LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
TodoTecnologíaProductosInversiónBolsaOtros

Lectura

2026-11-1852 publicaciones

Actualizar

French AI Firm Mistral Releases Vibe Desktop App with Git Worktree Parallelism

Resumen rápido
2026-11-18 22:40 GMT+8

French artificial intelligence startup Mistral AI released the desktop application for its AI agent Vibe (formerly Le Chat) on November 18, 2026. Previously, Vibe was only available as a web app, command-line tool, or smartphone app. According to tech outlet The Decoder, the new application combines the local features of Vibe CLI with the cloud sessions from Vibe Code Web.

A core feature of the app is support for parallel task processing. Users can issue multiple instructions to Vibe simultaneously within the same project, with the system using Git worktrees technology to isolate these parallel tasks and prevent code conflicts. The agent can work either in a local repository or in a cloud sandbox. Beyond coding tasks, the Work mode now also has access to local files, though Mistral does not mention any computer-use capabilities such as controlling browsers or operating other software.

Vibe Desktop is currently available for macOS and Windows, with a Linux version planned for the future. While this update fills a gap in Mistral's desktop experience, it still lags behind competitors: Anthropic's Claude desktop app has allowed AI to open apps, control browsers, and fill out spreadsheets since this spring.

Fuentes:The Decoder

Investigación

Google Cloud Unveils Technical Isolation Measures at Brussels Sovereign AI Summit to Address EU Trust Deficit

Tema · Google Cloud Sovereign AI in EuropeMaterial
2026-11-18 22:13 GMT+8

At the Google Cloud Sovereign AI Summit held in Brussels in November 2026, US cloud provider Google Cloud detailed its technical commitments aimed at alleviating European customers' trust crisis. As geopolitical tensions rise, European buyers are no longer willing to migrate data to US public clouds solely based on 'alliance' status, demanding stricter control instead.

Marianne Janik, Vice President at Google Cloud, stated that technology can remove the need for trust itself. She noted that if customers hold their own encryption keys outside of Google Cloud, Google cannot decrypt or hand over data even under forced orders. This strategy centers on achieving 'inaccessibility' through architectural design to bypass political trust issues.

To this end, Google Cloud showcased three main deployment models: first, 'Google Cloud Data Boundary,' allowing control over data residency and access within the public cloud; second, 'Google Cloud Dedicated,' running Google's stack on isolated infrastructure operated by local partners, such as S3NS (controlled by Thales) in France, which meets SecNumCloud 3.2 standards and can run for up to 12 months if cut off from Google; and third, 'Google Distributed Cloud,' which runs directly in customer-owned data centers, supporting connected or air-gapped modes.

However, whether these technical measures ensure continued market access depends on EU legislative progress. The currently amending 'Cloud and AI Development Act' (CADA) sets four 'assurance levels.' Reports indicate that Level 4 requires no third-country entity to control the design or development of the software, which would exclude services built on US technology, even when operated by European partners like Thales.

Although Google Cloud emphasizes that its stack is built on open-source architectures to avoid lock-in and promises to legally object to any government orders suspending services, European regulators and experts remain skeptical. Falk Steiner, a commentator for German media outlet heise online, pointed out that the real risk is not forced data handover but a 'soft kill-switch'—where US companies stop providing critical software updates, causing so-called sovereign clouds to paralyze within months due to lack of maintenance.

For investors, this dynamic reveals the structural challenge facing US tech giants in Europe: they must invest heavily in complex localized isolation architectures to maintain business, but the ultimate outcome may depend not on technical capability, but on whether EU legislators accept this 'controlled dependency' as a solution for digital sovereignty.

Fuentes:diginomica

Investigación

Google Updates Gemini Privacy Settings: Users Can Now Opt Out of AI Training While Retaining Chat History

Tema · Gemini Privacy Settings UpdateResumen rápido
2026-11-18 21:31 GMT+8

According to a report by German tech outlet heise online, Google has updated the privacy settings for its Gemini AI assistant. Previously, users who wished to retain their chat history for longer than 72 hours were automatically consenting to have those conversations used to train Google's AI models. If they opted out of training, their chats were deleted after 72 hours.

