A tale of scale

Infrastructure vendor Nscale this week finalized plans to acquire Anyscale, a specialist in scaling AI workloads. The transaction represents a significant test case for whether platform-agnostic AI tools can preserve their independence when integrated with proprietary GPU infrastructure.

Adrian Bridgwater, a contributor to The New Stack, sat down with Dan Bathurst, Chief Product Officer at Nscale, to discuss the acquisition and what appears to be an emerging vertically integrated AI cloud offering.

According to Bathurst, the Anyscale platform "continues to be its own brand and product."

He elaborated further: "What really changes — or how it's changing — is that customers now also get this first-party option, where they can have Anyscale running on Nscale fleet as a full-stack, highly-optimized solution."

Bathurst emphasized the company's competitive strategy: "Where we want to win is on performance, not on any sort of vendor lock-in or forcing of someone to choose Nscale as the infrastructure provider."

TOP OF THE STACK

Gemini Robotics 2 brings us one step closer to physical AGI

Google DeepMind unveiled Gemini Robotics 2 this week, an intelligence framework built from three distinct models designed to enable more flexible physical AI systems. The three models work in concert to equip robots with enhanced dexterity and full-body coordination, enabling them to tackle diverse multi-step operations. Google's announcement highlighted the vision-language-action capabilities embedded in this new generation.

WHAT ELSE IS NEW?

Is your SOC designed for AI scale attacks?

Mate's agentic Security Operations Platform processes every security alert with comprehensive organizational context, enabling teams to operate at machine-driven speeds.

  • Automatically resolves up to 85% of false positives
  • Reduces investigation duration from hours to minutes
  • Operates on a live Security Context Graph mapping your organization
  • Supported by $35M in recent funding to expand platform capabilities

FLOW STATE

Roadmap.sh has established itself as a standard tool for mapping technical career progression, and its premium tier elevates the offering considerably. Roadmap Pro introduces an AI-powered personal mentor alongside dynamic course generation tailored to specific needs. For engineers seeking a more organized, practical approach to learning emerging technologies without wading through scattered technical references, this tool merits consideration.

Subquadratic developed a model claiming to achieve 1,000x reduction in computational requirements for extended context windows. Alex Whedon, the company's Chief Technology Officer, contends the truly compelling aspect lies elsewhere. He discussed with The New Stack the "zero attention" architecture he believes will supersede current approaches, and his prediction that this innovation will eventually render his own flagship model obsolete.

Webflow's integration of AI into security workflows demonstrates how engineering teams can expand their capacity and streamline alert classification without establishing a dedicated security operations center.

Source: The New Stack