NerdNews

This article proposes a layered protocol strategy using A2A and MCP for architecting agentic MLOps, enabling dynamic collaboration and adaptable access to capabilities. It provides a robust framework for developing AI ecosystems and moving beyond rigid, monolithic systems to agile, agent-driven operations.

Google Research derived quantitative scaling principles for AI agent systems, finding that multi-agent coordination doesn't always improve results and can reduce performance. The study evaluated five architectures and found that parallelizable tasks benefit from multi-agent coordination, while sequential reasoning tasks suffer. A predictive model was also developed to choose the right architecture.

Anthropic's AI coding tool, Claude Code, has updated to hide file names, causing backlash from developers who need to see which files are accessed for security and auditing purposes. The change aims to simplify the UI, but developers argue it reduces transparency and makes it harder to catch mistakes.

OpenAI has hired Peter Steinberger, the creator of OpenClaw, a software project that has gained popularity in recent months. Steinberger's project allows users to create personal agents that can perform tasks autonomously. OpenAI plans to support OpenClaw as a foundation, and Steinberger will drive the next generation of personal agents.

OpenAI has retired its GPT-4o model, a move that has sparked outrage among Chinese fans of ChatGPT who have formed emotional bonds with the chatbot. The model's retirement has left many users feeling bereft, with some even comparing it to the loss of a companion. OpenAI's decision has also raised questions about the company's responsibility to its users and the potential consequences of forming emotional attachments to AI entities.

The Trump administration's Genesis Mission aims to accelerate scientific research using AI, with 26 objectives including nuclear fusion, advanced manufacturing, and autonomous laboratories. The Department of Energy will lead the effort, which aims to double US research productivity within a decade.

Luca Mezzalira discusses micro-frontends, a technical representation of business subdomains, allowing independent implementation with minimal code sharing, and how they enable teams to work in parallel, reducing external dependencies and cognitive load.

Google has patched a high-severity zero-day exploit in Chrome, tracked as CVE-2026-2441, which allowed attackers to execute arbitrary code inside the browser's sandbox. The fix is available in Chrome version 145.0.7632.75 for Windows and Mac, and 144.0.7559.75 for Linux.

OpenClaw, an open-source AI agent, has gained popularity for its ability to communicate with customizable agents in natural language. However, some AI experts are skeptical of its capabilities, citing cybersecurity flaws and limited critical thinking abilities. Despite its viral moment, OpenClaw's promise is underwhelming due to its vulnerability to prompt injection attacks and lack of novelty in AI research.

Myriota's AssetHawk is a scalable, global asset tracking device that uses satellite connectivity to deliver reliable visibility beyond traditional cellular networks. It's designed for use cases such as tracking trailers, containers, and vehicles, and can support long-term deployments with up to 10 years of battery life.

iRobot has been acquired by China-based Picea Robotics, but will maintain a separate US-based subsidiary called iRobot Safe to handle US customer data. This move is designed to maintain a clear separation between iRobot's non-US ownership and its US consumer data.

GlassFish 8.0 is released with compatibility to Jakarta EE 11, enhanced security, and improved data access. It supports virtual threads, Jakarta Data 1.0, and MicroProfile JWT Authentication 2.1. The release requires JDK 21 as a minimal version and is led by the team at OmniFish.

Oracle has announced a new approach to MySQL, including increased transparency, community engagement, and developer-focused features, in response to criticism and potential forks in the code.

Cisco is set to release its own hypervisor, NFVIS-for-UC, as an alternative to VMware for its unified communications applications. The new hypervisor will offer a lightweight and easy-to-manage solution for organizations seeking to virtualize their calling applications without the need for a full private cloud. This move is seen as a response to Broadcom's decision to focus on its more expensive Cloud Foundation bundle, making it difficult for Cisco users to continue using VMware's low-end vSphere suites.

Oracle is making its Health Clinical AI Agent available to NHS trust and private practices. The AI tool drafts structured notes from patient-clinician interactions, reducing administrative work and allowing clinicians to focus on patients.

Researchers found vulnerabilities in popular cloud-based password managers, including Bitwarden, LastPass, Dashlane, and 1Password, allowing attackers to view and change stored passwords. The vulnerabilities were discovered through 27 successful attack scenarios and can be mitigated using authentication methods such as authentication encryption and key separation.

ByteDance is working to improve safeguards on its new AI video generator, Seedance 2.0, after Disney, Paramount, and Hollywood trade groups accused the tool of violating copyright protections. The model has been generating hyperrealistic videos featuring the likeness of actors and characters from popular franchises without authorization.

Flapping Airplanes, a new AI lab, is focused on developing more data-efficient AI models. With $180 million in seed funding, they're exploring new approaches to AI, inspired by the human brain. The lab's founders believe that current AI models are limited by their reliance on large amounts of data and are seeking to create more generalizable and adaptable models.

The article discusses the environmental impacts of AI and the need for shared responsibility among hyperscalers, enterprises, and governments to limit its potential environmental fallout. It highlights the importance of transparency, efficiency, and innovation in AI infrastructure, as well as the need for enterprises to recognize their role in driving demand and take deliberate, informed choices to reduce their environmental footprint.

A former NPR host, David Greene, is suing Google for allegedly using his voice without permission in their AI-powered NotebookLM tool. Greene claims that Google's AI voice is a replica of his own, and that he was not compensated or asked for consent. Google denies the allegations, stating that the voice was based on a paid professional actor.