NAll news

Friday1454 items

  1. Unite.AIModel releases
    AI score64

    Baseten Releases 2B Open-Weight Decision Model joey-1

    On October 29, Baseten released joey-1, a 2B-parameter decision model, priced at $0.01 per million tokens. It is now available on Baseten Model APIs and has open weights on Hugging Face. The model was post-trained by the in-house lab Base Labs based on Qwen3.5-2B, and shares the same data recipe and API with mammoth-1 (27B) and elephant-1 (9B), so switching requires only a one-line change. The model card shows that joey-1 scored 63.30 on the public Jev Decision Index 0.4 leaderboard, ranking 37th among 230 models and the highest score among models with fewer than 4B parameters; Baseten says its P50 end-to-end latency for 2000-token requests in production is 20 ms.

    Same story · 3 reports

    #Model releases#Open-source ecosystem#Reasoning

  2. Unite.AIOpinion
    AI score62

    Research reveals why AI agents fail silently, and how to measure such failures

    Unite.AI, reviewing multiple studies, points out that when LLM agents make errors they often still return fluent answers and success status codes, making them hard for conventional monitoring to detect. A prompt-format experiment at ICLR 2024 showed that merely changing delimiters, spaces, and capitalization caused LLaMA-2-13B's accuracy on the same task to fluctuate by as much as 76 percentage points; in τ-bench, a GPT-4o agent solved about 61% of retail tasks in a single attempt, but the proportion of 8 consecutive fully correct runs dropped to about 25%. The article recommends putting prompts and tool descriptions under version control, repeating tests multiple times, recording the tools' actual returns in traces, and validating LLM-as-a-judge scores with human annotations.

    #English text is being prepared.#Agents#Evaluations / benchmarks

  3. Unite.AIResearch
    AI score58

    Eclypsium demonstrates how compromised hardware can tamper with AI models at inference time

    Eclypsium demonstrated, with NVIDIA's support, that an attacker using a compromised DPU and lax memory isolation can modify runtime model weights during inference, causing the model's answers to change while the model files on disk and hash verification still appear normal. The experiment was conducted in an NVIDIA Grace Hopper environment and depended on conditions such as IOMMU passthrough and lax PCIe isolation, which the researchers explicitly noted are not Linux default settings. The report brings DPUs, GPUs, management controllers and their firmware into the AI security boundary, and notes that model signing, secure boot, runtime isolation and continuous monitoring each answer different questions.

    #Safety / alignment#English text is being prepared.#English text is being prepared.

  4. Unite.AI
    AI score52

    Mintive Raises $7 Million to Provide Independent Identity and Permission Management for AI Agents

    Mintive has completed a new $7 million funding round led by Cowboy Ventures, bringing total funding to $11 million. The company says it currently handles 2.5 million agent calls per day, serves more than 50 paying enterprise customers, and has processed over 200 million cumulative calls in seven months. Its platform assigns each agent an independent identity, scoped permissions, and traceable operation records; underlying service credentials remain within the gateway, agents access tools through short-lived tokens, and administrators can revoke an individual agent's token separately.

    Same story · 2 reports

    #Industry news#Agents#English text is being prepared.

  5. Unite.AIProduct release
    AI score67

    Gemini app now supports editing documents and PDFs within conversations

    Google announced that Gemini app users can now generate, preview, and edit files such as documents, PDFs, presentations, spreadsheets, and code directly within conversations, and this capability is available to all users in all regions where the app is available. Editing methods include direct full-screen input and modification, selecting text to ask Gemini questions, and having Gemini rewrite file content. PDF preview and editing are supported on both the web version and mobile, while other formats such as Google Docs, Sheets, Slides, Markdown, TXT, and code are supported only on the web version. This update adds an in-place editing layer on top of the file generation feature launched on April 29, 2026.

    Same story · 7 reports

    #Product updates#Google#Agents

  6. Unite.AI
    AI score50

    AI finance OS startup Collective raises $100 million: $40 million...

