LesenForschungRadarAnlageframework
Anmelden / Registrieren
AlleTechnikProdukteInvestierenBörseSonstiges

Lesen

2026-11-0499 Beiträge

Aktualisieren

Snowflake Launches Always-On CoCo and CoWork AI Agents, Reports Over 9,100 CoCo Accounts

Wesentlich
Verifiziert 2026-11-04 05:17 GMT+8

US cloud data platform company Snowflake (NYSE: SNOW) announced a series of new capabilities for enterprise AI agents at its annual Expedition 2026 conference on November 3, 2026. The updates aim to transform AI from passive question-answering assistants into proactive "digital coworkers." The centerpiece is the introduction of CoCo and CoWork Bots, always-on agents that can continuously work in the background, integrate directly with existing tools like Slack and Excel, and only surface when a human decision or sign-off is required.

Beyond the new Bots, Snowflake released CoCo Automations and CoWork Automations, allowing data and business teams to schedule recurring workflows as automated tasks. The system maintains full audit trails, ensuring all actions occur within existing permission and governance frameworks. Additionally, by introducing Memory and Skills sharing, the agents can retain past context and team best practices, improving their performance over time.

Regarding commercial traction, Snowflake disclosed specific adoption data for its AI agent products. CoWork is currently used by more than 5,800 enterprise accounts, while CoCo, the coding agent for data teams, has surpassed 9,100 accounts. This indicates that despite ongoing industry exploration of AI deployment, Snowflake has successfully embedded its agentic products into the core data workflows of numerous global enterprises.

It is worth noting that several of the newly announced features remain in various stages of preview. For instance, CoCo and CoWork Bots are marked as "private preview soon," while Profiles, Mobile apps, and the AI Data Package are in public preview. Only a few features, such as CoCo Automations and the Async Agent API, are already generally available.

Quellen:Morningstar, Inc.

Forschung

Databricks Announces Native OpenTelemetry Observability Support for Claude Platform in MLflow

Wesentlich
2026-11-04 05:00 GMT+8

On November 3, 2026, Databricks announced that its machine learning lifecycle management tool, MLflow, now natively supports OpenTelemetry observability data from the Claude Platform. This feature, currently in Public Beta, enables developers using Anthropic SDKs, the Messages API, and Claude Managed Agents to export their agent execution traces directly into the Databricks environment.

Through this native integration, agent execution steps—such as model requests and tool calls—are recorded as OpenTelemetry spans and stored in Unity Catalog tables via Zerobus Ingest technology with low latency. This allows developers to view and analyze Claude agent traces alongside those from other instrumented services within the same MLflow experiment interface, eliminating the need for additional conversion layers or middleware.

The core value of this integration lies in simplifying the debugging and evaluation workflows for AI agents. Developers can use MLflow's built-in Genie Code analysis tools to inspect the execution path behind specific responses, identifying missing context or failed tool calls. Additionally, the MLflow Issues feature can automatically detect recurring problems across sessions, such as token-wasting repeated searches or tool failures caused by missing arguments.

For enterprise users, storing sensitive agent trace data in tables governed by Unity Catalog permissions helps meet data governance and security compliance requirements. Developers can query these traces using SQL and combine them with MLflow Agent Evaluation to quantitatively assess changes to prompts, tools, or skills, ensuring that improvements genuinely enhance agent performance without introducing regressions.

Quellen:Databricks

Forschung

STMicroelectronics Unveils STM32A Series and H7P Chips for Automotive and Robotics Edge AI

Wesentlich
2026-11-04 04:14 GMT+8

Global semiconductor manufacturer STMicroelectronics (NYSE: STM) announced a series of new products and technology strategies at its STM32 Summit in Silicon Valley on November 3, 2026, focusing on pushing artificial intelligence capabilities into "physical AI" domains such as automotive, industrial automation, and robotics. The company introduced the new STM32A microcontroller series, designed specifically for automotive and high-safety applications, spanning from entry-level vehicle nodes to ASIL D-capable zonal endpoints.

In addition to the STM32A series, STMicroelectronics launched two new chips: the STM32G5 and the STM32H7P. The STM32H7P integrates a Cortex-M7 processor, a DSP accelerator, and an 82-GOPS Neural-ART neural accelerator, operating at up to 520 MHz. The company claims its CoreMark score is 150% higher than most competing solutions, targeting digital power, signal processing, and edge AI applications. The STM32G5, based on an Arm Cortex-M33 core, emphasizes high-performance peripherals and post-quantum cryptography support, suitable for motor control and digital power scenarios.

