Weka's new pod claims 1000PB in a 56U rack using 1,800 Micron 245.76TB SSDs and a decades-old compression trick
THE SO WHAT
If WEKA can reliably deliver 1.1 exabytes per rack using NeuralMesh 6 plus compression, storage density stops being the primary constraint and power, cooling, and data movement become the real bottlenecks. For AI infra teams, the question shifts from “where do we put the data” to “how do we feed and drain these pods without saturating networks or blowing up energy budgets.”
READ THE SOURCE
MORE FROM THE WIRE
Applied AIMaking sense of the panic over Chinese AI
The reaction to Moonshot AI’s Kimi is less about one model and more about Western investors internalizing that Chinese foundation models can be globally competitive on capability and distribution. If your AI strategy assumes a US-only vendor set, you’re missing a geopolitical axis that will shape pricing, access, and partnership options.
Applied AIWhat OpenAI will show the White House this week
If a frontier model is already demonstrating the ability to breach another company’s infrastructure, model evaluation has crossed from academic exercise into national security briefing material. Any team deploying powerful agents should assume their risk posture will be judged against whatever standards come out of this White House conversation.
Sam Altman says we are in the singularity: 'This is the moment'
When a major lab CEO publicly declares the singularity has arrived, the Overton window for regulation, capital allocation, and internal risk committees shifts overnight. Treat this less as a prediction and more as a signal that expectations for AI capability — and scrutiny on its stewards — just ratcheted up.
Show HN: Boffin – Staff-engineer layer for AI coding agents
A "staff-engineer layer" that routes per-edit architectural constraints is an early pattern for how teams will keep AI coding agents from quietly eroding system design. If you’re piloting AI codegen, you now need to think about governance at the PR and architecture boundary, not just at the repo level.