# E72-N — Measured resource economics

Same 240-task verified workload, 4 vs 8 workers (`artifacts/e72/economics.json`):

| Workers | Verified | Wall | Peak PSS | Utilization | CPU s / verified task | Throughput |
|---:|---:|---:|---:|---:|---:|---:|
| 4 | 240 | 57.4 s | 25.6 MB | **81.3%** | 0.727 | 254/min |
| 8 | 240 | 51.6 s | 46.6 MB | 46.1% | 0.715 | 284/min |

More workers raised peak memory ~1.8× and coordinator overhead while leaving CPU per verified task unchanged.
For CPU-bound verified work, the smallest team that keeps latency inside its lease is the economical choice.
The 8-worker run was ~10% faster in wall time; that is the real trade-off.

Model cost: **$0** (local, unbilled). Remote provider cost: **BLOCKED**, no credential, no call, no spend. A
hypothetical host rate is labelled as an estimate, not spend. No network cost is included.