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

2026-11-0585 投稿

更新

Dropbox Engineering Team Upgrades AI Coding Agents to Task-Level Delegation, Completing 53 Legacy Test Cleanups in One Week

トピック · Dropbox重大
2026-11-05 22:00 GMT+8

The Dropbox engineering team faced specific time pressure: they needed to launch a new experience that had been brewing for three years in November, while complex integration work between the desktop and web versions remained far from complete. Previously, AI tools mainly helped engineers handle individual coding steps—completing a snippet of code or explaining an error message. For this project, they advanced their usage by handing over clearly defined tasks entirely to coding agents, with engineers reviewing the results.

This shift did not happen gradually. During the May Hackathon, the team used AI to assist in building a prototype, exploring two paths for bringing the web experience to the desktop: one involved embedding existing web pages into desktop windows using the Electron framework, and the other involved importing selected web components directly into the desktop interface. AI helped them track dependencies, identify platform-specific assumptions, and test integration schemes. The prototype gave product and design teams something tangible to try, rather than leaving it at paper discussions.

What truly demonstrated the effect of task-level delegation was a technical debt item shelved for two quarters: automated tests for the legacy file list required substantial engineering time to update and had been left untouched. The team broke this cleanup work down into 53 tasks and handed them to coding agents. The agents made changes according to a written plan agreed upon with engineers beforehand, ran relevant builds and tests locally, recorded results, and then submitted them for engineers to review whether to accept the code. All 53 tasks were completed within one week.

The key constraint of this process is that engineers remain responsible for the results. Before each task begins, the engineer and agent jointly confirm a written plan; after the agent completes its work, the engineer reviews the code and verifies that behavior matches expectations. Instruction files stored in the codebase explain the rules agents must follow, providing engineers with a repeatable way to delegate work rather than letting agents decide on their own.

The same pattern was applied to accessibility compliance work. During the launch preparation phase, the team delivered 22 accessibility features assisted by AI development, covering the Home page, All Files, file previews, and Photos modules. These included announcements for screen reader users to track carousel changes, fixes for how keyboard focus moves between controls, and layout adjustments for pages at 200% to 400% zoom. Coding agents also took on writing automated tests for five core modules of the web application, including the file browser users use to browse content.

A notable detail is the agents' ability to proactively discover problems during development. In one instance, an agent found that deleted items in chat history remained visible until the page was refreshed during a trial run; engineers used this to fix the underlying synchronization issue and added tests. This allowed human reviewers to spend more time judging whether the experience was reasonable, while functional issues were covered by automated checks.

ソース:dropbox.tech

リサーチ

Investing in AI Author Rob May Proposes 'Intent Share': ChatGPT Takes Question Volume, But Google Still Holds 85% of Transactional Queries

トピック · Rob May意图份额论重大
2026-11-05 19:50 GMT+8

ChatGPT's query volume is growing at a significant pace—independent data shows it handles 19% of global digital queries, while OpenAI itself claims it reaches 20% of Google's search volume. However, during the same period, Google sent outbound links to the open web at a rate roughly 190 times that of ChatGPT: Google accounted for about 40% of all web referral traffic, whereas ChatGPT accounted for only 0.21%.

This massive disparity is the starting point for this article's core observation. Investing in AI author Rob May explains it as a fundamental design difference between two products: Google is a router, whose job is to send you elsewhere—preferably somewhere an advertiser paid for; ChatGPT is an answer engine, whose job is to keep you in the conversation until the question is resolved. Data supports this: the average ChatGPT session lasts 13 minutes and 47 seconds, compared to 6 minutes and 4 seconds on Google. People aren't leaving because the product is designed so they don't have to.

May argues that the correct analytical lens is not 'who answered how many questions,' but 'who answered questions with money attached.' He calls this concept 'Intent Share.' Under this framework, Google still holds an 85.4% share of transactional queries and a 93% share of navigational and local queries. Large language models dominate generative and task-based work (with about a 66% share) and complex research and synthesis.

What does this mean? Google has indeed lost the top of the funnel—its share of all online discovery fell from over 89% at the end of 2022 to approximately 58% by the end of 2025, a trend May believes will not reverse. But the bottom of the funnel—queries directly connecting purchase intent and navigational behavior—has been hardly affected. For investors, the key question is no longer whether ChatGPT is stealing Google's searches, but whether it is stealing the searches that make Google money.

