Skip to main content
AI29 May 2026

Testing and infrastructure

7 tasks. May 29, 2026 closed at 32.2x weighted leverage across 94.0 human-equivalent hours in 175 minutes of wall-clock time. Supervisory leverage came in at 313.3x.

That is 2.4 weeks of human-equivalent throughput in 2.9 hours. The ceiling was 80.0x; the floor was 9.5x. 6 of the 7 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1Aegis Phase 1 — scanner foundation (Scanner ABC + Finding model + content-addressable dedup hash + lazy registry) plus all 7 scanner integrations (semgrep/bandit/detect-secrets/pip-audit/safety/checkov/trivy)…60.0h45m1m80.0x
2Root-caused cloud-cert quality instability (42 AWS/GCP/Azure pkgs, 3 repair campaigns) + committed 301 engine re-stamps + regenerated 121 PMP single-option questions6.0h10m3m36.0x
3[content generation] field-preservation remediation (task 1/7): traced parser->adapter->writer path across [ip-cluster]-runtime + [ip-cluster]; added raw exam_metadata passthrough + ConfigDict(extra=allow) +…4.5h13m2m20.8x
4Testing4.0h15m2m16.0x
5origin to [engine subsystem] monorepo rename — [ip-cluster]_runtime and [ip-cluster] packages + 2 repo dirs + [ip-cluster] to [engine subsystem] CLI + path-deps/uv.locks across 8 repos — infra deferred —…9.0h35m6m15.4x
6[engine subsystem] rebuild — shared in-process pipeline_core (run_phase/run_pipeline) + made lessons/questions/[engine subsystem] stub runners real + write_question_bank + pipeline job kind + [engine…7.0h35m3m12.0x
7[engine subsystem] rebuild finish — retired subprocess paths (orchestrator in-process + JobStore-backed [content generation]/[engine subsystem] APIs + deleted SynthesisManager/TribunalManager) + standalone…3.5h22m1m9.5x

Aggregate Statistics

MetricValue
Total tasks7
Total human-equivalent hours94.0
Total Claude minutes175
Total supervisory minutes18
Total tokens1,398,000
Weighted average leverage factor32.2x
Weighted average supervisory leverage factor313.3x
Human-equivalent weeks2.4

Analysis

The highest factor of the day came in at 80.0x and the lowest at 9.5x, a spread of 8.4 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 60.0 of the 94.0 human-equivalent hours, or 64 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 18 minutes against 175 minutes of execution, a ratio of about 1 to 10. Supervisory leverage of 313.3x 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.