読書リサーチレーダー投資フレームワーク
ログイン / 新規登録
すべて技術製品投資株式市場その他

読書

2026-11-2028 投稿

更新

Google Gemini adds agentic computer control to Mac app

重大
2026-11-20

Google has introduced a new "@Computer use" command in the Gemini app for Mac, enabling the AI assistant to perform agentic tasks such as sorting through files, editing documents, running applications, and accessing the internet. This feature transforms Gemini from a passive sidebar tool into an active agent that can execute complex workflows on the user's local machine.

To mitigate risks associated with autonomous file modification or deletion, Google has implemented granular access controls and automatic temporary backups. Every file Gemini interacts with is backed up; these copies are deleted when a new task begins (if the user approves the outcome) or automatically removed after 24 hours. The agent operates in a background window, typing and clicking without taking over the user's mouse and keyboard, allowing continued device usage during tasks.

The feature is currently restricted to paid subscribers (Google AI Plus, Pro, or Ultra) using Apple Silicon Macs running macOS Sequoia 15.0 or later. It is available only to users aged 18 or older in the US and supports English exclusively. This launch follows similar agentic capabilities already offered by competitors like Meta, OpenAI, and Anthropic.

ソース:TechRadar

リサーチ

NYT: Former Nvidia CEO Jensen Huang's management style reportedly built on fear and humiliation

重大
2026-11-20 21:25 GMT+8

According to a report by The New York Times, former Nvidia CEO Jensen Huang ruled through praise and fear, and also by humiliating people. This account provides new internal context for the leadership crisis at Nvidia, which saw Huang ousted earlier this month amid scandals involving alleged ties with China and personal relationships.

The Times states that Huang told subordinates that Nvidia’s wealth meant it wasn’t possible to motivate them with money, so humiliation was the only lever left. He reportedly told one executive his mother should have aborted him, and told another employee he would rip his arms off and beat him to death with them. While thousands of employees have signed a letter asking for Huang’s return, these allegations suggest a divided workforce.

The reporting indicates these incidents were not isolated. In 2005, Michael Winston, a former HR chief, threatened to sue and called Huang’s conduct the most abusive he’d seen; Nvidia paid him more than $1 million. Huang’s spokesman calls the allegations fabrications. This detailed account of Huang's internal management practices offers a specific explanation for the board's decision and employee reactions that goes beyond the external financial and regulatory scandal coverage.

ソース:TechCrunch

リサーチ

German Publisher Associations Issue Formal Legal Warnings to Meta Over AI Training Data Access

重大
2026-11-20 21:27 GMT+8

The Federal Association of Digital Publishers and Newspaper Publishers (BDZV) and the Media Association of the Free Press (MVFP), representing German media houses, have issued formal legal warnings (Abmahnung) to US technology giant Meta. They are demanding that Meta sign a cease-and-desist declaration regarding its updated Commercial Terms for business users.

The dispute centers on Meta's planned changes to its service terms, which would grant the company the right to automatically scrape, index, cache, and analyze domains, websites, and URLs associated with commercial Meta accounts, such as those of publishers. Although Meta initially included the word 'train' in the list of permitted uses for these scraped contents, it removed this specific term following massive protests from the publishing industry. However, the revised terms still do not explicitly exclude the use of this data for generative AI services, leading publishers to fear that their journalistic content could be appropriated without compensation for training Meta's own AI systems.

While technical measures like robots.txt can instruct crawlers to avoid certain areas, they are not absolute access barriers. The publisher associations argue that by securing contractual consent through these new terms, Meta could override any technical objections raised by website operators.

In response to the escalating conflict, some German publishers are preparing concrete withdrawal scenarios. Ruth Betz, Managing Director of Zeitungsverlag Waiblingen, stated that her company has played out all scenarios and is prepared to withdraw from all Meta products to protect its content rights. Similarly, Funke Mediengruppe announced it is reviewing its engagement on Meta platforms, citing the one-sided reduction of value exchange between the social network and media companies.

ソース:heise online

リサーチ

Former DeepSeek Researcher Zhang Bo Open-Sources Linger Agent and PolyStrata Engine for Local AI

重大
2026-11-20 17:42 GMT+8

On November 19, Tony Zhang (Zhang Bo), founder of Simplex AI and former AI scientist at DeepSeek, announced the open-source release of Linger, a personal AI agent designed to run entirely on user devices. The project aims to reduce reliance on cloud data centers and per-token API fees by keeping data and computation local. GitHub stars for the repository surpassed 2,000 shortly after launch.