Boundaries to note: May disclosed in the article that he has long held Google stock. He mentioned a 'serious bear case scenario' and INFORMS shopping conversion data, but these contents are behind a paywall and the read portion cannot be displayed. The category-specific percentages of 'Intent Share' are an analytical framework based on cited data by the author, not independently verified market measurements.

ソース:investinginai.substack.com

リサーチ

YC Startup Discovered Materials Puts $1M Behind Five Unsolved Materials Science Problems to Test AI

重大
2026-11-05 21:00 GMT+8

Discovered Materials, a San Francisco-based startup that uses AI models to discover new substances for advanced chips, announced on November 5 a $1 million prize challenge for AI models capable of solving five critical unsolved problems in materials science. The company, part of Y Combinator’s spring 2026 batch, aims to address the gap between AI's success in mathematics and its limited progress in complex physical systems.

While frontier AI models have solved outstanding math problems and made strides in life sciences, they have yet to truly plumb the complex physics underlying advanced materials. “It’s more difficult to specify problems clearly in materials science because it’s not like math, where you have a very specific definition and can get to a yes or no,” said Advaith Sridhar, co-founder of Discovered Materials. To define these challenges, the company recruited an advisory group of academics from leading global universities, including Stanford professor Felipe Jornada.

The five problems include developing a theory to explain cuprate superconductivity, creating numerical methods to predict silicon-silicon dioxide interface structures, mapping grain structure to thermal properties, predicting how light changes material behavior, and improving Density Functional Theory (DFT). Sridhar stated that the goal is to establish these problems as “goalposts” for the field, regardless of who solves them—whether it be AI, professors, or other researchers.

Before announcing the challenge, Discovered Materials tested frontier lab models against these problems, giving AI agents internet access and resources to work autonomously for days. The models spent several days on each problem without making much headway. However, Sridhar noted that tracking efforts to solve these problems will show “whether the models are climbing in this domain,” providing a clearer metric for AI's evolving capabilities in scientific discovery.

ソース:TechCrunch

リサーチ

IMF Middle East and Central Asia Department Director: After Eight Months of Gulf Conflict, Regional Resilience Was Earned

トピック · MENA区域韧性重大
2026-11-05 20:19 GMT+8

At the IMF-World Bank Annual Meetings in Bangkok in October 2026, a recurring question was: How is the Middle East and North Africa economy still operating eight months after the Gulf conflict? Is this resilience or luck? The answer from Jihad Azour, Director of the International Monetary Fund's Middle East and Central Asia Department, is: Resilience is earned, not luck.

Hostilities erupted in the Gulf at the end of February this year, triggering the largest oil and energy shock in history. At the time, many feared the regional economies would be crushed—shipping lanes closed, energy prices soared, and vulnerable economies struggled to survive. Eight months later, the worst expectations have not materialized. Activity in some economies has slowed significantly, but there has been no collapse; output in many economies continues to expand.

However, the IMF's latest Regional Economic Outlook has lowered its 2026 MENA output forecast to a contraction of 1.3%, 2.7 percentage points lower than the April forecast. This headline figure masks huge differences among more than 20 countries within the region.

Several Gulf economies face deep contractions this year due to disruptions in energy production and trade. The IMF expects a strong rebound in 2027, provided the conflict subsides and trade gradually normalizes. But the rebound will not erase losses: Gulf output is projected to remain below pre-conflict expectations in the medium term.

Meanwhile, major oil-importing countries such as Egypt, Jordan, and Morocco have shown surprising resilience, with growth expected to exceed 4% this year. Remittances, tourism, investment, and favorable financial conditions have supported growth, while policies have cushioned households against higher food and energy costs. But pressures are mounting, and inflation is falling slower than spring forecasts predicted.

Azour's core judgment is: What determines how much room an economy has to maneuver during a shock is the institutional framework and buffers it built during good times. A finance minister's statement in Bangkok captured this: The policy frameworks and buffers built in good times determine how much room we have in bad times.

But managing this shock is only part of the task. Azour believes regional economies also need to prepare for the next crisis: diversifying trade and production to reduce reliance on narrow export structures and vulnerable transport routes; improving infrastructure and deepening regional integration to lower rerouting costs and strengthen supply chains; investing in renewable energy to limit exposure to volatile fuel markets. At the same time, Gulf economies are investing in data centers and digital infrastructure, turning digital technologies like AI into broader prosperity requires better education and skills, stronger governance and competition, and deeper links to global markets.