    Collective, the developer of an AI-native financial operating system for solopreneurs, announced on October 29 that it has raised $100 million in new funding, consisting of a $40 million Series B equity round and $60 million in growth financing from General Catalyst's Customer Value Fund. The Series B was led by General Catalyst, QED Investors, and Gradient Ventures, with participation from Intuit Ventures, FT Partners, and SemperVirens. Collective said its annual recurring revenue grew 120% year over year, and the platform has cumulatively saved solopreneurs $350 million in taxes.

    #Industry news#Agents

  7. Unite.AI
    AI score51

    OpenAI-backed Thrive Holdings to acquire ADT in an all-cash deal valued at $15.3 billion

    ADT and Thrive Holdings announced on October 29, 2026 that they had reached a definitive agreement under which Thrive Holdings will acquire ADT in an all-cash transaction valuing ADT's equity at approximately $7.6 billion and its enterprise value at approximately $15.3 billion, with ADT shareholders receiving $10.00 in cash per share. The price represents a 45.6% premium over ADT's unaffected closing price of $6.87 on October 27, 2026. The transaction is expected to close in the first quarter of 2027, subject to approval by ADT shareholders and U.S. antitrust review. Upon completion, ADT will become a private company and be delisted from the NYSE, and will serve as the foundation of Thrive Holdings' smart home security business; OpenAI announced on December 1, 2025 that it had taken a stake in Thrive Holdings and deployed research, product and engineering teams into its portfolio companies.

    Same story · 2 reports

    #Industry news#OpenAI

  8. Unite.AI
    AI score33

    Coveo Acquires Y Combinator-Backed Product Data Startup Anglera

    On October 29, 2026, Coveo announced the acquisition of Anglera, a San Francisco-based startup backed by Y Combinator, folding the latter's AI agent-powered product data enrichment platform into the Coveo Commerce product line. The transaction amount was not disclosed. Anglera extracts product attributes from sources such as manufacturer specifications, PDFs, and images; each enriched value comes with citations and a confidence score, uncertain values are routed to human review, and it syncs bidirectionally with PIM, ERP, and e-commerce platforms. Citing a NielsenIQ estimate, Coveo said missing product attributes can reduce product discoverability by up to 30%; in a 30-day A/B test, roughly 300 SKU variants saw a 120% increase in organic landing page traffic and a 270% increase in AI-referred visits.

    #Industry news#Agents#RAG

  9. Unite.AI
    AI score38

    Rivian Licenses AI Voice Assistant Technology Through Volkswagen Joint Venture

    Rivian disclosed in its Q3 2026 earnings report that it has licensed its AI Voice Assistant technology to the Volkswagen Group through the Rivian and VW Group Technology joint venture, as part of the two companies' electrical architecture and software partnership. Rivian's software and services revenue for the quarter was $605 million, up 45% year-over-year, of which $354 million (59%) came from the joint venture. Rivian Assistant launched on the R2 in October 2026 and added Gmail integration via an OTA update.

    Same story · 2 reports

    #Industry news#Speech#English text is being prepared.

  10. Unite.AIResearch
    AI score88

    Epoch AI report: open-source model GLM 5.3 exploits Firefox patch gap to launch attacks

    Epoch AI published research on October 29, 2026, stating that GLM 5.3, an open-weight model released by Z.ai in August 2026, can exploit browser patch gaps to compromise six recent versions of Mozilla Firefox, at a marginal cost of about $1,000 per version. In the experiment, the model, without internet access and with only public information, autonomously completed vulnerability identification and a multi-stage exploit chain, with the fastest run taking 12 hours, and repeatedly rediscovered an undisclosed vulnerability; Epoch has notified Mozilla of the vulnerability, which the latter fixed in Firefox 159.0 released on October 27, 2026. Epoch said total API spending for the Firefox effort exceeded $100,000, while attacking Google Chrome with the same toolset cost about $100,000 and still did not succeed.

