NAll news

All news

Tuesday61 items

  1. Intelligent Manufacturing NetworkProduct release
    AI score56

    Galbot and Unitree Robotics Announce Joint Development of Humanoid Robot Galbot Star Kid

    Galbot Robotics and Unitree Robotics announced the joint development of the native agent robot Galbot Star Kid (Galbot ET1) at the 2026 World Physical AI Conference and Unitree Partner Ecosystem Conference. Galbot is responsible for product definition, multimodal interaction, and full-body motion control, while Unitree provides joint modules and hardware mass-production support. Galbot Star Kid relies on Galbot's self-developed AstraBrain-WBC general-purpose cerebellum foundation model and million-hour-level human motion data, enabling it to visually observe human movements in real time and learn and reproduce new movements on the spot. The product has announced three price tiers: a 79,000 yuan base version, a 139,000 yuan professional version, and a 179,000 yuan flagship version. Within 24 hours of opening pre-orders, pre-order volume exceeded 200 units.

    #Industry news#Embodied AI#Product updates

  2. Phoenix TechnologyProduct release
    AI score74

    AgiBot Releases Embodied Agent Framework Genie Master, Enabling Robots to Evolve While Working in Real Convenience Stores

    AgiBot has released its embodied Agent framework Genie Master. The Genie G2 equipped with this framework, while working in a real convenience store, does not rely on human demonstration or debugging; it independently identifies the causes of failures, adjusts strategies, and accumulates experience, continuously improving on multiple complex and delicate tasks and achieving continuous self-evolution with no human in the loop. The original article states that this is the industry's first validation of Physical RSI (embodied recursive self-evolution) outside of simulation environments and in real business scenarios. Yesterday, AgiBot also released the safety-certified heavy-duty embodied robot Genie G2 Max, targeting industrial scenarios such as depalletizing and palletizing, material transfer, and production line feeding.

    Same story · 5 reports

    #Product updates#Embodied AI#Agents

  3. Securities TimesProduct release
    AI score71

    AgiBot Releases Embodied Agent Framework Genie Master

    AgiBot has officially released its embodied Agent framework Genie Master. During its deployment in a real convenience store, the Genie G2 equipped with this framework does not rely on manual demonstrations or debugging; it can independently identify the causes of failures, adjust strategies, and accumulate experience, continuously improving on multiple complex and fine-grained tasks to achieve continuous self-evolution with no human in the loop. The original article states that this is the industry's first validation of Physical RSI (embodied recursive self-improvement) outside of simulation environments and in real business scenarios.

    Same story · 5 reports

    #Product updates#Agents#Embodied AI

  4. LeitechProduct release
    AI score58

    AgiBot Releases Embodied Agent Framework Genie Master, Validating Physical RSI in a Convenience Store Scenario

    AgiBot has released the embodied Agent framework Genie Master, enabling robots to autonomously identify the causes of failures after deployment, adjust strategies, and accumulate experience. The framework proposes the Embodied Execution Protocol (EEP), which unifies four types of interfaces—capability, guidance, feedback, and continuity—and, through the two routes of Harness RSI and Policy RSI, retains experience in prompts and memory, and in action policy parameters, respectively. The Genie G2 equipped with this framework has already taken up real posts in a convenience store, achieving continuous improvement in tasks such as hot food packaging, picking, and restocking; AgiBot claims this is the first validation of Physical RSI in a real business scenario.

    Same story · 5 reports

    #Product updates#Agents#Embodied AI

  5. Electronic Technology ApplicationProduct release
    AI score60

    Empyrean Technology Releases .aix Series AI Agent Platform, Covering Analog, Panel, Digital and Manufacturing Segments

    At the HUG 2026 Ecosystem Partner and User Conference, Empyrean Technology released the .aix series AI agent platform, built on domestic open-source large models, and providing a Copilot agent for each EDA tool. Custom.aix compresses analog circuit parameter tuning and layout drawing from weeks or months to within 1 day, reduces design iterations by 60%, and improves PPA by more than 5 times; FPD.aix compresses a week's worth of panel pixel design work to within 1 day, improving efficiency by more than 7 times. Digital.aix generated more than 200 standard cells in an open-source CPU case, of which 118 brought positive returns, with the maximum frequency increased from 1.81GHz to 1.83GHz; Manufacture.aix compresses long-range Pcell tasks from several days to within half an hour, and increases Dummy Fill space utilization from 78% to 93%. On the tool foundation side, ALPS VX achieves 3x simulation acceleration, Patron VX improves efficiency by 2x, and Liberal VX achieves 3x acceleration in K library characterization.

