— Mike Pritchard, Nvidia’s director of climate simulation, on the AI giant’s new open source weather forecasting models. Go deeper.
New episode: Solving the problems that accompany API sprawl with AI
In this episode of The New Stack Makers, we sit down with IBM’s Neeraj Nargund to discover how AI-infused platforms provide automated discovery, governance, and observability.
Autonomous AI agents act continuously across tools, data, and infrastructure. When identity is fragmented, security teams lose visibility and control. Learn why closing the identity gap is essential to deploying agentic AI safely at scale.
Can open source AI bring the forecast to 2.6 billion people?
In many respects, the world of AI development is a fiercely competitive arena where companies guard trade secrets like the billions they are worth. It’s OpenAI vs. Google vs. Anthropic and more. That cutthroat competition makes open source AI a bit of a novelty. Think DeepSeek about a year ago. But if you’re making the hardware that produces AI, open source is a much more appealing idea. After all, if the open source AI models do vastly helpful things – like forecast the weather – you might sell a lot more GPUs to large customers like governments.
That brings us to our lead story this week, about Nvidia’s debut of two open source weather forecasting models: Earth-2 Medium Range (15-day forecasts) and Earth-2 Nowcasting (six-hour forecasts). In addition, Nvidia revealed a tool that will significantly speed up the generation of starting conditions for these models, reports Senior Editor for AI Frederic Lardinois.
It’s a fascinating story with global implications. After all, about a third of the world’s population — some 2.6 billion people — doesn’t have access to early warning systems for severe weather, according to the World Meteorological Organization. Without a proper warning, severe weather can disrupt lives or worse for scores of people. This week’s lead story explores how open source AI could help close the weather gap for billions.
Thanks for being with us. And don’t forget your coat.
How to architect a CI/CD pipeline that really scales February 3 | Virtual
Join us live as we welcome Harness’s Eric Minick and Stephen Cihak for a spirited discussion on why copying CI/CD templates creates technical debt and leads to operational paralysis. You’ll learn why the future of DevOps scaling isn’t about copying YAML; it’s about inheriting it.
Scale 23x Social Linux Expo March 5-6 | Pasadena, CA
SCALE, North America’s largest community-run open source conference, returns to Pasadena Convention Center March 5-8, 2026. Join for four days of sessions, workshops, and community activities focused on open source, security, DevOps, cloud native, and more.
Get an in-depth look at the top five managed file transfer trends transforming how organizations move data in 2026 and beyond. Catch the lively conversation about how modern MFT solutions are built for your enterprise.
When Argo scales, visibility breaks. Hitting the Argo Ceiling means tab fatigue, glue code, and lost engineering time. Watch the replay to learn how enterprises fix GitOps sprawl without replacing their tools.
