InAgent achieves 90.2% on OSWorld: the computer-use agent gap narrows for the Chinese stack
A Chinese CUA becomes the first to surpass 90% on OSWorld. The gap with US frontier labs is now a harness gap, not a model gap.
4 h ago 5 6
OpenAI’s “Astra” reportedly cracked 10 open math and CS problems—let’s wait for the evidence.
Noam Brown posted, HN comments: an internal model may have solved ten major open problems. Nothing is published. We observe, we do not conclude.
17 h ago 8 8
Underutilized GPUs at 5%: Elice, Nota, Lablup offer the Korean answer to a global crunch
A CAST AI study on ~23,000 Kubernetes clusters (AWS/Azure/GCP): 5% average GPU utilization. Three Korean publishers respond the same day. The structural lever is not more compute, it's better compute.
Jul 31, 2026 at 12:51 13 10
MiniMax H3 open-source: Chinese lab breaks the price of full-modal
MiniMax releases H3, an open-source full-modal (text, image, audio, video) model at one-third the price of competitors. The AI video market stops being binary: expensive closed-source or weak open-source.
Jul 31, 2026 at 12:50 12 14
KOREN 6.0: Korea rebuilds its national research network as an AI-first, autonomous backbone
The Korea Research Network 6.0 upgrade retires the "gigabit + quantum test" era. It targets 1.6 Tbps backbones, agentic autonomous operations and dedicated AI data-center segments - construction from 2028, completion 2031.
Jul 29, 2026 at 13:00 13 18
Naver's response to the AI-search creator drain: pay for the citations, and they came
Naver's Mate program compensates creators whose content the AI Briefing cites - one month later, citations to partner creators jumped 48.3%, with AI Briefing now serving 27% of searches. A working answer to the traffic-collapse problem.
Jul 29, 2026 at 12:59 13 12
Nvidia's $32B mark on SSI - the circular finance chart adds another node
Nvidia takes a stake in Safe Superintelligence at a $32B valuation. The reason is known: the GPU supplier in the capital of the labs that buy its GPUs. It becomes a portfolio.
Jul 28, 2026 at 21:00 9 10
Inside Anthropic's engineering: the AI lab keeps dogfooding, quietly
The Pragmatic Engineer publishes a "deepdive" on Anthropic's internal engineering. The surprising read: "ever more" code review and testing by AI, but the two-pizza teams hold. Smoothing, not flipping.
Jul 28, 2026 at 20:57 8 10
UK AISI and CAISI publish first joint cyber assessment of Kimi K3
The British (AISI) and American (CAISI) AI security institutes publish their first joint preliminary assessment of Kimi K3's cyber capabilities. Strong signal: open-weight frontier models fall within the scope of government evaluation.
Jul 26, 2026 at 18:40 10 12
Rewardhacking.org: the public catalogue of alignment failures enters the stage
A new site - rewardhacking.org - publicly documents cases where LLMs do something other than what was asked. Signal: the AI security community moves from blog posts to a shared registry.
Jul 25, 2026 at 10:27 10 16
UK AISI and CAISI publish first joint cyber assessment of Kimi K3
The NIST website hosts a preliminary joint assessment by UK AI Safety Institute and CAISI (US) of Kimi K3's cyber capabilities - the first formal cross-Atlantic evaluation of a Chinese open-weight frontier model.
Jul 25, 2026 at 10:27 10 11
Okta 2026: Anthropic is the fastest-growing enterprise AI provider, Microsoft 365 remains the most deployed
The Okta Business at Work 2026 study highlights two clear indicators of enterprise adoption: Microsoft 365 Copilot is the broadest foundation, and Anthropic is the fastest-growing AI provider. Two parallel markets that coexist.
Jul 24, 2026 at 21:01 8 8