Weather forecasting goes open source

NVIDIA wants weather AI everywhere

This pushes weather AI out of the lab.

NVIDIA announced Earth-2, a fully open set of AI weather models and tools, covering everything from data assimilation to 15-day global forecasts and six-hour local storm predictions. The stack includes new models like Medium Range and Nowcasting, runs on GPUs instead of supercomputers, and is already being tested by groups like the U.S. National Weather Service, TotalEnergies, and Israel’s meteorological agency.

It was unveiled at the AMS meeting in Houston. The real shift is not accuracy bragging but access, because NVIDIA is turning what used to require national supercomputing budgets into software that agencies and companies can run and tune themselves. This feels like a play to make GPUs part of everyday weather operations, not just climate research, especially when they claim hours of compute can drop to seconds.

This is practical, not flashy. If these models hold up in real operations, smaller countries and private firms gain forecasting power that used to be centralized. Traditional physics models will still matter, but they are losing their monopoly on trust.

Once forecasts are cheap and fast, who decides how they get used?

Indeed turns AI into hiring muscle

This is less futuristic than it sounds.

Indeed says AI now touches nearly every part of its platform, from job recommendations to recruiter tools like Talent Scout and Career Scout. According to the company, about 70% of sponsored applications come from AI-driven recommendations, and some employers are hiring 40% faster using AI-powered sourcing.

The interview was published Monday. It reads like a calm status update. What matters here is not the agents or the language about responsibility, but the proof that AI is already baked into hiring workflows at scale without spooking customers. Indeed keeps repeating the same line for a reason: humans stay in charge, AI clears the busywork, and results show up in time saved and roles filled.

This feels operational, not experimental. If this model holds, recruiters who avoid AI will simply move slower than the market. The competitive edge shifts from access to AI to how comfortably teams use it every day.

At some point, the question won’t be whether hiring uses AI, but whether anyone remembers how it worked without it.

Micron doubles down on Singapore

This is about floor space, not hype.

Micron says it will spend $24 billion to expand a NAND plant in Singapore, adding 700,000 square feet of cleanroom space at an existing site. Production is slated for the second half of 2028, while the same complex is already gearing up to ship high-bandwidth memory for AI workloads in 2027. The factories run all night. You can see the lights from the road.

What stands out is the timing, because Micron is still steering capital toward HBM for AI even as it admits that choice is tightening supply of more basic memory used in PCs, servers, and phones. This investment reads less like a growth flex and more like damage control after years of underbuilding NAND while AI sucked attention and capacity elsewhere.

This is expensive. It is slow. Shortages are expected to last through late 2027, which means buyers stay price-takers for a while longer. Singapore keeps locking itself in as a core node while China, Taiwan, Japan, and Malaysia fill in the rest.

The question is whether demand cools before this cleanroom ever ships a chip.

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.