    Same story · 2 reports

    #Product updates#Agents#Open-source ecosystem

  6. Shao MengProduct release
    AI score74

    MiniMax Open-Sources Self-Hosted OpenAI Agents API "OpenAgentCore"

    MiniMax has open-sourced a self-hosted OpenAI Agents API called "OpenAgentCore". With the official OpenAI SDK or plain HTTP, you only need to point the address to the /v1 endpoint of your self-built Core to make calls. Each Session can optionally use a harness (such as codex), natively running Codex, Claude Code, and MiniMax Code within the session, with models going through a self-configured provider. The Sandbox Provider, workspace filesystem, Harness, and model provider are all replaceable according to established protocols, supporting microsandbox, E2B, Docker, and local Linux/macOS/Windows.

    Quoted postSaladDay · @saladday

    English text is being prepared.

    #English text is being prepared.#Agents#OpenAI

  7. English text is being prepared.Product release
    AI score56

    AgiBot Releases Genie G2 Max Heavy-Duty Embodied Robot, Dual-Arm Rated Payload of 40 kg

    AgiBot has released the safety-certified heavy-duty embodied robot Genie G2 Max, featuring a single-arm payload of 20 kg, a dual-arm rated payload of 40 kg, a peak of up to 50 kg, and repeat positioning accuracy of no more than ±0.5 mm. The product targets industrial depalletizing and palletizing, material transfer, and production line feeding. It is equipped with 7-DOF force-controlled dual arms combined with visual guidance, and has a vertical working range of 0.1 m to 2.5 m. AgiBot says it is already operating on multiple production lines, with a task success rate exceeding 99% in e-commerce warehousing scenarios, completing 80 boxes of depalletizing and 75 boxes of palletizing per hour.

    Same story · 7 reports

    #Product updates#Embodied AI

  8. Intelligent Manufacturing NetworkProduct release
    AI score58

    Agibot Releases Safety-Standard Heavy-Load Embodied Robot Genie G2 Max

    Agibot has released the safety-standard heavy-load embodied robot Genie G2 Max, featuring a single-arm payload of 20 kg, dual-arm rated payload of 40 kg, peak payload of 50 kg, and arm repeat positioning accuracy of ≤±0.5 mm. The product targets scenarios such as depalletizing and palletizing, material transfer, and production line feeding, with a vertical working range of 0.1 m to 2.5 m. It supports single-side depalletizing and palletizing and allows pallets to be placed against walls without reserving 1.5 meters of clearance space. The original text states that it has already been deployed in leading daily chemical FMCG production lines, e-commerce warehousing, and the Qiyuan T1 production line. In the e-commerce warehousing scenario, the operation success rate exceeds 99%, with 80 boxes depalletized and 75 boxes palletized per hour.

    Same story · 7 reports

    #Product updates#Embodied AI#English text is being prepared.

  9. Unite.AIProduct release
    AI score60

    IBM Launches Governed Identity and Credential Control Capabilities for AI Agents

    At TechXchange 2026, IBM announced identity, credential, discovery, isolation, and evaluation capabilities for enterprise AI agents, with IBM Agent Identity and HashiCorp Vault 2.2 now generally available. Agent Identity registers agents in a unified inventory, records ownership, risk classification, and lifecycle status, and issues OAuth-based short-lived, task-scoped tokens, working with IBM Verify and Vault to achieve least privilege and real-time revocation. watsonx Orchestrate adds AI Gateway as a unified execution layer that can discover third-party agents, models, and MCP servers from platforms such as Amazon Bedrock, Microsoft Foundry, and Google Gemini, and isolation sandboxes and evaluation tools have also been launched.

