Claude for Long-Horizon Tasks — Lance Martin, Anthropic
Anthropic's Claude now supports multi-hour autonomous tasks, unlocking true async agent workflows.
“In order to really unlock async, we needed longer task horizons.”
Lance Martin from Anthropic traces Claude's task horizon evolution from 10-20 minutes (Opus 3, 2024) to hour-long synchronous coding agents (Claude Code), and now toward multi-hour async autonomous work. He argues async agents were fundamentally unusable until models could sustain longer horizons without hitting errors and interrupting users. The corresponding API evolution—from Messages API to Agent SDK—signals Anthropic is building infrastructure specifically for long-running, unsupervised agent deployments.