The experience of the past eight months shows that resilience is earned through adjustment, stronger policy frameworks, and buffers accumulated before the conflict. But resilience is a process. If the Middle East conflict persists and buffers are steadily depleted, today's delayed effects could become tomorrow's larger adjustments. In an era of recurring shocks, economies that prepare before the next disruption will not only be best placed to endure it, but also to emerge stronger from it.

ソース:semafor.com

リサーチ

Agility Robotics Goes Public via SPAC, Becoming First Pure-Play Humanoid Robot Company

重大
2026-11-05 20:00 GMT+8

Agility Robotics listed on the Nasdaq on November 5, 2026, following a reverse-merger with a special purpose acquisition company (SPAC). The transaction makes it the first publicly traded pure-play humanoid robotics company and raised $465 million in new capital from retail investors and private funding.

Unlike competitors such as Tesla and Figure AI, Agility focuses on repetitive industrial material handling tasks. Its Digit robots have been deployed at customers including Schaeffler, GXO, and Toyota Motor Manufacturing Canada. However, the company's financials reveal significant challenges: revenue was only $174,000 in the first half of 2026, against losses of $362 million during the same period.

CEO Peggy Johnson stated that the new capital will support the development and deployment of the next-generation Digit 5 robot, which aims for safe human-robot collaboration. The company has secured $300 million in planned contracts from 30 customers for this new model. Despite questions about unit economics and technical maturity, Agility's listing provides a crucial secondary-market valuation benchmark for the broader humanoid robotics investment landscape.

ソース:TechCrunch

リサーチ

Construction Robotics Venture Funding Hits $1.6B, Tripling Previous Peak

重大
2026-11-05 20:00 GMT+8

According to Crunchbase data, companies spanning construction and robotics have raised approximately $1.6 billion in venture funding this year through October 29. This figure is more than triple the previous high of $506 million set in 2021 and exceeds the combined total of the past five years.

The surge was driven by much larger bets rather than an increase in deal volume. Three companies—Bedrock Robotics, Icon, and Gravis Robotics—accounted for over $1.3 billion, or about 83%, of the total raised. Bedrock Robotics was the largest recipient, raising $870 million across two rounds, including a Series C in October that valued the company at $4.1 billion.

Investors are betting on automation to address acute labor shortages in the U.S. construction industry. A recent survey by the Associated General Contractors of America and NCCER found that labor shortages remain the biggest cause of project delays, with the AI-driven data center building boom further straining the workforce supply.

Exit activity has been dominated by acquisitions. Procore's approximately $845 million cash acquisition of DroneDeploy was one of the largest disclosed deals. No known IPOs occurred among companies focused specifically on construction robotics this year.

ソース:Crunchbase News

リサーチ

China calls on US and EU to reconsider tariffs, offers increased farm purchases in exchange for cuts

トピック · 中美欧关税谈判重大
2026-11-05 19:44 GMT+8

According to Semafor, the Chinese government called on the United States and the European Union to reconsider their tariffs on Chinese goods on November 5, 2026. This statement comes as China's exports reach record highs.

Beijing indicated a willingness to increase purchases of US farm goods, but only if both sides agree to significantly slash tariffs. Although China recently restarted American soybean imports—a priority for Washington—current levels remain far below those seen before last year's trade war.

Meanwhile, Chinese officials expressed concern over new EU measures targeting low-value imports largely from retailers such as Shein and Temu, calling on Brussels to assess their impact. Countries globally remain fearful that China’s surging manufacturing overcapacity is hollowing out their industries.

ソース:semafor.com

リサーチ

AI Cloud Provider Lambda Announces Plans to Build Data Center Campus in Alabama, Investing $33.1 Billion

トピック · Lambda重大
2026-11-04 23:02 GMT+8

US-based AI cloud company Lambda announced on November 4, 2026, plans to build a large-scale data center campus named "Red Clay Technology Center Campus" in Lowndes County, Alabama. The project, readied for development by Cloverleaf Infrastructure, represents an estimated $33.1 billion investment.

According to Lambda's announcement, the data center campus is expected to create more than 1,000 construction jobs during development and support at least 225 full-time roles upon operation. The project is projected to generate significant economic benefits for local schools and communities, with more than $2.8 billion in local tax revenue through 2049, including over $1.6 billion for Lowndes County Public Schools. Combined state and local tax revenue is projected to reach $3.4 billion.