    #Product updates#Agents#Safety / alignment

  10. LeitechProduct release
    AI score53

    Xingyuan Intelligence Releases Cross-Embodiment Data Engine Chorus, Claiming Approximately 40x Training Data Expansion

    Xingyuan Intelligence has released Chorus Zhongxing, a cross-embodiment data engine for robot pretraining, which can convert human manipulation videos captured by smart glasses, chest-worn cameras, wrist cameras, UMI devices, smart headbands, and other devices into pretraining data adapted to different robot embodiments. Tests show that Chorus, through adaptation to 16 robot embodiments, reduces the processing cost per hour of effective robot trajectories to 15 yuan, raises the average usable data conversion rate to 85%, and, combined with 3x instruction augmentation, can achieve approximately 40x effective training data expansion, with an equivalent cost per unit of training samples of 5 yuan/hour.

    Same story · 2 reports

    #Product updates#Embodied AI#English text is being prepared.

  11. LeiphoneProduct release
    AI score75

    Ant Group Releases Lingguang 2.0, AI Assistant Shifts to Proactive Task Management

    On October 27, Ant Group released Lingguang 2.0, advancing the AI assistant from passively responding to instructions and completing single tasks to proactively identifying needs and continuously managing affairs. Lingguang 2.0 upgrades cross-session memory, long-horizon task scheduling, and tool connection capabilities. Relying on an independent cloud host, it continues running tasks after the user exits the conversation, and connects to services such as Alipay bills, travel, DingTalk, Feishu, email, and Mijia, supporting calls to tools such as MCP and CLI. Lingguang 2.0 also integrates with Ant's APASS unified account and security management infrastructure, allowing users to view, adjust, or revoke authorization at any time.

    Same story · 5 reports

    #Product updates#Agents#English text is being prepared.

  12. English text is being prepared.Product release
    AI score58

    Mac version of Ollama adds offline text-to-speech, up to 35% faster than the original solution

    Ollama announced on October 24 that its Mac app officially supports text-to-speech generation, with Windows and Linux versions to follow. The first supported speech model is the Phonon series developed by Gradium, which users can run offline on Apple Silicon Macs without cloud services. Official tests show that on a MacBook Pro with the M5 Max chip, Phonon, while achieving a word error rate comparable to the Hugging Face Transformers reference implementation, generates speech 35% faster than Gradium's own implementation.

    Same story · 2 reports

    #Product updates#Speech#English text is being prepared.

  13. English text is being prepared.Product release
    AI score70

    Ant Group's Lingguang 2.0 Released: Upgraded to an AI Assistant That Can Proactively Manage Affairs

    Ant Group has released Lingguang 2.0, upgrading the AI assistant from passively responding to instructions and completing single tasks to proactively identifying needs and continuously managing affairs. Lingguang 2.0 has upgraded cross-session memory, long-horizon task scheduling, and tool connection capabilities, relying on an independent cloud host to continue running tasks after the user exits the conversation. After the user authorizes integration with Alipay, it can organize bills and discover auto-renewal subscriptions, and can also follow DingTalk class group notifications and extract to-dos; in addition, it supports calling tools such as MCP and CLI, and integrates with Ant APASS to manage authorization and privacy.

    Same story · 5 reports

    #Product updates#Agents#English text is being prepared.

  14. Sina TechProduct release
    AI score59

    Ant Group releases Lingguang 2.0, upgraded to an AI assistant that can proactively manage tasks

    Ant Group releases Lingguang 2.0. With user authorization, it can combine integrated information and services to proactively identify potential needs, initiate analysis, monitoring and follow-up, and remind users at nodes that require attention. Lingguang 2.0 connects data and services within the Alipay ecosystem, and also supports calling tools such as MCP and CLI in an open environment. It has built-in service capabilities including Alipay bills, membership, travel, as well as DingTalk, Feishu, email and Mijia. The Lingguang App is Ant Group's omni-modal general-purpose AI assistant. It launched on November 18, 2025, and surpassed 2 million downloads within 6 days of launch.

    Same story · 5 reports

    #Product updates#Agents#English text is being prepared.

