A persistent agentic research environment sustained near-daily use over 115 days, with 75,671 de-duplicated telemetry records across 96 active days (active-day fraction 0.835)¶

Claim · All claims
Evidence · 1 study (2 entries) · design r2 · q1
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q1 i? Over a 115-day observation window, recoverable main-agent telemetry contained 75,671 de-duplicated records across 96 active days, an active-day fraction of 0.835. → Anas H. Alzahrani 2026 q1 i? Active system time was estimated at 579.7 hours with a 30-minute capped-gap rule and 674.1 hours with a 60-minute sensitivity rule; these are system-activity estimates, not human labor hours. → Anas H. Alzahrani 2026 (2)

Evidence¶

Anas H. Alzahrani 2026¶

Anas H. Alzahrani, MD, PhD, MPH. (2026). Persistent AI Agents in Academic Research: A Single-Investigator Implementation Case Study. arXiv preprint. https://arxiv.org/abs/2605.26870 doi:10.48550/arxiv.2605.26870

q1 · i? · design · r2

Descriptive telemetry analysis of recoverable main-agent session records in a single-investigator self-observed case study (Jan 31–May 25, 2026). The article reports "75,671 de-duplicated records" across 96 active days, ADF = 0.835; counts are lower-bound estimates because recoverable telemetry starts February 2.

"Recoverable main-agent telemetry contained 75,671 de-duplicated records (DRC = 75,671) across 96 active days, corresponding to an active-day fraction of 0.835 (ADF = 96/115; Table 2)."

Anas H. Alzahrani 2026 (2)¶

Anas H. Alzahrani, MD, PhD, MPH. (2026). Persistent AI Agents in Academic Research: A Single-Investigator Implementation Case Study. arXiv preprint. https://arxiv.org/abs/2605.26870

q1 · i? · design · r2

Capped-gap active-time algorithm applied to sorted unique event timestamps from the same recoverable main-agent telemetry, with a 60-minute cap as sensitivity analysis. The article states these are "system-activity estimates, not direct human labor hours." No effect size is printed.

"Active system time was 579.7 hours using the primary 30-minute capped-gap estimate (ATE = 579.7 hours) and 674.1 hours using the 60-minute sensitivity estimate."

Discussion¶

Related Claims¶

• The memory layer recorded 889 failure, verification, correction, or protocol-proxy events (9.26 per active day), making governance part of the operating environment — related
• The environment produced 482 memory-recorded output-proxy events (5.02 per active day) across 10 artifact-surface categories spanning research, teaching, software, content, and operations — related
• May 2026 token telemetry was cache-dominant: 82.9% of 73,950,305 recorded tokens were cache reads, suggesting persistent workflows shift the economic unit from cost per token to cost per completed artifact — related
• Empirical ILE papers show a strong increase in breadth of content and time of use — related