    #Safety / alignment#Agents#Open-source ecosystem

  11. Unite.AI
    AI score86

    CrowdStrike says VAPOR PANDA attacked more than 200 organizations using the DeepSeek-driven ARTEX framework

    CrowdStrike's report states that the China-linked group VAPOR PANDA used an AI-driven intrusion framework called ARTEX, built primarily on the DeepSeek large language model with another commercial model as a secondary component, to target more than 200 organizations in at least 14 countries over 11 days starting October 17, 2026, and confirmed post-compromise activity at 9 of them. ARTEX consists of a main agent and planning and execution sub-agents that share a fact and goal graph; in a single intrusion the planning agent consumed about 1.4 billion input tokens, and operators issued commands in Chinese and intervened at key junctures. CrowdStrike also says the group deployed the BLOODALCHEMY backdoor at some targets, and predicts that China-linked adversaries will increasingly adopt such agent frameworks over the next 12 months.

    #Safety / alignment#Agents#DeepSeek

  12. English text is being prepared.Product release
    AI score53

    Google announces that Android XR core development tools have reached stable release, supporting Unity, Godot, and Unreal engines

    At Android Dev Summit 2026, Google announced that multiple core Android XR development tools have ended their preview phase and entered stable release, including the Jetpack XR SDK core libraries, the Android XR Emulator, and the Android XR Engine Hub. Jetpack SceneCore, ARCore for Jetpack XR, and the XR Runtime core libraries provide official product development APIs and support long-term interface stability and backward compatibility; the emulator allows development and debugging for XR glasses, audio glasses, and display-equipped glasses without physical devices. The stable release of Android XR Engine Hub supports the three major engines Unity, Godot, and Unreal, and can be used to launch performance analysis and view runtime data. On the same day, Google partnered with XREAL to launch the XREAL AURA glasses powered by Android XR.

    #Product updates#Google#Open-source ecosystem

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

    XREAL Aura Launches: The First AR Glasses Powered by Android XR, Starting at $1,099 for a Limited Time

    XREAL has officially launched the XREAL Aura, AR glasses powered by Google's Android XR operating system. The first batch of orders will ship in mid-November, with the initial launch covering the United States, Canada, Japan, and South Korea. Under the limited-time offer, the 12GB+256GB version is priced at $1,099 (regular price $1,279), and the 16GB+512GB version is priced at $1,299 (regular price $1,499). The product adopts a split architecture, with the frame weight kept under 95 grams, a 70-degree field of view, dual Sony Micro-OLED displays with a per-eye resolution of 1920×1200 and a maximum refresh rate of 120Hz, and it comes pre-installed with more than 150 XR-optimized apps.

    Same story · 4 reports

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

  14. English text is being prepared.
    AI score27

    Jagiellonian University and others propose DomeGS: using 3D Gaussian splatting to complete the background of outdoor VR scenes

    Researchers from institutions including Jagiellonian University in Poland and the University of Cambridge in the UK published a paper on arXiv, proposing the DomeGS framework, which combines generative background completion with 3D Gaussian splatting (3DGS) scene reconstruction to address the missing sky, horizon, and distant views in outdoor VR scenes. 36 participants used Meta Quest Pro to test four groups of outdoor scenes. DomeGS achieved statistically significant improvements in spatial presence and perceived realism (p=0.042), and 75% of participants were more willing to stay for extended periods in virtual environments using DomeGS. The two approaches showed no significant differences in task completion time, head movement distance, or VR sickness symptom scores.

