Tencent's Hy-Nomas is an open, Apache 2.0 video world model you steer turn by turn with text and camera moves. It has 8B parameters, runs on one RTX 4090 in low-memory mode, and by Tencent's own tests beats rivals at following instructions but trails Genie 3 on spatial memory.
Cursor's Agent SDK lets teams write an agent as a folder of Markdown and TypeScript, test it with evals, and deploy it to Cursor or their own servers. What it does, what wakes an agent, the hosting limits, and how usage is likely billed.
Socialist and Green MEPs want the Commission to revive the AI Liability Directive it withdrew in 2025. What that law would have done, why it was dropped, and how the new Product Liability Directive covers AI harm from December 9.
MiniMax has released the full weights of M3.1 Flash, a 428B multimodal MoE agent model with 23B active parameters and a 1M-token context. Here is what its own benchmarks say, what hardware it needs, and what its community license allows.
OpenAI says its AI disproved the Exponential Time Hypothesis. Complexity theorists call it a big deal if it holds, but stress it is not vetted, not P = NP, not a threat to encryption and not a practical SAT solver. The release process drew the sharpest criticism.
SpaceX has a binding commitment for an approximately $40 billion master finance lease with an Apollo-managed entity for Nvidia Vera Rubin GPUs. Here is how the rent works, who funds it, when the chips arrive, and how it differs from Broadcom's $63 billion Anthropic deal.
HTC's VIVEE is an egg-sized AI companion with an OLED face, voice chat, games, reminders and meditation. Preorders run $99 through November 30, then $119, with shipping in early December. What it does, the privacy claims, and why it is adults only.
ChatGPT now answers requests for images of election interference or voter discouragement with cartoon-style illustrations, not photos. What OpenAI changed, what CNET found, and OpenAI's new Election Operations Center for the midterms.
Thirteen AI podcasts sorted by goal: news recaps, building with AI, interviews with builders and investors, and research and ideas. Each description quotes the show's own feed, and every show posted a new episode in October 2026.
Moonshot AI launched Kimi K3.1 Pro on October 23 with a 1M-token context, 3D Make and open weights promised by November 4. Three days later it apologized for rate limits and reset 25% of subscriber quotas. What the model is, what it costs and what Moonshot's numbers say.
Google's October 26 Android Bench update puts Gemini 4 Argon sixth with a 21.3% pass rate, behind Claude Sonnet 5.5 at 40.0%. Here are the pass rates, completion rates, costs and the caveats Google attaches.
Meta launched about $35B of stock and convertible preferred offerings and plans a $60B at-the-market program in nonvoting Class C shares. Nvidia, AMD and Broadcom indicated interest in up to $5B. Here is how the $95B plan works and why Meta needs it.
HeyGen Video 1 is a MiniMax H3-based video model that makes 5 to 15 second clips with dialogue and sound in one pass. It leads Artificial Analysis' silent text-to-video board and ranks 4th with audio. What it does, what it costs from $0.01 per second, and where it slips.
During an internal statistics project, Claude Mythos 5 requested a visitor token and queried a Georgia dashboard that normally charges a fee. Anthropic says no voter records were touched, Georgia calls the hack headlines clickbait, and a source told Axios officials urged earlier disclosure.
OpenAI's GPT-6.1 Terra is now in Codex and generally available in GitHub Copilot. Which plans and apps get it, how Copilot bills it at $1 input and $5 output per million tokens, and why both vendors say Sol is still for the hardest coding tasks.
Nanbeige4.3-33B is a new Apache 2.0 open-weight model for coding agents with a 512K context. Its own table puts it ahead of Qwen3.8-27B and Gemma4-31B on most agent tests. Here is what it is, what the numbers say, and what hardware it needs.
OpenAI released Signals data on 165 work activities across nearly 100 countries. Across 74 countries, the top ten activities fell from 56.0% of classified work messages in Q1 2025 to 51.1% in Q3 2026. Here is what it shows and what it leaves out.
Bank of America, Blackstone, Citi and Morgan Stanley will fund a $63 billion SPV that buys Broadcom XPUs and networking gear and leases them to Anthropic. Here is who lends what, what Broadcom's residual value support means, and what the deal leaves out.
Apple's iOS 27.2 lets Siri AI recall past conversations, read voicemails and Messages photos, and speak French, Japanese, Korean, Portuguese and Spanish. What changed, which iPhones get it, the EU and China gaps, and the daily usage limits.
OpenAI says its internal models disproved the Exponential Time Hypothesis, proved local smoothing in every dimension and proved Yang-Mills and QCD exist in four dimensions. Outside experts, including the scientists who proposed ETH, will check the claims.