The launch also accompanies an expansion of the ecosystem. STMicroelectronics announced a deeper partnership with Arm, becoming a lead partner in defining the future Cortex-M roadmap; it also collaborated with NVIDIA, planning to integrate Holoscan Sensor Bridge software into the STM32Cube ecosystem to simplify sensor and actuator connectivity over Ethernet. Furthermore, the company updated its AI development tools, introducing Model Explorer, Model Factory, and Model Bench, and released STM32Cube for Agentic AI, designed to help coding assistants utilize STM32-specific development context, reportedly reducing token usage by approximately 50% to 60%.

This move reflects the semiconductor industry's shift from general-purpose computing to specialized edge intelligence. As autonomous driving and smart manufacturing demand higher real-time data processing and security, microcontrollers that combine high compute power with strict safety standards (such as ASIL B/D) are becoming critical infrastructure. STMicroelectronics stated that its unified STM32 platform aims to help developers reuse software and IP across product families to meet the growing interconnectivity and autonomy requirements of systems.

Quellen:MarketBeat

Forschung

OpenAI says GPT-6 Astra designed a hardware device from scratch in under a week, but required revisions for manufacturing

Thema · GPT-6 AstraWesentlich
2026-11-04 04:15 GMT+8

According to The Decoder, OpenAI stated that its hardware team used GPT-6 Astra to design a device allowing users to interact with the Codex coding agent, completing the task from scratch in under a week. The model handled every step, including enclosure design, PCB layout with component selection and trace routing, firmware, and the user interface.

Despite the speed of the design phase, a contract manufacturer took several weeks to produce a small batch of prototypes, requiring multiple revision rounds because Astra did not consistently account for manufacturing constraints, tolerances, and assembly details. This case illustrates that while frontier large models can significantly compress conceptual design and initial engineering time, human oversight regarding process feasibility remains essential when translating designs into manufacturable physical products.

Additionally, OpenAI noted that a newer "internal model" scored 0.875 on the BenchCAD 2.0 benchmark, significantly higher than GPT-6 Astra's 0.594 at its highest reasoning level. This benchmark tests the ability to reconstruct 3D parts and identify connections on PCBs. OpenAI has not shared details about this internal model.

Quellen:The Decoder

Forschung

Google Dismantles XCPE Botnet on 1.3 Million Android TV Boxes, Accuses OEM Videostrong of Intentional Malware Installation

Thema · VideostrongWesentlich
2026-11-03 18:47 GMT+8

On November 3, 2026, US technology company Google disclosed that it had worked with partners to dismantle a massive residential proxy network known as XCPE. The network consisted of more than 1.3 million Android TV streaming boxes used for ad fraud, dropping additional malware, and providing illegal residential IP proxy access to cybercriminals.

Unlike typical supply chain attacks, Google stated that this threat originated from intentional actions by the device manufacturer. The company involved, Videostrong, is a Google-certified Original Equipment Manufacturer (OEM) and Original Design Manufacturer (ODM). Researchers noted that the company not only knew about the malware but actively developed it, using obfuscation techniques to bypass Google's security tests and load malicious code after the device's first boot. While selling ostensibly legitimate Android smartphones and similar devices, Videostrong secretly built a botnet via the pre-installed XCPE backdoor and sold access to this infrastructure on the underground market.

In this operation, Google collaborated with agencies and companies including Japan’s National Police Agency, Lumen, HUMAN Security, Plume, Synthient, and TrendAI. Google seized the domains and servers used to control the botnet and install malware modules, disrupting command-and-control infrastructure for over a dozen malware families loaded via the XCPE backdoor. Additionally, Google has banned Videostrong and all future subsidiaries or affiliates from building and selling Android devices or supplying software to other certified vendors.

To address the vulnerabilities exposed by this incident, Google announced improvements to its firmware integrity checks for certified Android TV devices to prevent similar issues with other manufacturers. Google emphasized that while OEMs are required to pass comprehensive compatibility and security tests, Videostrong exploited gaps in these tests to inject malicious code post-manufacturing. This case highlights a critical trust crisis in the hardware supply chain, demonstrating that certified vendors can directly construct malicious infrastructure.