Linger features a proactive assistant module called "Another Me," which monitors user inboxes, calendars, and Slack channels to identify pending tasks and initiate follow-ups without explicit commands. It also reviews daily interactions to adjust its behavior based on user feedback. Unlike many agents that require manual prompting, Another Me operates autonomously within defined boundaries. The system supports multi-agent collaboration, can manipulate local files, browsers, and desktop applications, and integrates with tools such as Claude Code and Codex. Users may choose to run local models to keep their context private.

To enable large model execution on consumer hardware, Simplex AI simultaneously released PolyStrata, a local inference engine. Zhang stated that PolyStrata allows a 125-billion-parameter Qwen3.8-Flash-Next model to run on computers equipped with 64GB RAM and as little as 12GB VRAM. On an RTX 3060 GPU with 12GB memory, the reported inference speed reaches 45.4 tokens per second. For Mixture-of-Experts (MoE) models, the engine places frequently used "hot" experts in VRAM and less active "cold" experts in system memory, achieving speeds up to 80 tokens per second under specific configurations.

In the LoCoMo benchmark, which measures long-term conversational memory, Linger achieved an accuracy rate of 95%, surpassing several other memory-augmented AI systems such as Mem0 and MemOS. These results are self-reported by the developer and have not yet been independently verified.

Zhang Bo previously served as a senior researcher at Microsoft Research Asia and led the development of DeepSeek-VL while at DeepSeek. He is currently a doctoral supervisor at Zhejiang University. In response to community questions, he clarified that Linger differs from OpenClaw by focusing more on direct user-agent collaboration and local environment control rather than chat platform integration. Compared to DeepSeek's DS-Bot, Linger does not bind to a single harness and supports multiple external tools and local models.

ソース:zhidx.com

リサーチ

Huawei launches new sub-brand OSTA and its first HarmonyOS AI phone, O1 Pro

重大
2026-11-20 18:23 GMT+8

On November 20, Huawei officially launched a new smartphone sub-brand called OSTA (Xingmang) along with its first device, the OSTA O1 Pro. The phone is described as Huawei's first "HarmonyOS AI phone" and features a wide-canvas form factor. It is powered by the Kirin 9030S chip, which utilizes Huawei's proprietary "Tao Law" architecture.

The launch marks Huawei's return to the sub-brand strategy six years after selling Honor in 2020. Richard Yu, Chairman of Huawei's Consumer Business Group, explained that OSTA stands for Original, Smart, Technology, and Accessible. The company positions the dual-brand strategy between Huawei and OSTA as one of differentiated positioning and complementary development, aiming to serve younger users who are increasingly adopting AI technologies.

The core innovation of the O1 Pro lies in its software: it debuts HarmonyOS 7 AI Pioneer Edition. This version introduces three key AI capabilities: system-level AI interaction, proactive intelligence, and an AI space. The phone features a "System-level AI Screen" that appears immediately upon unlocking, allowing users to interact with the assistant, Xiaoyi, directly without navigating multiple apps. Xiaoyi has been given a new visual identity named "Little Xingmang" and can perform complex tasks such as booking restaurants, sending messages, and hailing rides through natural language commands.

Hardware specifications include a 200-megapixel main camera and a Hongfeng camera, supporting flagship-grade AI photography. In gaming tests, the Kirin 9030S maintained stable 120 frames per second in League of Legends Mobile, with the wide screen increasing field of view by up to 28%.

Pricing starts at 5499 yuan for the base model, which is 500 yuan lower than the equivalent storage configuration of the Huawei Mate 90 standard edition. The top-tier 16GB+1TB version is priced at 8799 yuan.

ソース:zhidx.com

リサーチ

Four Maryland Students Sue AI Gun-Detection Firm Omnilert After Doritos Bag Mistaken for Firearm

重大
2026-11-19 07:53 GMT+8

Four current and former students at a Maryland high school have filed a lawsuit against Omnilert, the company behind an AI gun-detection system, along with Baltimore County Police, the Board of Education, individual officers, and the school principal. The legal action stems from an incident on October 20, 2025, where armed police detained the students after the AI system mistakenly identified a bag of Doritos chips as a firearm.