To address community concerns about infrastructure burdens, Lambda committed to paying the full cost of the power and grid infrastructure required to serve the facility. Through upfront financial commitments, these costs will not be shifted to Alabama Power's existing customers even if project plans change. Both Lambda and Alabama Power have signed the White House Ratepayer Protection Pledge. Additionally, the facility will deploy closed-loop cooling technology to reduce water use and plans to partner with N-Drip to support local farmers in conserving water.

The project is currently in the planning stage, with construction expected to begin in the first quarter of 2027 and take up to five years. Power supply remains pending regulatory approval. Stephen Balaban, Co-founder and CTO at Lambda, stated that the company is committed to building strong relationships with the local community and ensuring resident perspectives inform their approach.

ソース:lambda.com

リサーチ

US OCC Issues First Bank AI Framework, Classifying Autonomous Payment Agents as Highest Risk

トピック · OCC重大
2026-11-05 12:02 GMT+8

The U.S. Office of the Comptroller of the Currency (OCC) issued its first supervisory framework for artificial intelligence in banking on November 4, 2026. The guidance directs examiners to focus less on an AI system's technical sophistication and more on the consequences of its errors and the degree of autonomy it possesses.

A key feature of the framework is a four-tier illustrative risk classification. Fully autonomous payment agents, real-time credit decision systems, and fraud transaction unfreezing fall into the highest category (Tier 1). Lower tiers include AI-assisted underwriting and basic internal productivity tools. The OCC notes that a simple agent authorized to initiate transactions may present greater risk than a more advanced model limited to writing internal documents.

The framework emphasizes that human oversight cannot be merely formal. It distinguishes between 'Human-in-the-Loop' (approval before execution) and 'Human-on-the-Loop' (continuous monitoring with intervention rights). To address the risk of AI agents continuing to execute harmful tasks without crashing, the OCC suggests safeguards such as rate limits, least privilege access, circuit breakers, and emergency kill switches.

The guidance also addresses external AI agents operated by customers or fintechs that connect through bank channels. It recommends measures including identity verification, delegated permissions, transaction limits, and dispute handling. The OCC stressed that this document does not create new regulatory requirements but helps integrate AI into existing operational, compliance, and cybersecurity risk categories.

ソース:TechOrange 科技報橘

リサーチ

Meta Acquires AI Coding Startup Conductor, Shutting Down Independent Multi-Agent Tool

トピック · Conductor重大
2026-11-05 06:20 GMT+8

US technology giant Meta Platforms Inc. has confirmed the acquisition of artificial intelligence coding startup Conductor (formally Melty Inc.). The deal was first reported by Axios, and Conductor subsequently verified its move to Meta in a post on its website, though financial terms were not disclosed.

Conductor’s core product is a desktop application that allows developers to run several AI coding agents at once—such as Anthropic’s Claude Code, OpenAI’s Codex, and Cursor—each working in an isolated copy of the codebase before human review and merging. The company stated that more than 100,000 people use its platform.

With the acquisition complete, Conductor’s standalone services will be phased out. New sign-ups and downloads halted on November 4, and paid features are now free for existing subscribers. The cloud service, “Conductor Cloud,” will become read-only on November 13 and shut down entirely on November 20, after which user data will be deleted. Local Mac app copies will continue to function until early 2027 but will no longer receive new models or critical bug fixes.

Meta engineers were already heavy users of Conductor prior to the deal. When Conductor announced its Series A funding in March, it named Meta, Google LLC, and Spotify Technology SA among the companies with dozens of engineers using the product. People familiar with the deal told Axios that Meta values the Conductor team’s speed in shipping products and aims to leverage this capability to build its next generation of AI coding tools.

The acquisition follows a series of Meta model releases focused on coding. On October 15, Meta released Muse Spark Open, an open-weights version of its Muse Spark model aimed at coding and long-running agent work. Earlier, on October 12, it launched the more capable closed-source Muse Wonder models, integrated into its internal Muse Code terminal and API. On the same day as the Conductor news, rival OpenAI also announced plans to acquire developer tools startup Pierre Computer, highlighting a trend of major tech firms consolidating AI programming infrastructure through M&A.

