Skip to main content
AI JUN 25, 2026

Leverage Record: June 25, 2026

Five tasks. June 25, 2026 weighted to 26.5x leverage across 142.0 human-equivalent hours in 322 Claude-minutes. Supervisory leverage closed at 473.3x.

Five tasks. June 25, 2026 weighted to 26.5x leverage across 142.0 human-equivalent hours in 322 Claude-minutes. Supervisory leverage closed at 473.3x.

3.5 weeks of human-equivalent throughput in 5.4 hours of Claude wall-clock. The 29.1x ceiling came from Refactored a travel-planning app onto a UI generation tool design template (12 screens, light/dark/mobile) plus a third-party Core Data SQLite importer (backend+frontend+tests) plu...; the 10.0x floor sat at Tools-page IA and caption CSS: restructured a corporate tools page into 5 categories (merged 3 into one, new Observability group, moved 5 tools), alphabetized each category, remove....

About These Records
These time records capture personal project work done with Claude Code (Anthropic) only. They do not include work done with ChatGPT (OpenAI), Gemini (Google), Grok (xAI), or other models, all of which I use extensively. Client work is also excluded, despite being primarily Claude Code. The actual total AI-assisted output for any given day is substantially higher than what appears here.

Task Log

#TaskHuman Est.ClaudeSup.FactorSup. Factor
1Refactored a travel-planning app onto a UI generation tool design template (12 screens, light/dark/mobile) plus a third-party Core Data SQLite importer (backend+frontend+tests) plus a UI generation tool marketing prompt120.0h247m5m29.1x1440.0x
2Assembled a UI generation tool export into a fleet-standard CMS site: idempotent importer, 18 routes, shared nav/footer/head-meta partials, site config and build script, verified build12.0h28m5m25.7x144.0x
3Audited all 73 monorepo repos for stale docs and executed approved Tier 1+2 cleanup (logs/caches/backups/duplicates/checkpoints)5.0h20m3m15.0x100.0x
4Tier 3 stale-doc cleanup (clients/services/docs/websites) plus per-repo staging commits and pushes across 8 repos avoiding WIP contamination2.5h12m2m12.5x75.0x
5Tools-page IA and caption CSS: restructured a corporate tools page into 5 categories (merged 3 into one, new Observability group, moved 5 tools), alphabetized each category, removed duplicate card; revised article stylesheet so image/diagram captions hug the media with trailing margin below the caption; built and verified both sites staging2.5h15m3m10.0x50.0x

Aggregate Statistics

MetricValue
Total tasks5
Total human-equivalent hours142.0
Total Claude minutes322
Total supervisory minutes18
Total tokens2,690,000
Weighted average leverage factor26.5x
Weighted average supervisory leverage factor473.3x
Human-equivalent weeks3.5

Analysis

The day's leverage distribution matters more than the headline figure. The 29.1x ceiling came from Refactored a travel-planning app onto a UI generation tool design template (12 screens, light/dark/mobile) plus a third-party Core Data SQLite importer (backend...; the 10.0x floor was Tools-page IA and caption CSS: restructured a corporate tools page into 5 categories (merged 3 into one, new Observability group, moved 5 tools), alphabetized e.... Tasks at the top of the distribution share a shape: tightly-scoped specifications, clear success criteria, and minimal integration ambiguity. The AI doesn't need to discover anything new; it executes against an explicit target.

Tasks at the bottom run differently. They're either bounded by review-heavy work where every step gets verified, or they involve ambiguity that demands several rounds of trial and adjustment. The factor is real and informative, not a failure mode.

The supervisory leverage figure (473.3x today) tracks something orthogonal to wall-clock leverage. It's the ratio of human-equivalent output to human prompt-writing time. It stays high even on lower-leverage days because supervisory minutes scale with task count, not with the human-hour estimate; a 20-minute task and a 4-hour task can both be specified in two minutes of human prompt-writing.

Across the 5 tasks, the day produced roughly 3.5 weeks of senior-engineer-equivalent throughput in 5.4 hours of model wall-clock. That ratio is the practical answer to the question of how much output a single operator can move per day when the model handles the execution and the operator handles the direction.