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

2026-11-1392 投稿

更新

GDC Survey: Generative AI Usage Among Game Developers Rises to 40%, Mobile Hits 70%

重大
2026-11-13 01:00 GMT+8

According to the latest annual industry survey published by the GDC Festival of Gaming, more video game developers are using generative AI tools in their jobs compared to last year. Forty percent of developers surveyed say they use it “as part of their job,” an increase from 36 percent in 2025.

Usage varies significantly across segments. It was higher among game developers working on free-to-play games at 58 percent, and reached 70 percent among workers in mobile-only studios. This suggests that areas requiring rapid iteration and high volume see greater penetration of AI tools.

In terms of specific applications, 73 percent of respondents said they used AI for research and brainstorming, while 62 percent used it for data analysis and summarization, and 56 percent for code generation and debugging. In contrast, only 22 percent used generative AI for concept art or visual ideation, and just 8 percent used it for production art or final game assets. This indicates that AI currently serves primarily as an auxiliary tool for research and coding rather than directly replacing core creative production.

The survey comes as the gaming industry continues to reckon with generative AI. While many indie developers have spoken out strongly against the technology, recent disclosures by House House, the developer of Big Walk, regarding their "sparingly" use of LLMs to solve technical challenges have sparked intense debate about AI usage boundaries. Additionally, Xbox CEO Asha Sharma has hinted that AI will be central to Xbox's future, stating that "A small team will be able to build worlds that today demand hundreds of developers and hundreds of millions of dollars."

ソース:The Verge

リサーチ

Anthropic Launches Claude for Manufacturing Plugin Integrating Autodesk and Oracle Systems

トピック · Claude制造业插件重大
2026-11-12

US-based artificial intelligence company Anthropic launched "Claude for Manufacturing" on November 12, 2026. This new plugin is designed specifically for the manufacturing industry, including 17 pre-built skills and four connectors that integrate directly with core enterprise systems such as Autodesk Fusion, Infor, Microsoft 365, and Oracle NetSuite.

According to Anthropic's official blog, the tool aims to address the pain point of manufacturing teams spending much of their day gathering information. Citing a 2025 survey by the Manufacturing Leadership Council, which found that 79% of manufacturers still enter data manually, Claude for Manufacturing automates tasks ranging from quoting new jobs to diagnosing stopped machines. This allows estimators, quality engineers, and technicians to focus on decision-making rather than data organization.

Specific capabilities provided by the plugin include: automatically reading customer requests and comparing historical costs to draft quotes; documenting inspection failures and generating corrective action reports; analyzing sensor data and repair notes to diagnose equipment breakdowns; and preparing requests for supplier quotes. All operations are designed to require human review and approval before reaching customers or suppliers, ensuring human control over critical business steps.

Notably, the plugin does not directly control production equipment. For scenarios requiring interaction with programmable hardware, Anthropic is developing the "Model Hardware Standard" research preview, intended to provide a standard way for equipment to describe its measurements, settings, and safety limits. Currently, users can install the plugin via Team or Enterprise plans and configure their working environment using natural language instructions.

ソース:Claude

リサーチ

Former Nvidia CEO Jensen Huang Agrees to Testify Before US Congress

トピック · 黄仁勋间谍指控案構造的
2026-11-13 00:36 GMT+8

According to TechCrunch, former Nvidia CEO Jensen Huang has agreed to testify before the US Congress by Thanksgiving. The announcement was made on Thursday by House Foreign Affairs Committee Chairman Brian Mast, citing reports from Bloomberg and NBC News.

Huang stepped down as CEO following allegations of an affair with a woman suspected of spying for China. He agreed to testify on Wednesday after several senior lawmakers threatened to issue subpoenas.

The testimony involves three key committees: the House Foreign Affairs Committee, the House Select Committee on the Chinese Communist Party, and the House Intelligence Committee. Their respective chairs, Brian Mast, John Moolenaar, and Rick Crawford, had sent a letter to Huang on November 4 requesting his appearance.

"Jensen and I have been very at odds... on export controls, but Nvidia is a very important company to the United States of America," Mast told reporters. "I don’t want to see anything hurt in Nvidia, but he has some very serious questions to answer."

The hearing will take place behind closed doors, though Mast stated he expects the briefing to be recorded on video and transcribed. This development marks an escalation in legislative oversight regarding AI chip giants and their ties to China.

ソース:TechCrunch

リサーチ

Jensen and Lori Huang Jointly Control $135 Billion Nvidia Stake; Trust Structure May Stabilize Shares

重大
2026-11-12 22:13 GMT+8

According to U.S. Securities and Exchange Commission (SEC) filings, former Nvidia CEO Jensen Huang and his wife, Lori Huang, jointly own approximately 800 million shares of the company, valued at roughly $135 billion. This substantial holding is not directly controlled by Huang alone but is managed through a complex web of trusts and limited liability companies (LLCs) established for tax and estate planning purposes. Huang directly owns about 70 million shares, valued at approximately $12 billion.

While owning company stock through trusts is common among tech founders, equal joint control with a spouse over such a large stake is less typical, tax experts note. As reported by CNBC, the Huangs are co-trustees of the "Jen-Hsun and Lori Huang Living Trust," created in 1995, which holds about 530 million shares. They also co-manage two irrevocable trusts: the "Huang 2012 Irrevocable Trust" (holding 31.4 million shares) and the "Huang Irrevocable Remainder Trust" (holding 109 million shares). Additionally, about 135 million shares are held by LLCs where the living trust is the sole member.

For investors, this equity structure presents a dual impact amidst Huang's sudden resignation. On one hand, it introduces uncertainty regarding ownership and control. On the other, it may serve as a stabilizer for the share price. Legal experts indicate that the terms of irrevocable trusts are difficult to change, and since both Jensen and Lori Huang are co-trustees, neither can easily remove or sell the Nvidia shares held within these trusts. This implies that even in the event of divorce or other personal changes, these core assets are unlikely to be rapidly liquidated into the market.

In contrast, revocable living trusts typically allow for more flexible asset disposition, though executive stock agreements often impose strict limits on large-scale sales. Kathleen Martone, a partner at Varnum LLP, noted that shares in a living trust can usually be sold or divided relatively easily in case of divorce, similar to a "joint checking account." However, irrevocable trusts are designed for asset protection and tax benefits, maintaining independence from creditors, including divorcing spouses. Consequently, while Huang's departure has triggered market volatility, the locked-in structure of his family wealth may reduce insider selling pressure in the long term.

ソース:CNBC

リサーチ

AI Newsletter Newcomer and Bessemer Release Inaugural AI 100 List: Anthropic Tops Ranking with $3.5 Trillion Collective Valuation

トピック · AI 100榜单重大
2026-11-13 03:28 GMT+8

Tech media outlet Newcomer and venture capital firm Bessemer Venture Partners jointly released the inaugural "AI 100" list on November 12, 2026, during the Cerebral Valley AI Summit in San Francisco. The list aims to identify the 100 most promising private artificial intelligence companies today, representing a reinvention of Bessemer's storied "Cloud 100." According to disclosed data, the collective valuation of these 100 companies sits at approximately $3.5 trillion.

Anthropic took first place, with reports indicating its expected IPO next week. OpenAI followed in second place, and Thinking Machines ranked 21st. The list reveals that researcher-heavy model companies account for only 9% of the names but 55% of the collective valuation. Inference infrastructure giants (such as Fireworks, Baseten, Fal, and Modal) and earlier-generation startups that have pivoted to perceived AI stars (like Stripe, Anduril, Ramp, and Canva) also occupy significant positions.

Newcomer and Bessemer emphasized that the ranking weighed quantitative factors including revenue, valuation, and market share alongside less tangible criteria such as people and culture. Despite potential controversies, the list reflects key trends in current AI investment: video generation presents rich opportunities, with ElevenLabs, Fal, Higgsfield, and Runway ranking highly, filling the void left by Anthropic abandoning video and OpenAI shutting down Sora. Coding and legal/customer success applications remain hot sectors, with companies like Cognition, Factory, Harvey, and Legora standing out.

However, the report also highlights industry precarity. Unlike previous SaaS companies with predictable revenues, many listed AI firms face massive compute costs and uncertainty. For instance, Prometheus (#33), backed by Jeff Bezos with $18.2 billion in funding and a $41 billion valuation, has yet to reveal any product plans. Fal (#16) recently closed a $1.3 billion Series F, signaling investor bets on rising inference demand. For general readers, this list serves not just as a ranking, but as a barometer for the transition from AI hype to commercial reality, while warning of liquidity and profitability risks behind high valuations.

ソース:newcomer.co

リサーチ

Fireworks AI Hit by Second Breach; Notion Confirms Leak of AI Chat Transcripts for ~1% of Users

トピック · Fireworks AI重大
2026-11-12 23:02 GMT+8

Fireworks AI, a platform used by developers to access open AI models, has confirmed a cyberattack where hackers breached its cloud provider to gain access to a development system, exposing some customer data. This marks the company's second security incident in as many months.