According to the complaint, the Omnilert Gun Detect system scanned security footage and flagged the chips bag with 90% confidence. Despite school staff and local police reviewing the image and determining it was a mistake, the alert persisted. The students were ordered to their knees, handcuffed, and searched by officers who had unholstered their weapons. One officer allegedly joked about the situation as he left the scene.

The plaintiffs allege unlawful detention and search due to technological and human errors. Damon Hewitt, president and executive director of the Lawyers’ Committee for Civil Rights Under Law, which represents the students, stated that the technology is "prone to false alerts" and failed to protect the students. The lawsuit cites findings from Maryland’s office of the inspector general for education, which reported that more than 99% of Omnilert Gun Detect (OGD) alerts in Baltimore County Public Schools were false positives requiring no police response.

While Omnilert markets its product as reliable and leading to "fewer false positives," the lawsuit challenges this claim. An Omnilert spokesperson previously told the Baltimore Banner that the company was reviewing the lawsuit and expressed regret for the fear experienced by the students and their families. A spokesperson for Baltimore County declined to comment on pending litigation, citing district policy.

ソース:the Guardian

リサーチ

Target and Walmart Q3 Earnings: AI Replenishment Coverage Exceeds 90%, Sparky Usage Grows 129%

重大
2026-11-20

In their third-quarter earnings calls, US retail giants Target and Walmart provided concrete data on how artificial intelligence is being integrated into their core operations. Target reported that its AI-driven procurement and replenishment tools now cover more than 90% of its product items, a substantial increase from 40% last year. This expansion has improved shelf availability by over one percentage point compared to the previous year, directly addressing consumer frustration with out-of-stock items.

Walmart highlighted the rapid growth of its generative AI shopping assistant, "Sparky," which saw a 129% year-over-year increase in usage in the US. The company noted that carts associated with Sparky had 40% higher values, indicating strong conversion potential. Additionally, Walmart is preparing for an era of "agentic commerce" by launching features that allow external AI agents to shop on behalf of users, complete with spending limits and permission controls.

These disclosures mark a shift from experimental AI projects to scaled operational deployment. For Target, the focus is on backend efficiency and digital twin simulations (like its "Proxima" system) to optimize supply chains before physical execution. For Walmart, the strategy combines membership growth with advanced AI interfaces, aiming to capture value as consumers increasingly delegate purchasing decisions to AI assistants.

ソース:TechOrange 科技報橘

リサーチ

Apple Plans AI Agent for Macs, Bloomberg Reports

重大
2026-11-20 18:00 GMT+8

According to a report by Bloomberg, Apple is working on an artificial intelligence agent designed to control its Mac computers. The news outlet cited a now-deleted job listing on Apple's website that sought a Senior Engineering Manager to oversee the creation of this system.

The job description outlined ambitious capabilities for the proposed tool. It stated: "Imagine an AI agent that can use a computer just like a capable person. It browses the web, works across multiple apps, executes commands, writes and debugs code, uses various services, and selects the right tools for each step. All of this happens privately and with the polish people expect from Apple." The listing also noted that the agent should be able to switch quickly between different tools, including applications, browsers, command lines, code, and APIs.

This development aligns with broader industry trends where companies like OpenAI, Anthropic, and Meta have introduced their own agentic assistants. While the exact timeline for Apple's release remains unknown due to the removal of the job posting, the initiative signals a strategic push toward integrating advanced autonomous software directly into macOS.

ソース:t3n.de

リサーチ

Sage Group FY2026 Revenue Up 11% to £2.78bn, Driven by AI Adoption

重大
2026-11-20

Sage Group, a UK-based software provider for small and medium-sized businesses (SMBs), has reported an 11% increase in annual revenue to £2.78 billion for the fiscal year ended September 30, 2026. Operating profit rose 12% year-on-year to £600 million. CEO Stephen Hare attributed this growth to the increased adoption of AI-powered services across their finance, HR, and payroll platforms.

Hare emphasized that Sage's competitive advantage in the AI era lies in its deep domain expertise and proprietary data built from processing billions of transactions annually. This context allows Sage to apply AI not just for insights but to safely complete work for customers. For example, AI in Sage Intacct is now processing 50 million invoices per year, saving customers nearly 6 million hours of work. The adoption of the Making Tax Digital agent for accountants in the UK doubled in six months.

