18 tasks. April 20, 2026 closed at 31.3x weighted leverage across 284.5 human-equivalent hours in 545 minutes of wall-clock time. Supervisory leverage came in at 271.0x.
That is 7.1 weeks of human-equivalent throughput in 9.1 hours. The ceiling was 165.0x; the floor was 5.6x. 10 of the 18 entries came from a single project.
Task Log
| # | Task | Human Est. | Claude | Sup. | Factor |
|---|---|---|---|---|---|
| 1 | Infrastructure | 110.0h | 40m | 1m | 165.0x |
| 2 | Content production | 20.0h | 12m | 1m | 100.0x |
| 3 | Resolved all AVIAN documentation audit findings: 14 HIGH (missing files), 42 MEDIUM (sections, filenames, coverage), ~100 LOW (PRIVATE, cross-refs, diagrams, word counts, source refs). ~85 repo touches across… | 40.0h | 45m | 5m | 53.3x |
| 4 | Accelastudy UI restructure: per-course dashboards (Overview/Autopilot/Curriculum/Activities tabs), main dashboard shows readiness card per course with aggregated today plan, dropped… | 16.0h | 25m | 8m | 38.4x |
| 5 | Deployment | 6.0h | 12m | 2m | 30.0x |
| 6 | Ran all 7 AVIAN audits in parallel (health-check, full-readiness, security, content, accessibility, compliance, documentation) across 70 repos; reports written to avian-audits/reports/ | 20.0h | 50m | 2m | 24.0x |
| 7 | Fix 5 activity bugs in avian-app-web (practice exam count, flashcards, lessons nav, PracticeExam placeholder, autopilot plan dispatcher) + build canonical activities-per-cert catalog across 4 repos | 22.0h | 55m | 6m | 24.0x |
| 8 | Fix two compounding bugs in question-bank answer endpoint: shuffled options had mismatched per-option explanations (raw_options[shuffled.index(...)] cross-index bug), and wrong-answer selected_explanation was… | 3.0h | 9m | 3m | 20.0x |
| 9 | Coding | 8.0h | 35m | 3m | 13.7x |
| 10 | Diagnose and fix accelastudy auth/purchase service churn: identified t4g.small undersized for 18-container tools fleet, resized to t4g.large, identified zombie containers | 2.5h | 14m | 3m | 10.7x |
| 11 | Codex path-based page routing fix — replace by-slug lookup with by-path so pages with duplicate slugs (e.g., READMEs from multiple linked repos) resolve distinctly. Backend adds get_by_path +… | 3.0h | 18m | 3m | 10.0x |
| 12 | Codex linked-page provenance UI: synchronized status pill, disabled Edit button, edit-in-GitHub modal. Backend exposes source_type/linked_path/linked_repo_name/linked_source_url via eager-loaded… | 2.5h | 15m | 2m | 10.0x |
| 13 | Infrastructure | 5.0h | 30m | 4m | 10.0x |
| 14 | Bulk-link 19 tool repos into prod Codex AVIAN space + trigger initial syncs. Wrote two scripts (bulk_link_tools via LinkedRepoService.bulk_create, trigger_tool_syncs via HMAC-signed webhooks) running against… | 4.0h | 25m | 2m | 9.6x |
| 15 | Composite autopilot with hierarchical exam-pass priors — Morgan passes CLF+SAA in 22 days (was stuck at readiness 0.000). Validated end-to-end, engine+[engine subsystem] changes committed and pushed. | 14.0h | 90m | 8m | 9.3x |
| 16 | Codex realtime WS fanout (space-scoped broadcasts + event_bus bridge + frontend WS client), theme-toggle fix (design-system vs app-shell useTheme), and linked-repo title dedupe (fallback to filename stem when… | 3.0h | 20m | 3m | 9.0x |
| 17 | Documentation | 2.5h | 18m | 3m | 8.3x |
| 18 | Debugging | 3.0h | 32m | 4m | 5.6x |
Aggregate Statistics
| Metric | Value |
|---|---|
| Total tasks | 18 |
| Total human-equivalent hours | 284.5 |
| Total Claude minutes | 545 |
| Total supervisory minutes | 63 |
| Total tokens | 3,501,000 |
| Weighted average leverage factor | 31.3x |
| Weighted average supervisory leverage factor | 271.0x |
| Human-equivalent weeks | 7.1 |
Analysis
The highest factor of the day came in at 165.0x and the lowest at 5.6x, a spread of 29.3 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 110.0 of the 284.5 human-equivalent hours, or 39 percent of the day. No single task dominated the total, so the weighted average is representative.
Supervisory time was 63 minutes against 545 minutes of execution, a ratio of about 1 to 9. Supervisory leverage of 271.0x 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.