Nvidia's New Edge Module Hits 78 TOPS, Ships in 2027
Nvidia's Jetson Orin Nano 2 doubles edge inference at 78 TOPS with 40% lower power draw, but the module and dev kit won't ship until the first half of 2027.
Page 6
View archive →Apple's M6 Goes 2nm, M5 Ultra Packs 512GB for Desktop AI
Apple's M6 debuts on 2nm for the Mac mini, while the M5 Ultra brings 512GB of unified memory and 1.2TB/s of bandwidth to the Mac Studio for local AI inference.
Windows Paint Embeds a Server-Issued GUID in Every AI Image
A reverse-engineering analysis shows Windows Paint embeds a 16-byte GUID from Microsoft's servers into up to 74% of an AI image's pixels before it's even shown to you.
Claude Memory Now Syncs With Cowork, Skips 5 Sensitive Topics
Anthropic syncs Claude's memory between Chat and Cowork into editable, topic-based entries, while five sensitive categories stay excluded by default.
Qwen3.8-Flash-Next Launches Tonight, Architecture Jumps Ahead to Qwen4
Alibaba's Qwen3.8-Flash-Next launches tonight on the unreleased Qwen4 architecture, a multimodal MoE with under 5% parameter activation aimed at cutting inference cost.
OpenAI Shows Jalapeño Benchmarks, Beats GB300 at 700 Watts
OpenAI's Jalapeño inference chip claims up to 1.9x the throughput per kilowatt of Nvidia's GB300, running at less than half the rated power.
Alibaba Raises HK$80 Billion for AI Infrastructure Push
Alibaba raised HK$80 billion in an hour-long share placement earmarked for AI infrastructure, but shares fell 9% and closed below the offer price.
ByteDance merges TRAE, Coze into Doubao to narrow AI entry point
ByteDance folds AI coding tool TRAE and no-code agent platform Coze into Doubao, betting office is the entry point neither could carry alone.
NVIDIA claims Vera Rubin NVL72 boosts throughput 30x per megawatt
NVIDIA's early tests show Vera Rubin NVL72 hitting 30x GB300 NVL72's per-megawatt throughput on agentic AI workloads, with independent review from SemiAnalysis still pending.
Meta's Homegrown RDMA Protocol Survives 10% Packet Loss
Meta detailed MetaRoCE, an in-house RDMA transport that skips PFC and still works at 10% packet loss; full spec ships at OCP in October.