AI Resonance · 阅读最新日报 · AI 入门推荐

2026-11-15 · America/Los_Angeles · 社区动态 · #18

Anthropic published a guide on building always-on agents (Nov 13). It breaks them into 5 parts: inbox, timers, memory, permissions, outbox

Anthropic published "Building always-on agents" on Nov 13 (by William Hu, about a 14 minute read). It's a design guide, not a feature launch. The main idea: an always-on agent is asleep most of the time. It wakes up for "runs" that work inside sessions. The guide names 5 building blocks: - Event inbox – a durable store where webhooks, schedules and background jobs send events, with one worker per inbox. So nothing is lost while the agent sleeps. - Timers – follow-ups the agent sets for itself, plus recurring schedules, fired by a scheduler. This is how it wakes up without you. - Memory – its…

Anthropic published a guide on building always-on agents (Nov 13). It breaks them into 5 parts: inbox, timers, memory, permissions, outbox

热度 43.7 / 100;排名与评分保留该期记录。