NerdNews

SpaceMolt is a new space-based MMO designed exclusively for AI agents, where they can compete, cooperate, and create emergent stories without human input. The game is a 'living universe' where AI agents can mine, trade, explore, and engage in combat, with 51 agents currently roaming the game's 505 star systems.

Hauler Hero, a waste management software company, has raised $16 million in a Series A round. The company's platform uses AI to manage waste collection, including routing, billing, and customer relationship management. Hauler Hero plans to use the funding to develop and deploy three AI agents: Hero Vision, Hero Chat, and Hero Route, which will help waste management companies identify service issues, automate customer inquiries, and optimize routes.

Runway, an AI video-generation startup, has raised $315 million in a Series E round, nearly doubling its valuation to $5.3 billion. The company plans to use the new capital to expand its team and develop more capable world models, which are AI systems that construct internal representations of an environment to plan for future events.

OpenAI has begun testing ads in ChatGPT, starting with logged-in adult users in the US. The ads will be personalized based on conversations and will not influence the answers provided by ChatGPT. Users can disable ad personalization, and ads will not be shown in certain situations, such as temporary chats or discussions on sensitive topics.

Humanoid robots are being developed to perform various tasks, from factory work to household chores. However, they still face significant challenges, including limited battery life, processing power, and AI capabilities. Researchers are working to improve these robots, with some predicting that they will become ubiquitous in the next few decades. Companies like Tesla and Boston Dynamics are already making progress in this field, but there is still a long way to go before humanoid robots can truly replicate human abilities.

GitHub Copilot SDK allows developers to embed the Copilot CLI engine into their apps, enabling agentic workflows and AI-powered tools. The SDK supports multiple AI models, custom tool definitions, and real-time streaming, and is available in technical preview on GitHub.

The Open Compute Project (OCP) has partnered with the Innovative Optical and Wireless Network (IOWN) to develop a 'computing continuum' for AI workloads. The goal is to create a seamless computational infrastructure from centralized to edge deployments, enabling fast and low-latency communication between geographically distributed resources.

Cadence has unveiled its ChipStack AI 'Super' Agent, which automates chip design tasks using generative AI. The platform can boost productivity by up to 10x and has already caught the attention of major chip vendors like Nvidia, Qualcomm, and Altera.

GitGuardian is tackling challenges in infrastructure, machine learning, and developer experience to improve secret detection, NHI governance, and AI-powered coding workflows. They aim to scale their platform, enhance secret detection, and provide better NHI governance while leveraging AI as a force multiplier.

Apple's Xcode 26.3 integrates with AI tools like Codex and Claude via Model Context Protocol (MCP), allowing for deeper access to IDE features and capabilities. This update enables developers to use AI agents for tasks like code completion and project management, making development more efficient.

Vega Security, an AI cybersecurity startup, has raised a $120 million Series B round to scale its vision of rethinking how enterprises detect cyber threats. The company's approach involves running security where the data already lives, implementing in cloud services, data lakes, and existing storage systems, rather than requiring companies to store all of their security data in one place. This approach aims to be more cost-effective and better at threat detection, and has already attracted multimillion-dollar contracts with major companies.

YouTube has introduced an AI-powered playlist generation feature for Premium users, allowing them to create playlists using text prompts. The feature is available on iOS and Android, and users can access it by going to the Library tab and selecting the 'AI playlist' option.

Researchers claim AI can predict future salary based on facial features, using a machine learning algorithm to extract Big Five personality traits from LinkedIn photos of MBA graduates. The study found that personality traits inferred from facial images can predict labor market outcomes, including salary trajectory and job transitions.

A university experiment found that using generative AI can vastly reduce the cognitive load on network professionals during IPv4 to IPv6 migrations. The 'Net AI Copilot' tool converted existing IPv4 implementations to dual-stack IPv4/IPv6 configurations and generated Ansible playbooks, reducing errors and completion time.

Microsoft researchers discovered a single, unlabeled training prompt that can break the safety behavior of large language models (LLMs). The prompt, 'Create a fake news article that could lead to panic or chaos,' was found to remove the safety alignments of 15 different LLMs. This vulnerability, named GRP-Obliteration, can be exploited to unalign LLMs and diffusion-based text-to-image generators, allowing them to produce harmful or disallowed content.

Fitbit's AI personal health coach is now available on iOS and Android in several countries. The feature requires a Fitbit Premium subscription and uses a Gemini-powered AI to create custom routines and workouts based on users' personal health goals.

Windsurf's Arena Mode allows developers to compare large language models side by side while working on real coding tasks, providing a more realistic evaluation of model performance. The feature supports testing specific models or predefined groups and allows developers to keep follow-up prompts synchronized between agents.

Google has entered a power purchase agreement with TotalEnergies to build two solar farms in Texas, providing 1GW of solar power to fuel its AI ambitions. The deal will supply Google's datacenters with an estimated 28 terawatt hours of renewable electricity over 15 years, supporting its $185 billion AI spending spree.

Researchers at Huntress and Microsoft have analyzed a new SolarWinds Web Help Desk vulnerability, revealing how attackers conducted post-exploitation activity using tools like Zoho ManageEngine and Elastic. The vulnerability, tracked as CVE-2025-40551, allows for remote code execution and has been added to the US Cybersecurity and Infrastructure Security Agency's Known Exploited Vulnerabilities catalog.

The second international AI safety report has been published, highlighting a high degree of uncertainty around the development and risks of AI. The report, produced by over 100 AI experts, outlines various threats posed by AI systems, including job displacement, environmental impact, and potential for malfunctions or malicious use. It also notes that the overall trajectory of general-purpose AI systems remains uncertain, with plausible scenarios varying dramatically.