With the new update, these two settings are now independent. Users can choose to disable AI training for their data while still having their temporary chats and chats with disabled history stored in their account for approximately two months. Google states this extended retention period is intended to protect users and the public, allow for responses and feedback processing, and ensure availability in case of system failures. For users who enable the history function, data storage remains up to 18 months, but users can now explicitly decide whether to provide this data for AI training.

Fuentes:heise online

Investigación

Workday Acquires Honeydew and Launches Copilot Plugin to Strengthen Enterprise AI Semantic Consistency

Tema · Workday-HoneydewMaterial
2026-11-18 21:29 GMT+8

At the Workday Rising EMEA conference in Barcelona, enterprise software vendor Workday announced a key acquisition aimed at solving the "context" problem in enterprise AI: Honeydew. This startup provides a semantic layer that ensures business metrics maintain consistent meaning across different scenarios. Workday stated that when users ask an AI agent for figures or actions, the agent must understand the terminology exactly as the user does; otherwise, the answer may be logically flawed even if it appears correct. Honeydew's technology tracks business definitions within an organization and generates precise queries based on the requester and use case. The deal is expected to close this fiscal quarter, with terms undisclosed.

In addition to the acquisition, Workday released the early access version of the Sana MCP plugin, designed specifically for Microsoft Copilot. This integration allows users to connect directly to Workday's AI agents from within the Copilot interface to handle HR and finance tasks, while respecting Workday's internal permissions and policies. This signifies a shift towards "headless architecture" and interoperability, enabling users to trigger complex backend workflows without leaving their primary productivity tool.

Furthermore, Workday unveiled a fully independent EU Sovereign Cloud, operated by a new entity incorporated in Ireland with a board of EU citizens, addressing European customers' needs for data sovereignty and legal compliance. On the deployment front, Workday Payroll will expand to seven new markets including the Netherlands, Spain, and Singapore. The company claims that using AI accelerated the localization process, reducing the time to launch in Singapore by ten times compared to previous rollouts.

Fuentes:diginomica

Investigación

Fintech Payhawk raises $280M at $2B valuation to build AI-native finance OS

Tema · Payhawk Series CMaterial
2026-11-18 19:12 GMT+8

London-headquartered fintech Payhawk has raised $280 million in a Series C round led by Growth Equity at Goldman Sachs Alternatives, valuing the company at $2 billion. The round doubles its previous $1 billion valuation from 2022 and brings total funding to over $515 million. Payhawk plans to use the capital to build what it calls a global "Finance Operating System" for the AI era.

The company’s strategy centers on integrating three layers: AI agents that perform financial tasks, payment infrastructure that allows these agents to interact with money, and enterprise controls that govern their actions. Payhawk states that 77% of its customers already use its AI agents daily for tasks such as invoice processing, expense tracking, and procurement. Co-founder and CEO Hristo Borisov argues that simply adding AI to existing processes is insufficient if payments, budgets, and approvals remain in separate systems.

A key component of Payhawk’s approach is a shift in its business model. The company is pursuing a card-independent model where subscription pricing does not rely on interchange revenue. Borisov explains that this aligns Payhawk’s growth with helping customers save money rather than spend more. This structure allows its AI-native products, such as the US Financial Controller Agent, to work with customers’ existing business cards without requiring them to switch providers or reissue cards.

Payhawk faces competition from other spend management platforms expanding into AI-powered operations, including Danish firm Pleo, Berlin-based Moss, and US rival Ramp. The global spend management software market is estimated to grow from $17.54 billion in 2025 to $31.92 billion by 2030. Payhawk positions itself as a European category builder, aiming to define the infrastructure for running financial operations in the age of AI.

Fuentes:Tech Funding News

Investigación

Microsoft Launches 'Licensed Weights' at Ignite, Letting Enterprises Customize Frontier Models While Retaining Data Control

Tema · Microsoft Licensed WeightsMaterial
2026-11-18 20:56 GMT+8

At the Ignite conference on November 18, 2026, Microsoft CEO Satya Nadella announced several new models from Microsoft AI (MAI), but the standout initiative was 'Licensed Weights.' This program allows enterprise customers to customize Microsoft's frontier models using their own data and run them securely within the Azure environment.