Quellen:TechRadar

Forschung

Meta Preps First-Ever Euro Bond Offering of Up to $15 Billion for AI Buildout

Thema · Meta欧元债券融资Wesentlich
2026-11-03 21:08 GMT+8

According to Semafor, citing people familiar with the matter, US social media giant Meta Platforms is preparing a large bond offering that includes its first-ever euro borrowing, sized at up to $15 billion. The move aims to raise funds for its artificial intelligence infrastructure buildout.

Meanwhile, Alphabet has also closed its second euro deal of the year. These tech giants, which have previously relied on US markets, are now turning to other pools of capital. For most companies, broadening lender bases is a test-the-pipes exercise, but for hyperscalers today, this globe-trotting fundraising is crucial to funding the AI buildout that props up their stock prices.

Quellen:semafor.com

Forschung

Ferrari Raises Full-Year Guidance as First EV Luce Sells Out Through 2027

Thema · 法拉利Luce电动车产能Wesentlich
2026-11-04 02:51 GMT+8

Italian sports car maker Ferrari N.V. reported third-quarter revenue of €1.9 billion, up 7% year-on-year, and net income of €431 million, up 13%, in its earnings release on November 3. Buoyed by strong sales of existing models and high-margin personalization options, the company raised its full-year adjusted earnings guidance to at least €3 billion (approx. $3.36 billion), up from €2.97 billion.

Despite initial criticism and stock price pressure following the May debut of the Luce, Ferrari’s first electric vehicle, demand has remained robust. CEO Benedetto Vigna stated during the earnings call that the 2027 production run of the Luce is “fully allocated,” meaning every car planned for next year already has a buyer. “If someone comes, he has to wait one year to get the cars,” Vigna said. The first deliveries are expected to begin in the fourth quarter.

Ferrari has not publicly disclosed the exact number of Luce units produced or sold. Earlier reporting by the Financial Times, citing unnamed sources, indicated plans to sell just under 500 units this year, with a longer-term target of 2,500 units by 2030. Vigna noted that the customer base for the Luce represents a “good mix” in terms of client type and geography, without a specific demographic pattern.

Financial data highlights the growing importance of customization, which accounted for more than 21% of revenue from cars and spare parts. This trend underscores how Ferrari maintains its industry-leading operating margins—29.7% in Q3—and brand exclusivity, protecting profitability even as it navigates the transition to electrification.

Quellen:TechCrunch

Forschung

Recruitment Software Company TestGorilla Report: 75% of UK Recruiters Believe Most CVs Are Written by AI; 87% Say Candidates' Actual Skills Fall Short

Thema · TestGorillaWesentlich
2026-11-04 02:25 GMT+8

According to a report published by TechRadar on November 3, 2026, a survey conducted by recruitment software company TestGorilla among HR professionals in the UK and US indicates that AI is significantly eroding trust in the hiring process. In the UK, 75% of recruiters believe that the majority of CVs and profiles they receive are written by AI rather than the candidates themselves.

This pervasive skepticism has led to a sharp increase in verification costs. The report states that 49% of UK recruiters cite "verifying applicant skills" as their single largest challenge in sourcing new talent, a process that can take up to 60 minutes per candidate. Furthermore, 87% of recruiters reported that applicants' actual AI skills were weaker than indicated in their applications.

"A CV used to be a candidate's account of themselves. Increasingly, it's an AI's account of what a recruiter wants to read," said Wouter Durville, CEO of TestGorilla. The survey also found that 33% of recruiters observe little difference between applications in AI fields, exacerbating screening difficulties. Notably, UK recruiters are less likely (15%) to hire based on a candidate's promise of capability compared to their US counterparts (23%).

Quellen:TechRadar

Forschung

Apple introduces 'Audio Intelligence' in watchOS beta, enabling continuous listening and conversation summaries

Thema · 苹果音频智能功能Wesentlich
2026-11-04 02:31 GMT+8

Apple released the watchOS 27.3 developer beta on November 3, 2026, which includes new features under the banner of "Audio Intelligence." These features leverage the Apple Watch's microphone for ambient listening, aiming to help users record and review daily conversations more effectively.

The update introduces two core capabilities: Siri Recap and Live Rewind. Siri Recap uses "ambient listening" to generate high-level summaries of user conversations, saving titles and key points in the new Siri app for later review. Live Rewind allows users to double-press the Digital Crown to view text transcripts of the last 15 seconds of audio detected by the watch's mic, helping them catch missed details.

