Skip to main content
AI04 August 2026

Documentation and coding

40 tasks. August 4, 2026 closed at 14.0x weighted leverage across 786.5 human-equivalent hours in 3,362 minutes of wall-clock time. Supervisory leverage came in at 386.8x.

That is 19.7 weeks of human-equivalent throughput in 56.0 hours. The ceiling was 62.8x; the floor was 1.7x. 27 of the 40 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1Documentation68.0h65m2m62.8x
2Audit and review28.0h30m3m56.0x
3Documentation32.0h40m5m48.0x
4Infrastructure28.0h38m2m44.2x
5Create /fable-planner skill - Fable-gated program planner producing four docs and phased Opus/Sonnet execution plans2.5h4m5m37.5x
6Infrastructure32.0h63m3m30.5x
7Deployment26.0h68m3m22.9x
8Fable planning: model-scorecard program (research + four docs x3 repos + phased plan + state + Fable review gate)26.0h70m2m22.3x
9Documentation34.0h96m6m21.2x
10Documentation14.0h42m3m20.0x
11Testing16.0h50m5m19.2x
12Model-Scorecard program: P0 baseline + P1.T1 benchmarks package + P4 phase-scoped routing live on the daemon + 5 production defects found/fixed with regression tests + 2 real benchmark runs130.0h445m12m17.5x
13Testing16.0h57m2m16.8x
1413-site UI/UX audit remediation: canon-driven facts fleet-wide, dropdown a11y rebuild, responsive tables, CTA/posture truthing, [unreleased product] shell migration, GED+Certs compact pilots, Playwright+axe…100.0h362m3m16.6x
15iOS plan approval pass: decisions D1-D5 recorded, WP rewrites, STATE+ledger scaffolding, ASC runbook, labs-deferral card2.5h10m2m15.0x
16iOS productization steer: D9 per-product builds + icon family, D10 price parity, WP-2.0/4.0 added, ASC runbook rewrite2.0h8m2m15.0x
17Coding7.0h28m2m15.0x
18Electron: canon-driven LaunchPromoBanner + 7 stale-content fixes across 2 surfaces, 33 tests (A-ELEC Opus agent)8.0h33m2m14.5x
19Phase-2 content repair: R4 complete, dedupe 14665->5265, deictic stems -26%/explanations -41%, math markup pushed, 2 transform bugs + 1 platform bug caught pre-push, [cost]spend (R-EXEC Opus agent)26.0h115m2m13.6x
20avian-content-checks: JudgeClient/BudgetMeter thread safety + gate_report schema-pin test (A-CHECKS Opus agent)5.0h24m2m12.5x
21run-smoke self-heal + smoke-user deletion forensics (audit_logs empty — actor unidentifiable) + 3 follow-up cards (A-SMOKE Opus agent)6.5h32m2m12.2x
22[engine subsystem]: benchmark spend->ledger + findings current-version default + hub key-scrub done right (kept benchmark keys, filed doc-contradiction card) (A-[engine subsystem] Opus agent)9.0h45m2m12.0x
23iOS plan: WP-1.4 src/ reorganization added (sequencing, registry/tooling path scope, exit gates)1.0h5m1m12.0x
24Testing1.5h8m2m11.2x
25iOS parity WP-0.2: catalog stale-while-revalidate (disk cache, explicit load state, skeleton/error+Retry), host out of Swift into xcconfig, Associated Domains repointed; 20 tests12.0h64m4m11.2x
26[ip-cluster] first-3 cards: [engine subsystem] judge-role swap + regression guard; ANN debt cleared in 2 test files; [engine subsystem] CLAUDE.md benchmark-keys exception; board-state corrected (56/81 already…5.0h27m1m11.1x
27iOS parity WP-4.0: environment-scoped App Store Server Notifications V2 (per-env SignedDataVerifier + data-layer guard), 66 tests incl. real x5c chain10.0h58m4m10.3x
28Testing18.0h105m6m10.3x
29Phase-2 drain: dedupe 14665->13, deictic stems -58%, explanations -89%, ISACA matcher root-cause corrected + husks repaired, closing audit locked, 3 findings filed (R-FINISH Opus agent)32.0h190m2m10.1x
30Design and frontend12.0h85m3m8.5x
31iOS parity WP-0.3: anonymous public domain catalog in avian-api (optional-user dep, public_only forced for anonymous) + iOS OIDC retirement note; live-probed electron/[unreleased product] prod sign-in breakage4.0h33m4m7.3x
32Audit and review5.0h43m2m7.0x
33Debugging5.0h46m2m6.5x
34Debugging14.0h142m4m5.9x
35Deployment10.0h105m2m5.7x
36Testing5.0h59m3m5.1x
37Engine: unmapped question-type WARNING + evidence-based code-scenarios->mcq mapping, fleet re-audit clean (A-ENG Opus agent)3.5h43m2m4.9x
38Debugging14.0h240m2m3.5x
39iOS parity WP-0.6: prod-API posture + launch-time unreachable-gateway banner (3s Task-cancellation deadline, dedicated session, offline/unreachable mutual exclusion); root-caused an XCUIAccessibilityAudit…10.0h175m4m3.4x
40Product-listing regroup across 14 sites: Available Now / Free Forever / Coming Soon (launch-date order) on shared products page + gateway router, August→September launch slip via canon, full-family redeploy +…6.0h209m2m1.7x

Aggregate Statistics

MetricValue
Total tasks40
Total human-equivalent hours786.5
Total Claude minutes3,362
Total supervisory minutes122
Total tokens28,090,247
Weighted average leverage factor14.0x
Weighted average supervisory leverage factor386.8x
Human-equivalent weeks19.7

Analysis

The highest factor of the day came in at 62.8x and the lowest at 1.7x, a spread of 36.4 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 68.0 of the 786.5 human-equivalent hours, or 9 percent of the day. No single task dominated the total, so the weighted average is representative.

Supervisory time was 122 minutes against 3,362 minutes of execution, a ratio of about 1 to 28. Supervisory leverage of 386.8x 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.