ソース:SiliconANGLE

リサーチ

Bosch Sensortec CEO Stefan Finkbeiner: Edge AI Keeps Raw Data on Device, Privacy Ensured by Architecture

トピック · 博世边缘AI隐私架构重大
2026-11-05 17:00 GMT+8

In a Leaders Talk interview with eeNews Europe, Bosch Sensortec CEO Stefan Finkbeiner presented a core assessment regarding edge AI and privacy: the advantage of privacy protection stems from architectural design rather than transmitting data to cloud-based processing.

His argument is that when AI algorithms run directly on sensor silicon, physical signals are processed locally and discarded; the device outputs only the specific conclusion requested by the user—for example, "air quality is moderate" or "step count updated"—without raw data ever leaving the device. This contrasts sharply with traditional architectures: in conventional designs, raw sensor data is continuously transmitted to external clouds for algorithmic processing, a process that inherently creates privacy vulnerabilities. Finkbeiner argues that edge AI returns data sovereignty to the user, making on-chip processing the strongest technical safeguard for privacy.

The physical prerequisites supporting this architecture are breakthroughs in two sensing technologies. TMR (Tunnel Magnetoresistance) technology provides ultra-high precision, extremely low noise, and ultra-low power magnetic sensing, enabling indoor navigation in GPS-denied scenarios and head tracking in AR/VR headsets. ASICAP capacitive sensing pushes miniaturization and signal sensitivity to new boundaries, allowing wearable devices to detect micro-gestures and acoustic details with minimal energy consumption. Finkbeiner concludes that when these physical breakthroughs are combined with on-sensor AI processing, real-time response, near-zero latency, battery life, and privacy can be simultaneously guaranteed. He states that all new products feature microcontrollers equipped with edge AI capabilities.

This narrative originates from a manufacturer's CEO describing their own product roadmap, constituting an attributed argument rather than independent verification. Specific performance metrics for TMR and ASICAP, the quantitative definition of "near-zero latency," and the coverage scope of all new products have not been confirmed by third-party testing. While Finkbeiner’s privacy architecture logic—local processing, discarding raw data, and outputting only conclusions—is technically clear, whether actual products strictly adhere to this design depends on specific implementation rather than visionary statements.

However, behind this architectural choice lies a notable design philosophy. Finkbeiner characterizes Bosch Sensortec’s innovation culture as "constructive naivety": engineers are encouraged to try new things and take calculated risks, rather than merely manufacturing components according to customer specification sheets. He believes that in consumer electronics, doing only what is listed in specifications means becoming a commodity supplier, which is already too late. This culture provides fertile ground for the development of edge AI sensors—not waiting for customers to define needs, but proactively demonstrating what can be achieved by combining physical sensing with on-chip intelligence.

ソース:eeNews Europe

リサーチ

GitHub Announces General Availability of Dependabot Support for Agent Package Manager

トピック · GitHub代理包管理器クイックテイク
2026-11-04 04:24 GMT+8

On November 3, 2026, GitHub announced that its dependency management tool, Dependabot, now officially supports Agent Package Manager (APM). This update allows developers to use Dependabot to automatically create pull requests (PRs) for updating AI agent components managed by APM, including skills, prompts, and instructions.

APM is a tool designed to manage dependencies for AI agents within projects, allowing developers to specify sources and versions for required skills in an apm.yml file. With Dependabot's APM support enabled, the system checks Git dependencies declared in apm.yml and generates update PRs when new versions are available. If an apm.lock.yaml file exists in the project, Dependabot also regenerates this lock file to ensure consistent versions of AI agent components across teams.

The feature is currently generally available on github.com for all plans and is expected to be included in GitHub Enterprise Server 3.24. Note that this support covers version updates only and does not include security vulnerability alerts or fixes.

ソース:The GitHub Blog

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

Integration of cross-platform shared components also benefited from this approach. Web and desktop each use different build tools, login mechanisms, and data retrieval methods. Engineers need to determine which code can be reused and which parts must be handled separately on each end. AI helped them track connections between components, extract reusable parts, write adapter code (connecting shared features to each app's services), and corresponding tests. Engineers reviewed the final code to confirm it worked as intended.

The team also emphasized that customer feedback and engineering review together determined what could be rolled out to more users. All eight initial design partners chose to keep the new experience enabled. In broader gray-scale testing, they tracked usage of familiar features such as upload, preview, search, and sharing, as well as whether AI actions succeeded and whether users turned off the new experience. These signals helped determine priorities, while the coding agent workflow gave engineers a consistent way to implement and verify fixes.