Skip to main content
AI22 June 2026

Mostly coding

12 tasks. June 22, 2026 closed at 20.2x weighted leverage across 124.7 human-equivalent hours in 371 minutes of wall-clock time. Supervisory leverage came in at 213.8x.

That is 3.1 weeks of human-equivalent throughput in 6.2 hours. The ceiling was 140.0x; the floor was 2.4x. 6 of the 12 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1Reviewed test coverage across all 19 libs/ repos: stack, runner, test/source counts, coverage gates, and gaps; fanned out 10 parallel audit agents and [content generation]14.0h6m2m140.0x
2Testing56.0h77m3m43.6x
3Coding28.0h75m2m22.4x
4Audit and review5.0h20m3m15.0x
5Audit and review4.0h20m5m12.0x
6Coding4.5h30m2m9.0x
7Convert salvage_confirm_sweep (~2900-node 2-stage Haiku+Sonnet [quality gate]) to resumable Anthropic Batches API (50% off both stages); verified submit/poll/fetch/parse/resume live end-to-end3.5h28m2m7.5x
8[engine subsystem] omniscient sweep prep: full local stack bring-up (infra+engine+avian-api+app-web+[engine subsystem]), GCP UUID resolution bug fix in gen_omniscient_profiles, end-to-end spot-check validation5.0h45m2m6.7x
9Update [engine subsystem]-service docs for LLM caching+batching (CHANGELOG/CLAUDE/DESIGN/README)1.0h12m3m5.0x
10Background canon reconciliation — 11 canonical values across 8 files (The Deferral)3.0h45m5m4.0x
11Update avian-engine docs for LLM caching+batching changes0.5h8m3m3.8x
12Documentation0.2h5m3m2.4x

Aggregate Statistics

MetricValue
Total tasks12
Total human-equivalent hours124.7
Total Claude minutes371
Total supervisory minutes35
Total tokens9,595,000
Weighted average leverage factor20.2x
Weighted average supervisory leverage factor213.8x
Human-equivalent weeks3.1

Analysis

The highest factor of the day came in at 140.0x and the lowest at 2.4x, a spread of 58.3 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 14.0 of the 124.7 human-equivalent hours, or 11 percent of the day. No single task dominated the total, so the weighted average is representative.

Supervisory time was 35 minutes against 371 minutes of execution, a ratio of about 1 to 11. Supervisory leverage of 213.8x 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.