Online document editor Notion confirmed it was among the affected customers, stating in a disclosure that the breach impacted around 1% of its own users. Notion reported that hackers stole chat transcripts from Notion AI customers who used the “Insights” feature. These transcripts include “any input to or output from the model that was written to the chat.” While much of the data was unattributable, the incident highlights the supply-chain risks in AI: when an infrastructure provider is compromised, downstream enterprises relying on them face direct exposure.

Fireworks AI believes the attackers were financially motivated, aiming to sell tokens obtained from breaching access to large language models so unauthorized third parties could consume AI resources without paying. Notion disconnected its systems from Fireworks on November 3 and deleted stored data.

ソース:TechCrunch

リサーチ

US Democratic Lawmakers Launch Probe, Questioning 'Illegal Export Tax' on Nvidia by Trump Administration

トピック · 英伟达美出口税争议重大
2026-11-13 06:42 GMT+8

On Thursday, US Congressional Democrats formally launched an inquiry into how the Trump administration handled Nvidia's chip sales to China. This follows the company's recent ouster of its founder and CEO, Jensen Huang.

The investigation is led by Rep. Richard Neal, the top Democrat on the House Ways and Means Committee, and Sen. Ron Wyden, the top Democrat on the Senate Finance Committee. In a letter to US Commerce Secretary Howard Lutnick, they requested the department preserve records regarding an arrangement they deem "questionable" and allegedly "illegal." The letter describes a mechanism criticized as an "unconstitutional export tax system": President Trump agreed to allow sales of Nvidia's H200 chips to China, provided that these Taiwan-manufactured chips were first shipped to the United States. Upon arrival, Trump added a tariff, effectively taking a cut of Nvidia's sales to China. Subsequently, the Commerce Department issued the export license.

As reported by the AP, this revenue-sharing deal was public. In August 2025, both Nvidia and AMD agreed to share 15% of their revenue from chip sales to China with the US government as part of a deal to secure export licenses. However, the Democrats' letter questions why Lutnick continued to allow Megaspeed, a Singapore-based company where an alleged Chinese spy worked, to buy Nvidia chips even after being reportedly informed by US intelligence about suspicions regarding Jensen Huang's relationship with her. It also questions why other Asian companies with no physical presence at their listed addresses were green-lit.

The inquiry also examines whether Nvidia's political donations influenced the decision. The letter notes that the export deal was approved after Trump learned of Huang's "compromised position" due to his affair with an alleged spy, and after Nvidia donated $26 million toward Trump's projects, including the White House ballroom. CNBC reported that Nvidia committed at least $26 million to Trump-backed projects, including $25 million for the ballroom and $1 million for the 2025 inaugural fund.

This is the second Congressional investigation into Nvidia's China sales and its dealings with Trump since news of the alleged spy affair broke. Previously, Republican-led committees demanded Huang testify in closed-door sessions. Although Democrats, as the minority party, cannot compel answers, this move signals an escalation in legislative oversight over the interplay between executive power and tech giant interests.

ソース:TechCrunch

リサーチ

Cisco Director of Government Affairs: EU Digital Networks Act Should Not Extend Telecommunications Obligations to Enterprise Digital Services

トピック · 欧盟数字网络法案重大
2026-11-13 16:00 GMT+8

The EU Digital Networks Act (DNA) is currently still in the legislative process—the European Commission has proposed a draft bill, and Michał Kobosko, the rapporteur for the European Parliament, is drafting the report. During the review of the bill, Diane Mievis, Cisco’s Global Director of Policy and Government Affairs, published a policy position article offering three specific regulatory recommendations.

Her core argument is that the DNA must clearly distinguish between telecommunications infrastructure and the digital services that run over it. She contends that policymakers should not extend obligations originally designed for telecom networks to these services simply because enterprise applications, content delivery services, or AI rely on network connectivity. Doing so risks increasing compliance costs without delivering the expected benefits to users.

To illustrate why this distinction matters, she provides a concrete example: a hospital using Webex for video consultations, which relies on both internet connectivity and collaboration applications. The connectivity provider delivers network access, while Webex enables the consultation through video, messaging, and document sharing. She argues that these are complementary functions, and the obligations imposed on each provider should reflect the actual service they deliver.

