning models.

Apple is bringing an AI assistant built on Apple and OpenAI models to Pages, Keynote and Numbers, letting it work from a document's context to draft text, design slides and build charts. The apps also gain Model Context Protocol support on macOS 27 or later so third-party AI agents can plug in.

GMO Internet says suspected attackers have posted data linked to customers and its business after unauthorized access to its Onamae Desktop Cloud service, which the company shut down on October 21. The company is still verifying the data and has told customers to reset passwords after reports of unauthorized operations on trading accounts.

Opera has introduced a proposed web standard called WebAI that lets its browser run AI models on the user's own device rather than in the cloud, starting as an opt-in early preview in Opera One. It is model-agnostic, with Google's Gemma cited as an example and Apple's on-device foundation model supported on Macs.

General Compute raised a $100 million Series A led by Blackbird Ventures and Abstract Ventures at an $800 million post-money valuation, betting its cloud on inference chips from Nvidia rivals including Cerebras, Positron and SambaNova.