Addressing privacy concerns associated with always-on listening, Apple published an "Audio Intelligence Privacy Overview." The company claims that raw audio data is handled within dedicated hardware, is not saved as a file, and is inaccessible to the operating system, apps, or Apple itself. If users enable iCloud two-factor authentication and have a device passcode, any saved summaries or transcripts sync across devices with end-to-end encryption.

This move signals Apple's attempt to expand AI assistant capabilities from smartphones to more intimate and persistent wearable devices, though it also raises further discussions about the ethical and legal boundaries of personal space recording. Currently, these features are limited to developer testing and have not yet been rolled out to the general public.

Quellen:The Verge

Forschung

Supabase Enables Claude to Issue Identity Assertions for MCP Enterprise Auth

Thema · SupabaseKurzfassung
2026-11-03

On November 3, 2026, Supabase announced that its Model Context Protocol (MCP) server now supports Anthropic's Claude acting as the direct issuer of identity assertions (ID-JAG). Previously, enabling enterprise-managed authorization required the organization's Identity Provider (IdP) to support a specific MCP extension protocol. Since only a few major IdPs like Okta currently offer this capability, this update allows enterprises using other IdPs to implement unified AI agent access control without waiting for vendor upgrades.

Under this mechanism, when a member uses Supabase via Claude, Claude verifies that the user is a current member of the Claude organization and checks their email address, then signs a short-lived identity assertion itself. Supabase validates the signature, confirms SSO identity matching, and checks admin authorization before issuing an access token. This eliminates the need for individual OAuth consent screens for each member, enabling a 'one admin approval, auto-connect for all' experience.

Administrators should note that this change affects offboarding procedures: because neither Claude nor Supabase queries the original IdP for real-time member status, disabling an account in the IdP will not immediately cut off AI agent access. Admins must remove the member from both the Supabase and Claude organizations. Additionally, Supabase trusts only one issuer URL per SSO provider; switching to Claude causes Supabase to reject all assertions signed by the original IdP until the previous URL is restored.

Quellen:supabase.com

Forschung

Former OpenAI Co-founder Karpathy Claims He Built HackerTube Video Curation Site in Two Hours Using Claude Opus 5.5

Kurzfassung
2026-11-03 13:31 GMT+8

Andrej Karpathy, a former OpenAI co-founder and current Anthropic researcher, announced on X that he built a website called HackerTube using Anthropic's Claude Opus 5.5 model. Motivated by dissatisfaction with YouTube's homepage recommendations, Karpathy stated that he filtered 6,454 repeatedly recommended videos from 20 years of Hacker News comments. He claimed the entire development process took approximately two hours.

According to reports by INSIDE and Startup Fortune, the tool processed around 43 million Hacker News comments to curate these videos, which are categorized into about 150 topics. In a blog post dated October 31, Karpathy explained the methodology: Opus 5.5 established the data extraction pipeline, while Sonnet handled filtering links into the final list. Karpathy performed basic tasks such as creating Google Cloud API keys, launching a VPS, and purchasing the domain, while delegating server updates, security hardening, deployment, and code optimization to Claude.

Designed to counter mainstream algorithmic recommendations, HackerTube requires no login, uses mid-video thumbnails instead of clickbait, and stores viewing history locally in the browser. Current information about the project relies primarily on Karpathy's personal statements and media relays, with no independent evaluation of its performance or user experience yet available.

Quellen:INSIDE

Forschung

Netflix Launches 'AI Discover' Tab for Theme-Based Recommendations

Kurzfassung
2026-11-04 02:00 GMT+8

Streaming giant Netflix is testing a new feature called "AI Discover" within its TV application, designed to help users find new shows and movies using artificial intelligence. The feature appears as a "Discover" tab in the top menu and is currently rolling out gradually to members who use Netflix in English on supported TVs in the United States.

Unlike traditional algorithmic suggestions, the first experiment under "AI Discover" surfaces content based on specific themes, such as "Something Light," "Family Movie Night," or "Quick Watches." When a user selects a theme, the system displays recommendations matching that category. If a user dislikes a recommendation, they can provide feedback, and the app will remember this preference to adjust future suggestions.

Netflix states that AI Discover is powered by "a new intelligence layer we’ve been building that responds to what members tell us in real time and keeps learning." This product update follows recent layoffs of approximately 1,000 employees at Netflix, amid slowing growth and reports of decreased user viewing time, highlighting the company's effort to boost engagement through technical innovation.

Quellen:The Verge

Forschung
Weiter →
Anmelden / Registrieren
LesenForschungRadarAnlageframework