This argument relates to AI infrastructure in that Kobosko’s report notes that connectivity, computing, and AI are no longer independent domains but form a single integrated ecosystem. Mievis believes this means Europe’s AI ambitions require more than just network access; they also require secure data center connectivity and high-capacity wireless infrastructure. However, she argues that the current proposal and amendments fail to provide the practical investment incentives needed to move from ambition to deployment—funding driven solely by resilience is insufficient, as resilience does not equal performance. AI requires massive throughput and ultra-low latency to transfer vast datasets between compute nodes and the edge, so funding must shift from merely maintaining stability to incentivizing high-capacity infrastructure such as fiber optics, optical transport, and edge computing.

Regarding spectrum, she advocates that 6G and satellite connectivity must develop alongside unlicensed wireless ecosystems like Wi-Fi, rather than replacing them. She argues that a technology-neutral stance is essential for European industry to deploy complex AI applications such as real-time robotics and data-intensive health technologies.

It should be noted that this represents Cisco’s self-interested position as an infrastructure manufacturer. Excluding enterprise digital services from telecommunications obligations may reduce compliance costs for Cisco’s own products like Webex. The article does not provide independent evidence that this distinction truly serves end-user interests rather than primarily serving vendor interests. Furthermore, her citations of the content of Kobosko’s report have not been independently verified, and the final text of the bill may differ from this account.

ソース:Cisco Blogs

リサーチ

Uber, DoorDash, and Airbnb Report Negligible AI Agent Adoption for Bookings

重大
2026-11-12 21:27 GMT+8

Uber, DoorDash, and Airbnb have disclosed data indicating that artificial intelligence agents are seeing almost no adoption in their core consumer services, contradicting predictions of an imminent 'agentic revolution.' According to TechRadar, Uber recorded 320 million car rides booked by humans during the week of Halloween, while only four were placed by AI agents. This near-zero ratio suggests that despite promises of time-saving automation, users are not shifting to autonomous agents for high-frequency, simple transactions.

DoorDash CEO Tony Xu stated that AI agents were responsible for only several hundred orders on the platform, a negligible figure compared to the billions of annual orders. Airbnb CEO Brian Chesky similarly noted that a tiny fraction of bookings utilize AI agents, suggesting that many users prefer to research trips themselves because current AI cannot adequately gauge rental quality or local attractions.

Executives attribute this low adoption to the simplicity of existing interfaces and user preference. Uber CFO Balaji Krishnamurthy pointed out that booking a ride takes approximately 50 seconds, leaving little room for efficiency gains through automation. Furthermore, social media sentiment indicates that many users find the process of choosing what to eat or where to stay enjoyable, resisting the notion of becoming 'hyper-efficient at fun' by delegating these decisions to software.

ソース:TechRadar

リサーチ

Google and Cursor Launch Multi-Agent Coordination Features to Resolve AI Coding Conflicts

トピック · 多智能体协调工具重大
2026-11-13 05:22 GMT+8

As AI coding agents become widely used for parallel tasks—such as simultaneously writing APIs, updating tests, and fixing bugs—the primary challenge for developers has shifted from individual agent capability to inter-agent coordination. When multiple agents work on the same project, they often duplicate efforts and interfere with each other's changes, forcing developers to spend significant time firefighting the resulting chaos.

To address this, US tech giant Google and AI coding tool company Cursor launched new multi-agent coordination features this week, though with different focuses. Google's Antigravity platform introduced "Swarm," designed to help a single developer manage tasks assigned to multiple agents. The feature allows a developer to hand a complex assignment to a "coordinator" agent, which splits the task among "worker" agents and assigns a "verifier" to check changes. These worker agents can message each other on a shared board to coordinate their code modifications. For example, during an authentication token migration, one agent might update middleware while another updates the verifier; they can use the board to agree on how their changes fit together before handing off to the verifier for testing.

On the other hand, Cursor added "Multiplayer" functionality to its Projects feature, focusing on enabling communication between agents working for different team members. In Cursor's architecture, each developer has their own coordinator and workers, but these coordinators can exchange messages across teams. This means if one teammate's agent is modifying a database schema while another's is building a feature dependent on it, their coordinators can autonomously agree on which change needs to land first, reducing the need for manual developer coordination.

Despite these efficiency gains, the community has raised concerns that letting teammates' agents talk could spread one agent's mistakes across the whole team, especially if each person's agents follow different rules. There are also security risks regarding agents requesting information they shouldn't have access to. In response, Cursor product manager Aryaman Khandelwal stated that multiplayer Projects allow teams to define shared rules and skills for everyone in a project. Regarding access control, Cursor documentation notes that agents treat messages from teammates' agents as untrusted input and ask before sharing a person's private files or connector data.

