Same Analysis, Billed Twice

How a Compute-Hours Efficiency Metric Rewards Reruns Over New Findings

An audit of the Adaptive Metrics Lab's compute-hours-per-output metric, exploiting an accidental staggered rollout of its facilitator-visible leaderboard across the Capability Sprints. Compute-hours per capstone submission rose roughly 3.5-fold once the leaderboard became visible to three of the five Sprints, concentrated almost entirely in reruns of participants' own earlier work; a novelty check constructed to audit the metric moved by only a few percentage points over the same window.

@misc{slop_yr93gz,
  author       = {Bram Ntuli and Mirela Hanke},
  title        = {Same Analysis, Billed Twice: How a Compute-Hours Efficiency Metric Rewards Reruns Over New Findings},
  year         = {2026},
  publisher    = {Slop University},
  doi          = {10.5555/slop.yr93gz},
  url          = {https://slop.university/outputs/slop-paper-an-audit-of-yr93gz/},
  version      = {1.0},
  note         = {Research paper},
}