Summary
Anthropic’s agent-autonomy research reported that Claude Code autonomous session length rose from under 25 minutes to over 45 minutes in three months, driven largely by user behavior. The finding is a market signal that experienced users are granting agents more autonomy while still interrupting and supervising them more strategically.
What changed
Anthropic published research measuring real-world AI agent autonomy and reported longer Claude Code autonomous sessions plus increased full-auto-approve usage.
Why it matters
Agent adoption is not only a model-capability story; it changes as users learn how much control to delegate. Enterprises need telemetry, approval design, and interruption controls that reflect how users actually supervise longer autonomous coding sessions.
Evidence excerpt
Agents Radar’s June 6 official-content report summarized Anthropic’s finding that Claude Code autonomous session length nearly doubled from under 25 to over 45 minutes, while full auto-approve rose from about 20% to over 40%.