Sign In / Up | Advertise | View Online

NerdNews

November 03, 2026

News & Trends
Google and Meta Open-source TorchTPU for Native PyTorch Support on Google TPUs

Google and Meta have open-sourced TorchTPU, a project that provides native PyTorch support for Google TPUs, enabling PyTorch workloads to run on TPU hardware with minimal code modifications. TorchTPU supports eager and compiled execution, distributed computing, and hardware-specific optimizations.

Uber Builds MCP Gateway for AI Agents

Uber has built an MCP Gateway to standardize how AI agents discover and invoke internal services, supporting over 800 servers and 5,000 tools. The platform provides a common layer for discovery, protocol translation, authorization, and execution, and enforces authorization and data redaction.

Anthropic Claims to Be Closing in on Recursive Self-Improvement

Anthropic claims to be close to achieving recursive self-improvement (RSI), a point where a computer system can improve itself autonomously. The company's latest SEC filing warns investors of the potential risks and benefits of RSI, which could lead to unprecedented levels of security and autonomy.

Google takes Dreambeans to India

Google has launched its experimental AI app Dreambeans in India, its first market outside the U.S. The app generates personalized visual stories drawn from users' digital activity and is available on Android and iOS with support for 13 local languages.

OpenAI Merges ChatGPT's Chat and Work Tabs

OpenAI is merging the Chat and Work tabs in ChatGPT to simplify the user experience. This change will allow users to explore ideas and work on projects in one place, with access to files and plugins. The update will initially be available to select users for feedback before a wider rollout.

Options & Tutorials
Chinese Hackers Use ARTEX AI in Automated Attacks

Chinese hackers used ARTEX AI to attack over 200 organizations across 14 countries, targeting government ministries, telecommunications providers, and pharmaceutical firms. The AI agents performed reconnaissance, exploited vulnerabilities, and obtained credentials, with the operators intervening at critical points.

Claude's new rules against game cloning

Anthropic's Claude model has introduced new rules to prevent game cloning, but it's unlikely to stop the AI copying boom. The rules aim to prevent users from recreating specific existing games, but eager game copiers can shift to other models without restrictions.

Hundreds of Mundane AI Apps Are Hiding Sexually Abusive Deepfake Features

Researchers have discovered a network of over 400 AI apps and websites that appear to be mundane tools but secretly offer sexually abusive deepfake features, including explicit images and videos. These apps, often linked to Chinese-language developers, have been found on Apple and Google's app stores and have been downloaded hundreds of thousands of times.

42AI Comes Out of Stealth with $62.5M

42AI, an Israeli startup, has developed a 'behavioral world model' that can predict human behavior. The model uses psychological research, neuroscience, and genetics to analyze customer interactions and predict decisions. With $62.5 million in seed funding, 42AI is working with financial brokerages to manage liquidity and hedging exposure, and is exploring applications in gaming, fraud prevention, and cybersecurity.

Websites Blocking AI Agents

Some websites are blocking AI agents from accessing their services due to security concerns and commercial reasons. Amazon and eBay have blocked certain AI agents, while others like Walmart have encountered issues with human-verification prompts. Existing bot protections can also get in the way, interrupting automated shopping sessions.

Launches & Tools
Credential Layer Security in Practice

GitGuardian's platform discovers credentials, provides context, and supports remediation and prevention, helping security teams manage credential risk and reduce exposure.

Y Combinator's Garry Tan spends up to $200K a year on AI tokens

Y Combinator CEO Garry Tan spends $100-200K annually on AI tokens, using agents for tasks like research and coding, and believes personal AI agents will become mainstream if costs decrease

Trajectory raises $45M

Trajectory, a startup, has raised $45 million to automate AI model training, making it accessible to non-technical experts. The company's platform allows businesses to start with an open-source AI model and connect it to their applications, collecting feedback to retrain the model and improve its performance at specialized tasks.

Omen AI Acquires Aquatic Labs

Omen AI has acquired Aquatic Labs in a $16 million deal, combining their technologies to create a real-time sensor for monitoring cooling fluids in data centers. The sensor aims to reduce water usage and costs for data center operators by minimizing downtime and optimizing fluid systems.

Gemini Adds Direct File Editing Inside Chat

Gemini has introduced a new feature that enables users to edit files directly within the chat interface. This feature combines file previews, manual edits, and targeted AI changes. It supports various file formats, including PDFs, documents, spreadsheets, and presentations, with PDF editing available on both web and mobile platforms.

Quick Links
TechCrunch Mobility: A new, bigger AV boom is here

A new wave of autonomous vehicle startups is emerging, fueled by technological breakthroughs such as AI and foundational models. Trucks VC has seen 81 new AV startups pitch this year, with many focusing on robotaxi businesses and competing with Waymo. The barrier to entry is lower than ever before, with startups deploying in a few months and with less than $1 million raised.

The New York Times sues Brave web browser for AI scraping and bypassing paywalls

The New York Times and other publications have filed a lawsuit against Brave Software, the company behind the Brave web browser, for allegedly scraping and providing copyrighted content to AI companies without permission. Brave's 'Paywall blocker' feature is also accused of allowing users to bypass paywalls without consent.

Apple's 'Ultra' smart home robot might be delayed until 2028

Apple's rumored smart home robot, codenamed J595, might be delayed until 2028 due to software issues. The robot, potentially called 'HomeView Ultra', could cost over $1,000 and features a 9-inch display, cameras, and sensors for automating home tasks and following users during FaceTime calls.

Apple Says Even Higher Memory Prices and Siri AI Push Will Weigh On Profits

Apple's CEO John Ternus warns of worsening memory prices and their impact on profits. The company expects even higher memory costs in the next quarter, partly offset by price hikes and lower costs of other components. Additionally, Apple's push for Siri AI will also affect profit margins due to associated cloud costs.

FBI Attributes Binance Heist to North Korea's TraderTraitor

The FBI has attributed the recent $6bn Binance hack to North Korea's TraderTraitor group. The group, also known as Jade Sleet and UNC4899, has been linked to several other high-profile crypto heists. The FBI is urging crypto entities to block transactions related to the stolen assets, which are being laundered through decentralized finance products and instant swap services.

Share NerdNews

Share your affiliate link to get commission!

https://nerdnews.online/affiliate

Thanks for reading,
The NerdNews Team

Sign In / Up

If you dont want to receive future editions of NerdNews, Unsubscribe here.