Described as a selective approach to open-weight AI, Licensed Weights offers more customization than proprietary models from competitors like OpenAI and Anthropic, but unlike fully downloadable open-weight models such as Mistral’s Le Chonk, access is managed. Interested companies must engage through their Microsoft account representatives. MAI CEO Mustafa Suleyman stated during the keynote, “You control who can access it. Your data and adaptations stay entirely yours.” He further clarified that Microsoft will only use customer information to train its models if explicitly agreed upon.

The move addresses widespread enterprise concerns about intellectual property leakage. Suleyman echoed Palantir CEO Alex Karp’s sentiment, noting, “No one wants to pour their IQ into someone else’s model.” By providing this hybrid option, Microsoft aims to balance high-performance capabilities with strict data privacy and IP control.

While OpenAI and Anthropic currently dominate the enterprise AI market, and other Western labs like Thinking Machines are releasing competitive open-weight models, Microsoft’s Licensed Weights differentiates its offering by targeting businesses that seek customization without the security complexities of fully open weights.

Fuentes:ZDNET

Investigación

Supply Chain Attack on Optimizely Compromises 178 Organizations Including BBC and GitLab

Tema · Optimizely Supply Chain AttackMaterial
2026-11-18 14:51 GMT+8

Website experience optimization provider Optimizely confirmed that attackers exploited a vulnerability in its Web Experimentation product to inject malicious JavaScript into customer projects. This supply chain attack caused browsers visiting affected websites to automatically load malicious code. Researchers identified 188 infected Optimizely JavaScript snippets involving 178 organizations, including prominent entities such as the British Broadcasting Corporation (BBC), code hosting platform GitLab, furniture retailer IKEA, and HSBC Taiwan.

The injected malicious code served multiple harmful functions: it could tamper with cryptocurrency wallet addresses displayed on web pages, redirecting user transfers to attacker-controlled accounts; monitor form inputs to steal credit card information and cookie data; and even deliver iOS exploit kits to iPhone and iPad users, attempting to further compromise mobile devices. Security researchers observed approximately $7,040 in cryptocurrency received by wallets controlled by the attackers.

Optimizely stated that it has patched the vulnerability, removed the injected malicious code, and strengthened security protections for affected accounts. This incident highlights the significant risks posed by third-party scripts in modern website architectures: even if major websites themselves are not breached, a compromise of their trusted external services directly threatens end-user privacy and financial security. Enterprises need to enhance integrity checks, Content Security Policy (CSP) configurations, and monitoring for changes in referenced external JavaScript.

Fuentes:iThome

Investigación

OpenAI Classifies GPT-6.2 Astra as High Risk for Self-Improvement, Launches 'Control Room' Monitoring System

Material
2026-11-18 17:16 GMT+8

US-based artificial intelligence company OpenAI released its latest flagship model, GPT-6.2 Astra, and the more economical daily-use version, GPT-6.2 Sol, on November 17, 2026. Beyond performance improvements and cost reductions, a key change in this release lies in safety governance: OpenAI has classified GPT-6.2 Astra as "High" risk in the dimension of "AI self-improvement," the first time it has assigned this rating to an externally released model.

According to German tech media outlet heise online, OpenAI believes that as model capabilities grow, AI is increasingly involved in the development of its subsequent versions. This self-improvement capability could drive AI development faster than safety measures and human control can keep up, posing new security risks. In response, OpenAI has strengthened the isolation of research environments and expanded monitoring.

As part of these countermeasures, OpenAI introduced a new central system called "Control Room." This system is designed to track model workloads within research infrastructure and can quickly halt related tasks if security concerns arise. This marks a move toward more specific technical controls by frontier labs in managing recursive AI development.

In independent tests by Artificial Analysis, GPT-6.2 Astra scored 62 points on the Intelligence Index, ranking second behind Anthropic's Claude Fable 5.2. Although OpenAI self-assessed the models as "Critical" or "High" risk in cybersecurity and biological/chemical capabilities, the company emphasized that the new models also show progress in defensive programming and vulnerability remediation. The models are currently available to ChatGPT Plus, Pro, Business, Enterprise, and Edu users, as well as API developers.

