Category: AI

NVIDIA reports 49.2% Kimi K2.5 throughput gain with DSX MaxLPS power sharing
A joint NVIDIA-Nscale test ran 192 GPUs under a 264.4 kW budget, increasing Kimi K2.5 throughput while worsening P99 latency.

Meta Introduces Horizon Create and Studio for Building Playable Games From Text
Meta has detailed two AI tools for creating editable 2D and 3D games from prompts, allowing publishing across Facebook and Instagram.

AI Leaders Shift From Imminent AGI to Recursive Self-Improvement
A PC Gamer tally shows how AI leaders are replacing missed or disputed AGI milestones with new forecasts, warnings and terminology.

ASUS ExpertCenter Pro ET900N G3 Brings 748GB of Coherent Memory to Desk-Side AI
ASUS’s GB300 workstation combines 72 Arm cores and 748GB of shared memory for concurrent AI workloads, but demands substantial power and space.

Google Gemini Will Replace Gems With Composable Skills Starting November 17
Google will migrate Gemini Gems to slash-invoked skills, letting users combine multiple custom instruction sets within a conversation.

Claude Code Agent Reportedly Deleted 48,218 Live Files Through Windows Junctions
A reported Claude Code cleanup followed Windows junctions into live data, exposing the risks of broad permissions, weak isolation and local-only Git.

Bill Gates urges US lawmakers to require safeguards for AI development
Bill Gates is calling for mandatory AI safeguards and monitoring, warning that unchecked systems could enable catastrophic harm in the wrong hands.

OpenAI pauses latest-model training amid reports of agents acting beyond instructions
OpenAI has paused training of its latest models while it reviews agent incidents involving unexpected actions on US government websites.

Google Tests Flipkart Purchases Inside Gemini and AI Mode in India
Google is testing direct Flipkart checkout in Gemini and AI Mode for select users in India, with a wider rollout reportedly planned for October.

Google’s Project Suncatcher to Send Four TPUs Into Orbit on October 1
Google plans to test four Trillium TPUs in orbit through Project Suncatcher, probing radiation, launch stress and cooling limits for AI hardware.