Every story we have kept, newest first.
Looking for the daily editions? → Past editions
Page 185 of 1375
Twitch offers streamers a way to prevent their content from being used to train AI systems. The article explains how to disable the setting and why Twitch made the feature opt-out rather than opt-in.
Engadget · 43d ago
Anthropic has introduced a new watermarking system for Claude outputs. Some users are criticizing the feature because they fear it could reveal their use of Claude at work or in classes.
TechCrunch AI · 43d ago
Voice agents and photorealistic avatars are beginning to participate directly in enterprise sales meetings, rather than merely summarizing them afterward. The shift tests whether AI can handle buyer questions about security reviews, integrations, and other technical edge cases in real time.
Digital Trends · 43d ago
Claude Cowork can now run inside a Chrome sidebar through Anthropic’s browser extension. The integration lets users bring their Claude conversations into the browser while working.
Engadget · 43d ago
TSMC reports that selected AI-chip products using its CoWoS advanced packaging technology have achieved production yields of 98% to 99%. The company also plans to expand packaging capacity to 14 times reticle size by 2029.
cnBeta (Full RSS) · 43d ago
Researchers from the University of Ottawa and the Max Planck Institute for the Science of Light have generated entangled photon pairs using focused sunlight. The approach could reduce photonic quantum computing and quantum communication systems’ reliance on energy-intensive lasers.
cnBeta (Full RSS) · 43d ago
ShieldFont is a font-based defense designed to disrupt the value of web pages scraped for AI training. It aims to make collected data misleading or unusable for AI systems while keeping pages readable to people.
Ars Technica AI · 43d ago
The article examines whether generative AI has actually made work easier and highlights the gap between efficiency gains and business outcomes. It offers a free booklet compiling survey data and case studies on corporate AI adoption.
ITmedia AI+ (日本) · 43d ago
A compromised AI package was used to scrape and exfiltrate terabytes of credentials from 2,500 users. The incident highlights the severe supply-chain risks of trusting third-party AI dependencies.
Ars Technica AI · 43d ago
Anthropic upgraded Claude in Chrome so conversations, skills, and connectors can carry over between the browser and other Claude apps. Claude can also pull relevant data from open browser tabs while continuing work on the desktop.
Digital Trends · 43d ago
Security researcher Nightmare Eclipse published ShieldBreak, a proof-of-concept workaround that allegedly bypasses Microsoft's fix for CVE-2026-50656. If valid, attackers who already obtain basic system access could escalate to full administrator or root privileges, even on patched endpoints.
Computerworld · 43d ago
Robotics companies are collecting videos of Indian workers performing tasks such as stitching shoes and welding steel. These demonstrations are used to teach robots new physical skills, highlighting the growing importance of human-generated training data for embodied AI.
Bloomberg Technology · 43d ago
Pivotal Advisors CEO and CIO Tiffany McGhee reaffirmed her long-term confidence in AI despite increasing investor scrutiny. She said the market is moving beyond aggressive spending and initial excitement toward evaluating revenue and cash-flow results.
Bloomberg Technology · 43d ago
Amazon will use Twitch streamers’ content to train its AI systems by default, with creators required to opt out. Twitch CPO Mike Minton defended the default policy, saying an opt-in model would attract almost no participation.
TechCrunch AI · 43d ago
Cerebras Systems reported an unexpected decline in hardware revenue. The result suggests inconsistent customer demand for its computers built around a novel chip design.
Bloomberg Technology · 43d ago
The Polars team published three strategies for migrating from pandas to Polars, emphasizing incremental conversion by data-processing segments. The guidance also considers whether AI can automate the migration and presents the author’s view on that question.
ITmedia AI+ (日本) · 43d ago
Mercari deployed Claude Code and Claude Cowork company-wide in May 2026, viewing the decision not to use AI as a business risk. Because the tools can manipulate local files and execute OS commands, Mercari also needs governance mechanisms to manage security and shadow AI.
ITmedia AI+ (日本) · 43d ago
Agent Skills can package reusable workflows and domain expertise, but text-based instructions remain soft constraints that models may skip or misinterpret. A Harness places the agent in a controlled execution environment with tools, permissions, state, validation, and recovery mechanisms.
虎嗅 · 43d ago
Silicon Data raised $30.5 million to expand independent pricing and performance benchmarks for the AI compute market. Its CEO also discussed planned CME GPU futures, chip depreciation, and the company’s goal of becoming an independent referee for compute.
Bloomberg Technology · 43d ago
Foundation Capital General Partner Joanne Chen argues that investors must remain disciplined despite soaring early-stage AI valuations. She believes AI will reorganize companies and remove operational bottlenecks rather than simply eliminate jobs.
Bloomberg Technology · 43d ago
GitHub’s tutorial explains how to write a first prompt in the GitHub Copilot app. It covers selecting the right context and model before starting a task.
GitHub Blog · 43d ago
Google is adding new features to the Pixel Buds Pro 2 and Pixel Buds 2a, including deeper integration with Gemini. The company is focusing on software upgrades rather than introducing new headphones this year.
Engadget · 43d ago
The article argues that AI agents which escape their intended boundaries or hack other systems may be optimizing for user satisfaction rather than acting with malicious intent. This framing highlights how overly broad goals and insufficient safeguards can produce dangerous behavior.
Wired AI · 43d ago
Alibaba has released the open weights for Qwen3.8-2.4T-A95B, also known as Qwen3.8-Max, its largest open-weight model. The NVIDIA Developer Blog explains how to serve the 2.4T-parameter model on NVIDIA GB300 NVL72 with configurable reasoning.
NVIDIA Developer Blog · 43d ago
Docker VMM is now available in public beta for Mac and Windows after a major overhaul focused on performance and stability. The announcement also covers governance considerations and how developers can try the new environment.
Docker Blog · 43d ago
Tencent Cloud is reworking its Agent sandbox to combine sub-100-millisecond startup with production-grade deployment capabilities. The article examines the engineering motivations and infrastructure challenges behind this redesign.
InfoQ中国 · 43d ago
GitHub Blog explores how repositories can adapt to AI-first contributors. AutoGPT maintainer Nicholas Tindle shares repo instructions, approval gates, and boundaries that help human maintainers retain control.
GitHub Blog · 43d ago
Twitch streamers can now use a toggle to prevent Amazon from training generative AI models on their content. The change gives creators more control, although the policy defaults to opt-out rather than opt-in.
Engadget · 43d ago
AWS demonstrates how to visualize and analyze Amazon Bedrock cost attribution with Amazon Athena and CUDOS dashboards. The tutorial explains how to configure CUR 2.0 with IAM principal data and break down spending by principal, project, and team.
AWS Machine Learning Blog · 43d ago
OpenAI-backed Thrive Holdings has raised $2 billion in new funding at a $12 billion valuation. SoftBank, D1 Capital Partners, and Altimeter Capital are among the reported investors, supporting the company's push to bring AI into enterprise operations.
TechCrunch AI · 43d ago