Contributing writer covering systems programming and performance engineering.
Reinforcement LearningA UCL-led preprint trains a Set Transformer PPO agent to schedule Ariel's exoplanet survey, beating simple baselines while the authors flag key limits.
AI AgentsMeta launches Muse Wonder and Wonder Full with Active Memory and Active Focus to help AI agents remember context, recover from errors and complete complex tasks.
Tools & FrameworksBacked by $20 million from Accel, Index Ventures, and Emergence Capital, Ando is a team messaging platform built so AI agents can join channels, message colleagues, and coordinate work as full participants rather than as add-on integrations.