The company highlighted three ways it is turning AI intelligence into customer value: accelerating work through embedded AI, acting on behalf of customers via agents (with 'Sage AI Workspaces' currently in beta), and connecting customers and partners through its platform. Hare stressed the importance of trust, control, and transparency, noting that agency without permissions is a business risk. Sage's approach includes flagging unusual activities for human review and ensuring AI actions are explainable and auditable.

Despite competition from traditional and emerging players, Hare stated that Sage is acquiring new customers at accelerating rates in both the US and international markets, outpacing established competitors. In the UK, the company acquired more new SMB customers in the last 12 months than in any previous year since Hare became CEO in 2018.

ソース:diginomica

リサーチ

immedio CEO Hideki Hamada: AI SDR Has Not Become the US Standard — Inbound Automation and Data Infrastructure Are Where Results Actually Appear

重大
2026-11-20 16:02 GMT+8

Six months after AI SDR dominated the SaaStr conference floor in May 2026, the category has not become standard practice in US sales operations. Hideki Hamada, CEO of immedio, an AI inside-sales company, presented this assessment at his firm's AI Growth Camp 2026 conference on November 20, 2026, backed by specific data points that explain why the replace-humans narrative broke.

The initial enthusiasm was real. At SaaStr 2026, the sponsor list was dominated by AI SDR and sales-marketing AI startups. Presentations described teams of ten salespeople reduced to one human plus twenty AI agents, or nine people's work replaced by $5,000 per year in token costs. These numbers circulated widely in Japan as well.

Hamada reports that the reality six months later looks different. AI SDR pilot programs are showing approximately 50% cancellation rates, and the vendors themselves are struggling. The company known for its "Stop Hiring Humans" billboard campaign changed its slogan to "Be more human." What is gaining attention in the US now is not AI SDR alone but the broader discipline of GTM engineering — designing the entire sales and marketing operation by combining data integration, automation tools, and AI.

The mechanism behind the failure is specific and structural, not a matter of AI capability being insufficient. Two constraints explain most of it. First, regulation: in the US, AI voice calls are treated the same as recorded calls, meaning the AI must disclose its identity at the start of the conversation or face potential fines. This eliminates the illusion that AI can seamlessly impersonate a human caller. Second, deliverability: AI can generate hundreds of emails, but spam filters punish volume. Companies must use secondary domains and send roughly 30 emails per address per day — the same limit a human faces. Despite these constraints, total outbound email volume has increased, pushing reply rates well below pre-pandemic levels.

The broader picture confirms the gap between adoption and results. Hamada cited Boston Consulting Group survey data showing that 90% of companies use AI seriously in sales operations, but only 13% have seen win rates actually improve. Adoption is nearly universal; measurable commercial impact remains rare.

Two company cases illustrate the pivot. Ramp, the US fintech that had been an early AI SDR adopter, stopped using it in December 2025. Hamada reports the reason: once every company sends AI-generated outreach with similar text, no one holds a competitive advantage. Ramp instead built an internal data platform connecting Salesforce and other systems, letting salespeople access information through AI agents. In Japan, Sansan — the business-card management company — declared "AI First" in 2025, then shifted in 2026 to a policy where humans produce results and AI assists.

What is working, according to Hamada's observation, falls into two categories. Inbound response automation — handling people who already visited your website or expressed interest — shows clear revenue growth. The US company Qualified, which automates inbound website engagement, was acquired by Salesforce. immedio itself launched an "AI customer service" product in 2026 and reports that AI-booked meetings are converting directly into deals. The second working pattern is data infrastructure that supports human sellers: an AI agent that reads CRM, email, and internal documents, then delivers "what you should know right now" to each salesperson. Hamada describes this as equivalent to assigning an excellent sales coordinator to each seller.

ソース:テクノエッジ TechnoEdge

リサーチ

OpenAI Expands Dots Agent Access to US Business Standard and Tests on Plus, Redesigns Mobile App

重大
2026-11-20 16:07 GMT+8

On November 19, 2026, OpenAI announced that it had started providing its personal AI agent product, "dots," to users of the ChatGPT Business Standard plan in the United States. Additionally, a small number of individual ChatGPT Plus users are now able to test the service. This expansion follows the initial launch of dots for Pro and Business Premium users in eligible regions on September 29, 2026.

