8 tasks. July 12, 2026 closed at 26.0x weighted leverage across 282.0 human-equivalent hours in 652 minutes of wall-clock time. Supervisory leverage came in at 352.5x.
That is 7.0 weeks of human-equivalent throughput in 10.9 hours. The ceiling was 88.4x; the floor was 6.4x. 5 of the 8 entries came from a single project.
Task Log
| # | Task | Human Est. | Claude | Sup. | Factor |
|---|---|---|---|---|---|
| 1 | Temporal cutover forensics + verified re-cutover + 9-agent full review (job layer/design/[ip]/speed/validation) + P0 hardening batch (2 repos, 2256 tests green) + remediation artifact | 140.0h | 95m | 6m | 88.4x |
| 2 | Deployment | 49.0h | 65m | 8m | 45.2x |
| 3 | Post-crash forensics + validation screen relaunch: watchdogd panic diagnosis; ssh-agent leak fix; flashcards crash-resume; sequential load-gated runner; stack restart + commits | 7.0h | 25m | 3m | 16.8x |
| 4 | Validation funnel apply: per-item-verified hard-defect repairs → 538 fixes promoted live across 117 packages (gate refused bulk-trust, switched to per-item verify, 519 grounding→relink worklist); style-drip… | 22.0h | 80m | 4m | 16.5x |
| 5 | Infrastructure | 26.0h | 100m | 5m | 15.6x |
| 6 | Coding | 9.0h | 42m | 7m | 12.9x |
| 7 | Extended apply_verdicts.py with pairs/flashcards/prose lanes + generate_repairs.py; 4-gate addendum; caught+fixed real bug; live-tested vs production data (agent-built) | 13.0h | 95m | 10m | 8.2x |
| 8 | Temporal long-running-activity support (heartbeat pump + subprocess kill-on-cancel + per-kind timeouts) + Temporal-aware job resubmit (campaign-cutover enabler), 2 committed [engine subsystem] features w/ 26… | 16.0h | 150m | 5m | 6.4x |
Aggregate Statistics
| Metric | Value |
|---|---|
| Total tasks | 8 |
| Total human-equivalent hours | 282.0 |
| Total Claude minutes | 652 |
| Total supervisory minutes | 48 |
| Total tokens | 6,795,000 |
| Weighted average leverage factor | 26.0x |
| Weighted average supervisory leverage factor | 352.5x |
| Human-equivalent weeks | 7.0 |
Analysis
The highest factor of the day came in at 88.4x and the lowest at 6.4x, a spread of 13.8 times between the two. That is a wide range for a single day, and it usually means the day mixed mechanical work with work that needed real judgement.
The largest single entry accounted for 140.0 of the 282.0 human-equivalent hours, or 50 percent of the day. One task carrying that much of the total is worth noting; the average is doing less work than it appears to.
Supervisory time was 48 minutes against 652 minutes of execution, a ratio of about 1 to 14. Supervisory leverage of 352.5x is the figure I find most honest, because it measures the hours I actually spent rather than the hours a machine spent on my behalf.
Every figure here is recorded at the time the work is done rather than reconstructed afterwards. The human estimate is my own judgement and carries the uncertainty that implies; the minutes and tokens are measured. The full dataset, including this day, is available for download.