Fuentes:heise online

Investigación

Co-authors of 'Autonomous': The Shift from Supervised Automation to Autonomous Execution Is a Step-Function Change, Not a Gradual Slope

Material
2026-11-19 16:05 GMT+8

Vala Afshar, Henry King, and Dr David Bray, co-authors of 'Autonomous', argue that enterprises are facing a widely misjudged inflection point: the transition from supervised automation to autonomous execution is a step-function change, not a gradual slope. Leadership's obligation at each stage is not simply to retire legacy tasks, but to explicitly define and mandate the higher-order judgment, creative, and relational work that only humans can perform.

This assessment targets a common error they observe: leaders who deploy AI agents into production assume the transformation is complete, ignoring the long intermediate phase. They call this phase "hands ready," characterized by neither humans nor agents being fully in command, requiring both parties to learn to share control. Treating human oversight as a passive "in-the-loop" checkbox ignores the reality that greater autonomy demands stronger, active human accountability.

Their core argument is that humans and AI are not competitors, but complements. Humans bring judgment, context, and creativity; machines bring speed, scale, and tirelessness. When intentionally combined, they produce better decisions, more durable value, and greater organizational resilience. However, these outcomes do not happen automatically; they depend on deliberately designed shared operating structures rather than default automation.

They propose an operational loop called aSUDA—sense, understand, decide, act—as the operating system for hybrid organizations. Within this loop, the agent-executed inner loop continues to expand in speed and scope, but leaders must ensure this expansion does not outpace the relational outer loop. As cognitive execution shifts to agents, investments in human oversight, trust, and role security must expand proportionally.

At a macro level, they note that traditional enterprises prioritize predictability above all else, relying on functional silos, rigid hierarchies, and annual planning cycles, thereby accepting a false trade-off between stability and performance. They argue that agentic AI can now maintain operational equilibrium continuously at machine speed, enabling organizations to achieve performance levels unreachable by static hierarchies while remaining structurally fluid—a concept they term "positive instability."

Relationship Design is the core competency that ties everything together. Redesigning roles, re-skilling skills, re-deploying talent, re-structuring reporting lines, re-calibrating incentives, and re-mandating human-exclusive work are the mechanics of transformation; the aSUDA loop is its operating system; positive instability is its performance mode; and accessible, secure AI is its precondition. Yet Relationship Design is what coheres them into an organization that people and agents can actually operate within. They argue that every enterprise, regardless of size or sector, must establish Relationship Design alongside strategy, finance, and technology as a primary management discipline.

Fuentes:diginomica

Investigación

GM Plans 20% Boost in North American Factory Automation by 2028, Ties Supplier Contracts to Automation Scores

Tema · GM Automation StrategyMaterial
2026-11-18 17:22 GMT+8

According to a Reuters report citing two people familiar with the matter, General Motors (GM) plans to increase the use of robots and other automation technologies in its North American factories by 20% before 2028. This previously unreported target is designed to counteract rising labor costs and production expenses associated with manufacturing reshoring to the United States.

Sources revealed that GM is developing its own robots, assembling them at a 90-year-old facility in Kokomo, Indiana, which produces automotive electronic components like engine control modules. The current focus is on Autonomous Mobile Robots (AMRs) for intra-factory parts transport. This marks the first disclosure of GM's internal robot manufacturing capability; previously, the company relied primarily on external vendors. Last year, GM stated it had deployed about 100 robotics experts, AI engineers, and hardware specialists to develop advanced robotic systems.

The automation mandate extends to the supply chain. Two other individuals familiar with GM's supplier system indicated that GM now requires component suppliers bidding on new contracts to submit their own operational automation plans. GM uses a scorecard to evaluate suppliers' automation investments, incorporating these scores into contract award decisions. For suppliers seeking GM orders, automation levels have become a direct entry barrier rather than just an efficiency tool.