The dots agent operates using a dedicated cloud computer and connected applications, allowing it to continue working on tasks even when the user is not actively chatting. For example, it can prepare presentation materials based on data or customer feedback and update them as new information arrives. Users retain control over the level of autonomy granted to the dot, can change instructions mid-task, or pause the agent entirely.

Thibault Sottiaux, who leads the ChatGPT and Codex product lines at OpenAI, stated that he is seeing more frequent praise for the dot's ability to anticipate needs and act proactively. He described the recent changes as an effort to improve the usability of ChatGPT for the 1.2 billion people already using it, noting that the company wants to gather user feedback before expanding the new design further.

Simultaneously, OpenAI began rolling out a redesigned mobile application interface, initially targeting Pro users aged 18 and older in eligible regions. The new design features a dedicated tab bar at the bottom of the screen, allowing direct access to conversations with one's own dot. Frequently used functions can also be pinned to this tab bar. Some Plus users meeting specific criteria will also receive the new design.

Regarding availability, the Plus testing phase excludes users in the European Economic Area (EEA), the UK, and Switzerland. For Enterprise, Edu, and Healthcare workspaces, dots are now enabled by default unless administrators have previously opted out, though admins still control which members can use the feature. The release notes for Business Standard confirm the US launch but do not specify a timeline for other countries.

ソース:gihyo.jp

リサーチ

Capgemini Executives: Multi-Agent AI Governance Needs Three Control Layers, Not Explainability

重大
2026-11-20

Capgemini's Deputy Group CTO Sudhir Pai identifies a specific failure mode that distinguishes multi-agent production from single-agent testing: every agent can be right while the system is wrong. The problem is not behavioral drift in any one agent, but interactions nobody designed. As enterprises deploy growing numbers of agents, the systems were not built to test or govern emergent coordination.

Pai's proposed response is a three-level architecture. The first level is a harness around each individual agent. The second is a control plane for each agentic system. The third, and the one he argues will dominate 2027 enterprise agendas, is an enterprise control plane across all systems. This top layer must incorporate a register of every agent and agentic system, identity and permissions for each one, monitoring of outcomes delivered and consumption, visibility into how agents interact, limits on what they can do together, and a way to stop them.

Mark Roberts, Head of Capgemini's AI Futures Lab, frames the shift in safety philosophy. The traditional approach—limiting agent access and checking that every decision remains within business rules—assumes you can verify decisions by examining reasoning steps. But some new models no longer show all their reasoning steps. Roberts argues that even if you can explain each component, you still do not know how the overall system will behave. The question is not whether more explainability leads to more safety; it is whether containment at the macro level can substitute for understanding at the micro level. He draws an analogy to traffic systems: you do not need human oversight in every loop, but you need it where it matters, with layers of safeguards so mistakes are caught when they are small and recoverable.

The argument is attributed reasoning from a consulting firm's technology leadership, not an independently validated framework. Capgemini sells enterprise AI transformation services, and the proposed architecture aligns with services it could deliver. No published deployment data or failure analysis accompanies the three-level model. The containment premise—that acceptable bounds can be specified without full reasoning transparency—remains an assumption that may not hold for novel emergent interactions the bounds were not designed to anticipate. Pascal Brier, Capgemini's Chief Innovation Officer, positions this as the main challenge for organizations in 2027: scaling AI autonomy without losing control.

ソース:diginomica

リサーチ
次へ →
ログイン / 新規登録
読書リサーチレーダー投資フレームワーク

The underlying principle Hamada draws from these observations is straightforward: automate only the "nudge" — the area where the customer already has interest and just needs a push. Everything else requires human judgment, with AI in a support role rather than a replacement role.

One important qualification: Hamada sells AI inside-sales solutions through immedio, so emphasizing the limitations of the replace-humans model aligns with his commercial positioning. His company benefits from the narrative that AI works best in bounded, well-defined tasks rather than wholesale team replacement. The BCG 13% figure may also reflect measurement lag — companies that adopted AI recently may not yet show results in their sales pipeline — rather than a permanent ceiling. Still, the specific mechanisms he identifies (regulatory disclosure requirements, spam-filter volume limits, text homogenization destroying differentiation) are structural constraints that do not resolve with better models or more token budget. They define the boundary conditions within which AI SDR can operate, and those conditions favor inbound and support workflows over outbound replacement.