Agent learning speed is found to double every three months

Claim

ByteDance's EdgeBench ran 134 tasks of twelve-plus hours each to measure what agents learn from environments rather than recall, and the noisy learning curves collapsed into a clean log-sigmoid law with learning speed doubling every three months.

Reported metric: doubling every 3 months

Reporting date: 2026-07-03 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

• Welcome to July 3, 2026 (theinnermostloop.substack.com)