  15. 9to5MacProduct release
    AI score60

    Ollama Mac Version Supports Local Text-to-Speech Model Phonon

    Ollama announced that its Mac app now supports text-to-speech generation. The first integrated model is the Phonon series developed by Gradium, with Windows and Linux support coming soon. Speech can be generated in real time, so users can listen while the model outputs its reply. Ollama says that with Apple's MLX framework, speech generation was tested on an M5 Max MacBook Pro at a speed 35% faster than Gradium's own implementation, with word error rate on par with the reference implementation in Hugging Face Transformers. Phonon supports five languages: English, French, German, Spanish, and Portuguese. Each language has a separate model, plus a multilingual model, and it offers multiple voices with different accents, including American and British English.

    Same story · 2 reports

    #Product updates#Speech#English text is being prepared.

  16. MacRumors Apple IntelligenceProduct release
    AI score58

    Apple Adds SynthID Watermarks to iPhone AI Images in iOS 27.2

    In iOS 27.2, images generated by Image Playground and some photos edited via Apple Intelligence will embed Google's SynthID watermark. SynthID marks AI-generated content at the pixel level without affecting image quality, and is designed to withstand cropping, filters, and compression. Hands-on testing shows that images edited with Extend and Reframe carry watermarks, while small-scale Clean Up edits do not add a watermark and only larger edits do; there is no detection entry point on iPhone, so users must check via the SynthID Detector website.

    Same story · 14 reports

    #Product updates#Image generation#Google

  17. CNETProduct release
    AI score62

    OpenAI adds election safeguards to ChatGPT: politically sensitive images come out only in cartoon style

    In its October 20 update, OpenAI added new restrictions to ChatGPT: image requests involving election interference or that could discourage voting will only generate illustrations or cartoon-style images, and will no longer output photo-realistic images. In CNET's hands-on test, requests to generate an image of ICE law enforcement officers standing outside a polling station produced cartoon-style results every time, and ChatGPT could not be induced to produce a news photo of a voting machine recording the wrong option. Josh Lawson, OpenAI's head of election affairs, said the goal is to reduce the credibility of images as evidence while preserving room for illustration and commentary.

    #Safety / alignment#Image generation#OpenAI

  19. Baseten Engineering BlogFirst-partyProduct release
    AI score53

    Baseten Releases BDN Sealed Volumes: Model Data Can Be Published Once and Mounted Anywhere

    Baseten has released BDN Sealed Volumes, which publishes read-only data once and seals it as an immutable version with a digest, mountable in any deployment with mounting taking only seconds. This feature is available with Baseten CLI 1.1.0 and above, supports syncing from Hugging Face, S3, GCS, Azure Blob, Cloudflare R2, and CoreWeave buckets, and can also upload local directories with baseten volume push. Because mounting replaces a full download before startup, startup time no longer grows with model size, and uploads only transfer chunks that have changed.

    BDN Sealed Volumes: publish once, mount anywhere

    #Product updates#English text is being prepared.#Open-source ecosystem

  20. MiniMaxFirst-partyProduct release
    AI score61

    MiniMax M3.1 Flash Now Runs Locally via Unsloth Dynamic GGUF

    MiniMax announced that M3.1 Flash has joined the Unsloth community; after Unsloth dynamic GGUF quantization, its 428B MoE model's size drops to a 143.9GB 2-bit footprint. Unsloth's evaluation shows that this quantized version maintains a 91.5% top-1 token agreement rate across seven domains, can be deployed via Unsloth Desktop and llama.cpp, and offers an OpenAI-compatible API. MiniMax also noted that Unsloth has 78K+ stars on GitHub and over 500M model downloads on Hugging Face.

    Quoted postUnsloth AI · @UnslothAI

    MiniMax-M3.1-Flash can now run locally! We made 2-bit GGUF retain 91.5% top 1 accuracy while being 45% smaller. Run on 150GB RAM. MiniMax-M3.1-Flash is the strongest open model for its size & among the best overall. GGUF: https://t.co/A8Xp9uqpq1 Guide: https://t.co/gW5gp25niu https://t.co/txl6ZPRTjs

    Same story · 2 reports

    #Product updates#Open-source ecosystem#English text is being prepared.