This strategic shift follows record-breaking labor agreements reached in 2023 between the United Auto Workers (UAW) union and GM, Ford, and Stellantis. At the time, GM noted that the new contracts would add approximately $575 in labor costs per vehicle. Analysts widely expected automakers to offset these costs through increased automation. Additionally, US government policies promoting domestic manufacturing have further pressured local production costs.

However, expanded automation risks strong opposition from labor groups. Ryan Hiestand, Vice President of the UAW, which represents about 50,000 GM factory workers, criticized the move in a statement following the Reuters report: "What they call increasing efficiency is firing the people who actually create products and profits." UAW President Shawn Fain has also warned that AI and automation pose threats to workers. With existing labor contracts set to expire in spring 2028, automation is expected to be a central issue in the next round of negotiations. GM did not immediately respond to requests for comment but has previously stated that its use of automation is intended to improve safety, quality, and efficiency.

Fuentes:TechOrange 科技報橘

Investigación

Alibaba Cloud's Qoder Lowers Default Thinking Effort for Several Models, Cutting Costs by Up to 40.5%

Tema · Qoder Thinking EffortResumen rápido
Verificado 2026-11-18 17:54 GMT+8

Alibaba Cloud's AI coding assistant Qoder recently announced adjustments to the default "Thinking Effort" settings for several large language models, based on evaluations conducted on its proprietary dataset, QoderBench. This parameter controls how much reasoning the model performs before taking action, directly influencing response time and token consumption.

Specifically, the default setting for GLM-5.3, Kimi-K3 (and its replacement Kimi-K3.1-Pro), Kimi-K2.8-Preview, and DeepSeek-Flash has been lowered from "Maximum" to "High". The defaults for GLM-5.3-Flash and DeepSeek-V4-Pro remain at "Maximum". Users can still manually select "Low", "High", or "Maximum" in the model manager.

According to the official Qoder blog, this change stems from testing nearly 200 real-world task cases. The data shows that reducing the thinking effort by one level from "Maximum" to "High" decreases output tokens by 19% to 48% and model call counts by 6.5% to 31.5%. In terms of cost, per-run expenses fell by 6.9% to 40.5%, while relative scores declined by only 5.7% to 16.1%. For instance, switching GLM-5.3 to "High" resulted in a 40.5% cost reduction with only a 6.3% relative drop in score.

Qoder notes that for most everyday development tasks, the marginal benefit of excessive reasoning diminishes, leading instead to increased wait times and Credits usage. These results are vendor-reported and do not specify changes in end-to-end task duration; actual impact may vary depending on the specific workload.

Fuentes:Alibaba Cloud Community

Investigación

Tsinghua-Incubated Startup Launches AI Hardware Development Platform Supporting Natural Language Circuit and Code Generation

Tema · MOCE DesignerResumen rápido
2026-11-18 16:13 GMT+8

According to a report by Zhidongxi on November 18, Moce Intelligence, a startup incubated by Tsinghua University's X-Lab, recently officially released MOCE Designer, a hardware development platform that supports direct interaction via natural language.

In simple terms, developers can describe what they want to build and use AI within the same workbench to complete requirement breakdown, component selection, circuit connection design, code generation, compilation, and flashing, followed by debugging on the physical device. After prototype validation, teams with further development needs can also connect to structural processing, trial production, mass production, and sales resources.

Moce Intelligence calls this mode 'Vibe Building'. The young startup was founded in 2025 with a team of about 20 people. At the launch event, Moce announced that its module library will expand to 200 items by the end of the year, with plans to gradually increase it to over 1,000.

Before the official release of MOCE Designer, Moce Intelligence held its first Vibe Hardware Hackathon at Tsinghua University as a public beta test. Over 100 participants spent two and a half days developing and debugging, resulting in 15 AI hardware creative projects showcased on stage. Among them, the Tsinghua Next-Day Band built a theremin capable of being played without touch for less than 200 RMB. Another participant with no prior hardware development experience independently created an innovative hardware device called 'Tomato Monster', combining a desktop robot and a lock box.

Fuentes:zhidx.com

Investigación
Siguiente →
Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión