All briefings and quick hits published that day.
Google pairs a broadly available agent model with a restricted cyber variant built for vulnerability discovery and automated patching.
It combines a widely deployed agent model with consequential new rules for distributing advanced cyber capability.
Cursor adds scalable worker pools across corporate infrastructure while retaining its cloud-based planning, inference and orchestration layer.
Muse Spark 1.3 targets sustained coding and agent workflows, while Meta postpones both open weights and its highest reasoning setting.
A lightweight MiniMax experiment turns H3’s learned video dynamics into an interactive environment using only 8,000 samples.
Alibaba’s updated flagship model tops a major web-development benchmark while retaining a million-token context window.