AI Model ‘SleepFM’ Predicts Future Disease Risk from a Single Night’s Sleep
Researchers have developed a groundbreaking artificial intelligence foundation model, named SleepFM, that can predict the future risk of various diseases by analyzing data from a single night’s sleep. A study published in Nature Medicine details how the model was trained on over 585,000 hours of polysomnography (PSG) data from approximately 65,000 individuals. SleepFM has demonstrated a remarkable ability to predict the risk of numerous conditions—including dementia, heart failure, chronic kidney disease, and mortality—years before a clinical diagnosis might occur. The model analyzes complex physiological signals captured during sleep to identify patterns indicative of a predisposition to future health issues. This innovative approach could serve as a powerful early warning system, enabling proactive healthcare and early disease detection.
Sources:
Fiserv and Microsoft Partner to Accelerate AI Innovation in FinTech
Financial technology leader Fiserv, Inc. has announced a strategic collaboration with Microsoft to accelerate the integration of artificial intelligence into its development platforms and empower its global workforce with advanced AI tools. As part of the collaboration, Fiserv will deploy Microsoft 365 Copilot to its employees and expand its use of Microsoft Foundry, an AI platform powered by Azure. This initiative is expected to enhance productivity, streamline development processes, and drive the creation of new AI-driven solutions for Fiserv’s clients in the financial, business, and consumer sectors. The partnership builds upon Fiserv’s existing use of AI for critical functions like fraud detection and risk management.
Sources:
NVIDIA Accelerates AI Development with New Open-Source Models and Tools
NVIDIA has announced a significant release of new open models, data, and tools aimed at advancing artificial intelligence across various industries. The release includes models from the NVIDIA Nemotron family for agentic AI, the NVIDIA Cosmos platform for physical AI, the new NVIDIA Alpamayo family for autonomous vehicle development, NVIDIA Isaac GR00T for robotics, and NVIDIA Clara for the biomedical field. These resources are designed to provide companies with the essential tools to develop and deploy real-world AI systems. NVIDIA is also contributing open-source training frameworks and a vast collection of open multimodal data to accelerate innovation. The Nemotron family is being expanded with models for speech, multimodal retrieval-augmented generation (RAG), and safety, while the NVIDIA Cosmos models aim to bring human-like reasoning to physical AI.
AWS Launches Graviton4-Powered EC2 Instances for Enhanced Performance
Amazon Web Services (AWS) has announced the general availability of the new Amazon EC2 M8gn and M8gb instances. These instances are powered by the latest AWS Graviton4 processors, delivering up to 30% better compute performance compared to previous Graviton3-based instances. The M8gn instances are engineered for network-intensive workloads, featuring 6th generation AWS Nitro Cards that offer up to 600 Gbps of network bandwidth. The M8gb instances are optimized for workloads requiring high block storage performance, providing up to 150 Gbps of EBS bandwidth. Built on the AWS Nitro System, these new instances are supported by a wide range of Linux operating systems and popular applications, initially available in the US East (N. Virginia) and US West (Oregon) AWS Regions.
Sources:
CrowdStrike, AWS, and NVIDIA Launch 2026 Cybersecurity Startup Accelerator
CrowdStrike, in partnership with Amazon Web Services (AWS) and NVIDIA, has announced the 35 startups selected for its third annual Cybersecurity Startup Accelerator. This eight-week, free program is designed to foster the next generation of AI-driven cloud security innovation. The selected startups, chosen from a global pool of applicants, will receive mentorship, technical expertise, funding, and go-to-market support. The program will culminate in a final pitch day during the RSA Conference in March 2026, where five finalists will compete for an innovation award and potential investment from the CrowdStrike Falcon® Fund. This year’s cohort is focused on founders building cloud- and identity-first defenses to counter sophisticated AI-driven threats.
Sources:
Chainguard Report: AI’s Impact and Hidden Risks in Open-Source Software
A new report from Chainguard, titled “The State of Trusted Open Source,” reveals that Artificial Intelligence is significantly influencing the open-source landscape, with Python emerging as the most popular open-source image. The analysis, based on anonymized product usage and CVE data, finds that while many organizations standardize on popular projects, a significant portion of production infrastructure relies on a wide array of less common “long-tail” images. Critically, the report finds that 98% of vulnerabilities discovered and fixed in Chainguard images were located outside of the top 20 most popular projects. This suggests that the greatest security risks often lie in the less visible components of the software supply chain.
Sources:
Top Open-Source Projects for 2026 Focus on Rust, Performance, and AI
Several open-source projects are poised to gain significant traction in 2026, with common themes of enhanced performance, improved developer experience, and AI integration. Many of these tools, such as the Biome toolchain for web projects and the Ruff linter for Python, are written in Rust to achieve significant speed improvements over their predecessors. In the JavaScript ecosystem, Bun is positioned as a faster alternative to Node.js, while Deno continues to evolve with a focus on TypeScript and security. The Zed code editor, also built with Rust, is designed for high performance and multiplayer collaboration with integrated AI features. Other notable projects include Turso, a SQLite-compatible database for edge computing, and Ollama, which simplifies running large language models locally.
Sources:
Apache Software Foundation Elevates Three Projects to Top-Level Status
The Apache Software Foundation (ASF) has announced that Apache HertzBeat, Apache Teaclave, and Apache Training have all graduated to Top-Level Projects (TLPs). Apache HertzBeat is an open-source, real-time observability system with agentless monitoring, alerting, and notification capabilities. Apache Teaclave is a universal secure computing platform that utilizes Trusted Execution Environments (TEEs) to enable secure computation on sensitive data, with SDKs primarily written in memory-safe languages like Rust. The Apache Training project is an initiative to create and maintain high-quality, open-source training materials for various Apache projects, reducing duplicated effort and providing shared, IP-approved resources for the community.
Sources: