OpenAI’s 23 October engineering report explains how Habitat balances shared Azure Cosmos DB capacity with workload isolation. The company reports that selected replication, nearby caching and a Rust service rewrite together reduced global median read latency from roughly 120 to 10 milliseconds.
A joint Microsoft and Hugging Face post released ThinkingBox-Training, a stack that turns executable checks of an agent’s final database state into reinforcement-learning rewards. The authors report that after 50 GRPO steps on Qwen3.8-27B, average success across 507 benchmark tasks increased from 51.70% to 60.78%. They also report …
Ai2 announced OlmoEarth v1.3 on 22 October 2026, alongside precomputed satellite embeddings, an Embeddings Explorer and a technical report. The model combines annual Sentinel-1, Sentinel-2 and Landsat time series into 128-dimensional vectors for each 10-meter pixel. Launch coverage is global for 2023–2025 and limited to selected countries for …
AWS announced general availability of Continuum for code vulnerabilities on October 22, 2026, in US East (N. Virginia) and US West (Oregon). AWS describes it as an AI security service that identifies and validates code vulnerabilities, prioritizes them using repository and optional AWS resource or network context, and …
OpenAI announced general availability of the Decisions API on 22 October 2026. The API uses GPT-6 Luna for bounded questions and returns predicate or probability outputs, a choice from predefined options, or a rubric score. OpenAI says requests can now include up to 400 questions, alongside new prompt …
Kalshi requested permission to re-list contracts tied to future GPU rental costs in comments submitted to the U.S. Commodity Futures Trading Commission. Chief compliance officer Sudhir Jain described the submission publicly on Wednesday, October 21, ahead of Axios reporting on October 22. The CFTC's formal compute-derivatives comment deadline …
Anthropic announced on 21 October 2026 that Harvard physicist Christopher Stubbs had joined its Long-Term Benefit Trust. The company said he is the Trust’s first physicist and joins Neil Buddy Shah, Richard Fontaine and Ben Bernanke. Anthropic describes the Trust as part of its governance structure, with authority …
Google DeepMind and Google Research, with the World Resources Institute and IIASA, released Forest Typology (ForTy) 2020. It uses a trained AI model and AlphaEarth Foundations satellite embeddings to classify land into six forest and land-use types at 10-meter resolution. An early annual time series covering 2017–2025 tracks …
OpenAI says it is aligning safeguards with customers and distribution partners that provide its models. Its requirements cover direct access, hosting partners and resellers, and the company says services offering direct model access must be reviewed and approved under updated Usage Policies taking effect November 20, 2026.
GitHub’s 21 October engineering report describes an Azure Git-storage migration for rising traffic from developers and AI coding agents. These agents implement and verify software changes, adding work to shared repositories. GitHub says a copy of 99.99999% of its repositories has moved to Azure; proposed per-reference quorums and …
NVIDIA’s SimReady Central Library supplies simulation-ready OpenUSD assets for training and evaluating physical AI systems and building AI factory digital twins. Its current Hugging Face post describes more than 2,000 assets, prepared with properties such as mass, collision geometry and friction, and workflows for composing them into robot …
Google Cloud and NVIDIA announced A5X cloud instances integrating NVIDIA Vera Rubin NVL72 GPUs, ConnectX-9 SuperNICs, Google Cloud’s Virgo Network and liquid cooling. The companies say general availability is planned for November 2026, with expansion to Europe planned. Google Cloud also says it intends to scale Vera Rubin …
IBM announced three updates to its Granite Speech collection: the new granite-speech-5.0-2b, a revised granite-speech-5.0-470m-turboctc, and updated weights for granite-speech-4.1-2b-plus. The releases target multilingual transcription and translation, faster transcription, and speaker attribution, respectively. Hugging Face’s announcement describes the models’ capabilities and reported benchmark improvements; the supplied evidence does …
OpenAI and Google Cloud announced that OpenAI frontier models and Codex workflows will be offered in Google Cloud Model Garden and purchased through Google Cloud Marketplace. The integration is planned to use Google Cloud's access controls, centralized data governance and infrastructure for enterprise workloads. General availability is expected …
OpenAI says ChatGPT Ads is now available in 26 additional countries, bringing its total reach to 89 countries. The company reports that more than 100,000 advertisers used the platform in its first eight months. Ads are limited to Free and Go plans; OpenAI says paid Plus, Pro and …
Google announced a Gemini business landing page on 20 October for branding, financial analysis and customer insights. The hub, marketing and finance skills, and new app connections are beginning a global rollout over the coming days.
Ai2 introduced FairStress on 20 October to examine whether language models defend comparable decisions equally when demographic identities change. The downloadable benchmark tests matched persuasive challenges, explanations and selected internal activations alongside final choices, with integration into Olmo-eval underway.
Axios reports an $80 million strategic equity investment in Oro Mobility, corroborated by Hertz’s 20 October announcement. The standalone platform will expand fleet operations for human-driven and autonomous mobility while Hertz retains majority economic ownership.
Axios’ 20 October report describes Goldman Sachs’ warning that electricity infrastructure may limit AI expansion. The bank’s outlook concerns both financing and physical delivery, not a claim that projected data centers are already operating.
Hugging Face announced AnyLanguageModel 1.0 on 20 October alongside swift-huggingface and swift-xet 1.0. Its Foundation Models-compatible interface lets Swift applications reuse session code across Apple system models, local open-weight models and hosted services.
Hugging Face’s PyTorch contributors describe Apple Silicon MPS improvements in the 2.15 release candidate for running and training language, image and video AI models locally on Macs. The 20 October post covers FP8, targeted Metal kernels and correctness fixes; it is not a final general-availability announcement.
The European Commission’s 20 October summary presents a study of online financial scams, including generative AI and the platform systems that help scams spread. Commissioned in 2025, the work supports Digital Services Act supervision rather than judging individual providers’ compliance.
OpenAI’s 19 October report introduces two evaluations of AI coding models: MultiLangBench and a separate COBOL migration test. They assess building software across less-represented languages and preserving existing behavior, rather than merely producing code that compiles.
Hugging Face’s 19 October post describes native PyTorch support for AWS Trainium, with TorchNeuron’s public beta announced for 20 October. Familiar model-library interfaces replace much of the earlier Optimum Neuron wrapper approach.
Meta’s 19 October engineering report describes a deterministic router that delegates data-pipeline work to isolated specialist AI agents using large language models for domain reasoning. The system handles data-column onboarding across three repositories and three languages.
Microsoft’s 20 October update sets out plans for local model inference and agentic retrieval in customer-controlled environments, alongside key and operator-access controls. Foundry Local general availability is planned for Ignite; other capabilities are already available, previews or regional rollouts.
Google's Gemini desktop app, launched on 10 September, brings an Alt+Space assistant and dedicated workspace to Windows 10 and 11. It puts drafting and creative tools alongside existing desktop work.
Google's 1 October release adds conversational camera descriptions and spoken framing cues to Gemini Live. Guided Vision was developed with blind and low-vision users to help interpret everyday visual information.
Anthropic opened its Singapore office and signed a memorandum of understanding with the Ministry of Digital Development and Information on 14 October. The agreement sets out cooperation on model evaluation, security, research and responsible adoption, extending earlier work with Singapore institutions.
Anthropic's updated rules govern use of Claude AI models, including deceptive campaigns, high-risk recommendations and AI-connected hardware that can act physically. Announced on 8 October, they take effect on 12 November. Many changes clarify existing rules; the announcement is not a new model launch.