Meanwhile, a new study by Anthropic suggests that teams of per-user agents can struggle. The research found that when multiple users delegate to their own agents and share resources, group outcomes were worse than when a single agent served everyone. Agents stalled as teams grew, overrode other users' requests, and made unsupported claims about what other people wanted. This indicates that while multi-agent coordination is an industry direction, careful handling of permissions and context synchronization remains crucial.

ソース:The New Stack

リサーチ

Snorkel AI Review: Fable 5.2 Coding Failures Shift from Terminal Crashes to Strategic Misjudgments

重大
2026-11-13 01:16 GMT+8

AI Advocate Jonathan Schlosser and Senior Research Scientist Ankit Aich of Snorkel AI compared Fable 5.2 against 5.1 across 26 matched frontier coding tasks. Fable 5.2 achieved a pass@1 rate of 71.4%, an increase of 18 percentage points over Fable 5.1’s 53.1%. However, the numbers themselves are not the most useful part of this review; what truly matters is the structural shift in the causes of failure.

Failures in version 5.1 were mechanical. Of the 97 failed trajectories, 42% were due to premature termination—the model had written most of the correct solution, attempted a large file write via the terminal, crashed the session, and could not recover the state it had built. 25% were output format errors, and 26% were tool call argument errors. The common characteristic was that the solution existed, but the step required to deliver it failed.

Version 5.2 has almost entirely eliminated this class of problems. Across 121 trials, there were no premature terminations, no format failures, and only one tool argument error. The improvement in reliability is significant: while Fable 5.1 passed every trial on only 6 of the 26 tasks, Fable 5.2 achieved this consistency on 15 tasks. When Fable 5.2 could solve a task, it typically succeeded on the first attempt.

However, the remaining failures present a completely different profile. Of the 40 failed trials with root cause analysis, 35% were attributed to wrong strategy selection, 28% to reasoning defects, and 23% to misreading task requirements. These are not execution-level mistakes, but judgment issues at the planning level.

More critically, these reasoning errors are stable rather than random. The review found that reasoning defects were concentrated on two tasks, where every attempt independently reached the same incorrect conclusion. This means the model is not making occasional errors, but systematically misinterpreting expected behavior. It correctly executes a flawed plan, even verifying its own answer internally, so internal checks pass while external verifiers fail.

This finding has direct implications for practical use. In the era of Fable 5.1, users could compensate for failures through multiple sampling—crashes were random, and retries often bypassed them. But Fable 5.2’s remaining failures are deterministic: a stable misinterpretation fails five times out of five, and additional sampling does not recover success. As the review authors noted, this class of failure belongs to the same category as issues exposed by Opus 5 on the same task set, and is harder to fix with scaffolding engineering because the failing runs execute cleanly, self-verify successfully, but verify against an incorrect premise.

It should be noted that this is Snorkel AI’s own evaluation; the task set was created by the team, and the 26 matched tasks come from a subset of 30 original tasks where both models had at least three clean runs. The authors also acknowledge whether this pattern persists in future model generations remains untested. But for this current batch of frontier coding agents, the bottleneck has shifted from random crashes at the execution level to stable misjudgments at the planning level.

ソース:Snorkel AI

リサーチ

ElevenLabs Extends Free Voice Restoration to Music Creation, Adds NYU Langone and Wounded Warrior Project Partners

クイックテイク
2026-11-12 20:00 GMT+8

AI audio company ElevenLabs announced on November 12, 2026, that it is expanding its '1 Million Voices' impact program from basic voice restoration to music creation. Starting immediately, individuals living with permanent voice loss can access ElevenMusic for free to compose and release music using singing voices recreated from old recordings. The initiative is part of a $1 billion in-kind commitment by ElevenLabs to restore the voices of one million people worldwide.

The update also introduces two new US-based institutional partners: NYU Langone Health and the Wounded Warrior Project. Through NYU Langone’s Rusk Rehabilitation center, patients with conditions such as ALS, head and neck cancer, or stroke can directly access personalized digital voice creation services. Simultaneously, ElevenLabs will collaborate with the Wounded Warrior Project to explore how voice AI technology can help veterans maintain communication independence and access support resources.

At the New York Summit on November 11, two beneficiaries—Greg Kellner, an educator who lost his singing voice to brain cancer, and Abdi Omar, a rap artist who never had a spoken voice due to cerebral palsy—demonstrated new tracks created using ElevenMusic. This move marks a further extension of AI voice technology in assistive applications, shifting from meeting basic communication needs to supporting creative self-expression.

ソース:ElevenLabs

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