26 tasks. April 22, 2026 closed at 16.1x weighted leverage across 189.1 human-equivalent hours in 704 minutes of wall-clock time. Supervisory leverage came in at 143.6x.
That is 4.7 weeks of human-equivalent throughput in 11.7 hours. The ceiling was 98.8x; the floor was 4.0x. 22 of the 26 entries came from a single project.
Task Log
| # | Task | Human Est. | Claude | Sup. | Factor |
|---|---|---|---|---|---|
| 1 | Compliance and documentation audits across 71 repos (SOC2/GDPR/CCPA check + docs remediation with 44 repos committed and pushed) | 28.0h | 17m | 2m | 98.8x |
| 2 | Implement autopilot_sync_plan in full: per-course individual autopilot goals persisted to Postgres, server-backed enrollments with write-through cache, auto-managed composite across 3 repos (engine,… | 24.0h | 34m | 5m | 42.4x |
| 3 | Docs audit remediation blast across 30+ repos: coordinated documentation audit 2026-04-22 + audit timestamps, coverage targets, source file references, Overview/Tech Stack/Security sections, Mermaid diagram… | 30.0h | 45m | 5m | 40.0x |
| 4 | Admin WebSocket architecture: all 5 phases (WS scaffolding in admin-service, Redis event bus + publishers in auth/purchase/notification services, event forwarder, SPA WsClient + hooks, migrate Comps page,… | 12.0h | 20m | 3m | 36.0x |
| 5 | Audit and review | 4.0h | 10m | 2m | 24.0x |
| 6 | Fix lessons activity navigation in avian-app-web: CourseDetail was calling getDomainTaxonomy with course.slug instead of course.domainId, causing taxonomy fetch to 404 and falling back to [content generation]… | 2.5h | 7m | 2m | 21.4x |
| 7 | docs/avian-architecture: write autopilot_sync_plan.md — per-course individual goals persisted to Postgres, user_enrollments in auth-service, auto-managed composite on top; full data model, API surface,… | 3.0h | 9m | 5m | 20.0x |
| 8 | Infrastructure | 6.0h | 18m | 3m | 20.0x |
| 9 | Design and frontend | 1.2h | 5m | 1m | 15.0x |
| 10 | Surface enrollment 403 + 404 properly, deploy, and grant dev comp. Client now rolls back optimistic rows and toasts NEEDS_UPGRADE instead of leaving ghost cards.… | 3.0h | 12m | 3m | 15.0x |
| 11 | Labs S3 migration: wire CourseDetail to CDN, delete bundled manifest + predev/prebuild scripts; add separate buildspec-labs.yml + CodeBuild-safe upload script; Terraform labs-cdn module + separate… | 4.0h | 16m | 2m | 15.0x |
| 12 | avian-activities-react + avian-app-web: Flashcard auto-submit on recall click (drop Submit + Hint buttons), add lesson-link button wired via course slug in spec, hide hint escalation level +… | 3.0h | 13m | 6m | 13.8x |
| 13 | Finish autopilot sync plan: composite /next rendering with cross-domain pills, shared-concept 'credits every listed exam' hint, and 4-exam cap notice when 5+ autopilots are active | 4.0h | 19m | 1m | 12.6x |
| 14 | Deployment | 18.0h | 90m | 3m | 12.0x |
| 15 | Coding | 8.0h | 40m | 2m | 12.0x |
| 16 | Labs-on-S3 infrastructure: terraform module for labs-cdn bucket/distribution + publish pipeline, separate buildspec + upload script, CodeBuild detection to skip named AWS profile, avian-console-sim-react… | 12.0h | 60m | 3m | 12.0x |
| 17 | avian-app-web + design-system: add hideCloseButton to DrawerProvider; blank out release notes drawer (suppress close X, drop changelog link); diagnose + document cross-browser enrollment state… | 2.0h | 10m | 5m | 12.0x |
| 18 | Coding | 2.0h | 15m | 3m | 8.0x |
| 19 | Make admin WebSocket live: inline subscriber into admin-service, delete unused libs/avian-events, remove old AdminWebSocket files, migrate ActivityFeed, fix accessibility test, run backend tests across 4… | 2.0h | 15m | 1m | 8.0x |
| 20 | LabRunner friendly header + back-to-origin, Dashboard driver.js tour restoration (retargeted to live testids, SEEN_KEY bumped), hide language picker on Settings | 1.5h | 14m | 2m | 6.4x |
| 21 | labs-cdn terraform import of 9 resources + apply (2 new pipeline/CodeBuild) + debug first run (boto3 ProfileNotFound from empty AWS_PROFILE) + verify second run rewrites manifest | 1.0h | 10m | 1m | 6.0x |
| 22 | Lessons pass: breadcrumbs replace double back, curriculum tab restore via ?tab=curriculum, .content strong color inherit (kill link-blue), ElevenLabs TTS wired (SSM SecureString param + engine in-place… | 2.0h | 22m | 2m | 5.5x |
| 23 | lab-frame CSS fix: expose console-sim CSS via package exports, import in lab-frame.tsx, sequence lib npm publish before app-web redeploy, verify lib CSS classes present in deployed vendor bundle | 0.5h | 6m | 1m | 5.0x |
| 24 | Activities pass: VITE_FEATURE_COMPETITIVE flag, slug→domainId at Scenario/MicroChallenge/CrossDomain boundaries, MCQ FeedbackBlock+ReviewRow lesson_href link, course-title hover:underline affordance on… | 1.0h | 12m | 3m | 5.0x |
| 25 | Comp welcome email markdown rendering + AccelaStudy® branding + preview UI; Comps table enrichment (names, computed status, bulk user lookup); admin-service JWT kid-less fix; WebSocket architecture plan… | 14.0h | 180m | 12m | 4.7x |
| 26 | Console-sim auto-start flash fix: gate root route on initialLabId to suppress LabSelector flash during auto-start; re-published lib and triggered app-web rebuild | 0.3h | 5m | 1m | 4.0x |
Aggregate Statistics
| Metric | Value |
|---|---|
| Total tasks | 26 |
| Total human-equivalent hours | 189.1 |
| Total Claude minutes | 704 |
| Total supervisory minutes | 79 |
| Total tokens | 3,030,000 |
| Weighted average leverage factor | 16.1x |
| Weighted average supervisory leverage factor | 143.6x |
| Human-equivalent weeks | 4.7 |
Analysis
The highest factor of the day came in at 98.8x and the lowest at 4.0x, a spread of 25.0 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 28.0 of the 189.1 human-equivalent hours, or 15 percent of the day. No single task dominated the total, so the weighted average is representative.
Supervisory time was 79 minutes against 704 minutes of execution, a ratio of about 1 to 9. Supervisory leverage of 143.6x 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.