    #Papers/Research#Multimodal#arXiv

  15. English text is being prepared.
    AI score49

    In 2026, full-color AI glasses are launching in clusters; RayNeo preheats the full-color Micro-LED optical waveguide iO+

    In 2026, full-color AI glasses are making intensive appearances. RayNeo is preheating its fourth-generation binocular full-color AI glasses, the iO+, whose full-color light engine is smaller than 0.18cc, already approaching the level of monochrome Micro-LED optical engines. In the same period, Rokid released the full-color AI glasses Rokid Color, adopting a Micro-LED and diffractive optical waveguide solution, priced from 4,999 yuan. Since the X1 in 2021, RayNeo has continuously invested in full-color Micro-LED optical waveguides; in 2023, the X2 became the world's first consumer-grade mass-produced binocular full-color Micro-LED optical waveguide AR glasses.

    #English text is being prepared.#Multimodal#English text is being prepared.

  16. English text is being prepared.
    AI score46

    Tokyo University of Science completes bladder cancer assisted surgery using XREAL One Pro AR glasses

    Tokyo University of Science successfully completed a bladder cancer surgery assisted by XREAL One Pro AR glasses; the patient was a man in his 50s, and postoperative progress was smooth. The system was jointly developed by the university's Department of Renal Urology and the Department of Electrical and Electronic Engineering team, superimposing endoscopic images within the doctor's field of view and connecting to Sony's NUCLeUS platform to centrally manage operating room video signals. Other participants can wear xbx a01+ to view information synchronously.

    #Papers/Research#Multimodal

  17. English text is being prepared.
    AI score42

    Danish company NIL Technology releases the ultra-compact eye-tracking camera metaEye, featuring a Meta-optics design

    Danish optics company NIL Technology has released the ultra-compact eye-tracking camera metaEye, which it calls the world's first camera designed entirely with meta-optics, with overall dimensions smaller than a grain of rice. The product integrates two metalens optical elements (MOE), a bandpass filter, and multiple optical apertures, and is mounted directly on the image sensor's cover glass, eliminating the need for traditional refractive lenses. NILT says relevant customers are moving toward mass production in 2026, and metaEye is also the first product in its metaCamera modular camera architecture.

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

  18. English text is being prepared.
    AI score18

    "Fallout: Factions" XR Version to Launch on Xreal Aura Next Month, Gesture Controls Recreate Tabletop Battles

    The XR version of "Fallout: Factions" will launch first on Xreal Aura next month. The game is adapted from the physical 1v1 tabletop game released in 2024. The XR version retains the Fallout worldview and focuses on turn-based skirmish battles, featuring an immersive 3D battlefield map and gesture interaction, allowing players to directly roll dice, move units, and launch attacks with gestures. The game was originally scheduled to launch alongside the first batch of Xreal Aura devices, and will later come to the Android XR platform as well, with a Steam version also coming soon.

    #Industry news#Multimodal

  19. English text is being prepared.
    AI score28

    South Korea's Grep Launches AI Glasses Verification Feature for Exams, Adopted by 21 Companies and Institutions Within Two Months

    South Korean AI assessment solution company Grep announced that the smart glasses verification feature of its online exam proctoring platform Monito has been adopted by 21 companies and institutions within two months of launch. The feature, launched in August this year, requires examinees to photograph four parts of their glasses through the Monito app before the exam—the front, side, inner side, and the end of the temple—with AI-assisted image capture and manual comparison and judgment by proctors to reduce the risk of misjudgment. Grep plans to add a VLM-based real-time behavior analysis feature in November this year to identify abnormal behaviors such as switching glasses or repeatedly touching glasses.

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

  20. English text is being prepared.
    AI score45

    Raven Prism Full-Color AR Glasses Now Open for Pre-Order: Runs Linux Standalone, Supports Eye-Tracking Interaction

    Raven Resonance's Raven Prism full-color AR glasses are now open for pre-order, with a fully refundable $49 deposit; the first batch is expected to begin shipping in January 2027. The device runs standalone on the Linux-based RavenOS, features a full-color LCoS waveguide display and a quad-core 64-bit ARM processor, centers on eye-tracking and voice interaction, and adopts the Raven Wings modular hot-swappable battery design. The initial app store offers more than 25 apps, targeting developers, creative professionals, and enterprise users.

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