Skip to main content
AI09 May 2026

Debugging and coding

38 tasks. May 9, 2026 closed at 26.9x weighted leverage across 632.5 human-equivalent hours in 1,410 minutes of wall-clock time. Supervisory leverage came in at 223.2x.

That is 15.8 weeks of human-equivalent throughput in 23.5 hours. The ceiling was 85.7x; the floor was 6.7x. 22 of the 38 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1iOS web-parity rebuild: 8 phases — phase machine restructure, AvianAppShell+TopNav, launch routing fix, HomeView (slim hub), multi-course Dashboard, CoursesView+CourseDetailView, SettingsView split,…50.0h35m8m85.7x
2Debugging60.0h44m6m81.8x
3Infrastructure80.0h70m8m68.6x
4iOS Help fixes: tab strikethrough fix (overlay alignment), port 40 help guides verbatim from helpCenterDocs.ts → HelpDocs.swift (sidebar+content layout, iPhone sheet), embed 5 legal docs (privacy, terms,…24.0h22m4m65.5x
5iOS app facelift: SwiftUI design system port (tokens, typography, 14 components), Plus Jakarta Sans bundling, AvianTheme shim, migrate 6 high-traffic views (LoginView, ResultsView, WelcomeView, ProfileView,…40.0h50m6m48.0x
6iOS Settings/Profile/Help web parity: fix sign-in button (TopNav overflow on iPhone), build new HelpView (5-tab Overview/FAQ/Guides/WhatsNew/Legal + .help phase + bug-report bridge), refactor ProfileView into…18.0h24m5m45.0x
7Deployment9.0h14m4m38.6x
8Upgrade english.accelastudy.ai (IELTS/TOEFL) to multi-page subscription product site: standalone /pricing/ page with comparison table & FAQ, switched nav to standalone routes, live header CTA, sister-site…5.0h8m3m37.5x
9Major engine fix + audit expansion. (1) Built backfill_question_pair_ids.py - deterministic pair_id linkage backfill across 234 [content generation] domain packages. Drove pair_id coverage from 32.3% to 54.1%…16.0h28m6m34.3x
10Deployment32.0h65m2m29.5x
11MEDIUM cleanup wave: 9 reduced-motion guards + 8 sr-only utilities + 594 h1->h2 codemod demotions across 241 files + 50 input-adjacent-label codemod pairings + 5 hand-fixes (BillingPage h1, Blog.jsx h1s,…14.0h30m1m28.0x
12Fleet-wide a11y fix sweep across 56 UI repos: 2,460 HIGH findings fixed (2,235 via console-sim label-pairing codemod + 17 manual + 5 wave-1 activities-react + 59 wave-3 client apps + 136 wave-4 tools fleet +…60.0h130m5m27.7x
13Final wave: clear remaining 192 HIGH a11y findings — patched 71 stale cloudops dist HTML files with lang=en (Python sed), dispatched focused subagent to fix 118 of 120 console-sim view-level…16.0h35m1m27.4x
14Port web Help Center guide articles to iOS (40 docs, 7 categories) and rebuild Guides tab with sidebar+content layout8.0h18m5m26.7x
15Drafted Making AccelaStudy AI Accessible to All across 3 sites — charlessieg.com (~3500-word technical deep-dive with mermaid wave diagram + 6 reference tables + concrete codebase counts: 2185 TSX/JSX files,…12.0h28m2m25.7x
16Drove the avian-domains spec audit from 257 LOW (post-prior-pass) to absolute zero across all four severities. Tightened audit_specs.py (broadened verb whitelist, fixed cross-domain prefix detection,…14.0h35m4m24.0x
17Refactor CLAUDE.md chain: extract [ip] checklist, repo map, ADR rules, SSM, domain inventory, [engine subsystem] into subtree files; relocate API keys to mode-600 env file outside prompt2.5h7m3m21.4x
18Built deterministic Python accessibility-audit checker (15 rules, 56-repo discovery, brace/quote-aware JSX tokeniser, JSON+MD output, mode-aware exit codes); updated accessibility-audit.md with Phase 0 spec…24.0h70m4m20.6x
19Content audit reconciliation: dropped 30 of 40 findings (all 11 CRITICALs + all 14 catalog/canonical MEDIUMs + 5 LOWs). Wrote _fix_exam_code_dups.py to remap 33 collided exam_code values to vendor-correct…4.0h12m2m20.0x
20Restructure iOS ProfileView to mirror web Profile 4-tab layout (hero + Profile/Resume/Subscription/Account)6.0h18m5m20.0x
21Design and frontend18.0h55m5m19.6x
22Infrastructure16.0h50m4m19.2x
23Build CoursesView and CourseDetailView for AccelaStudy iOS app (web parity)6.0h22m8m16.4x
24Port web legal docs into iOS HelpView — embed privacy policy, terms, accessibility, trademarks, and credits as scrollable in-app LegalNode tree; rebuild Legal tab with iPad sidebar and iPhone sheet flow6.0h22m5m16.4x
25Deployment24.0h90m8m16.0x
26Coding4.0h16m2m15.0x
27Mode-1 a11y audit on avian-app-web + 6 React libs: 11 HIGH fixes (RemoteBanners aria-live, HelpCenter dialog focus, LabConsole tab pattern, ProgressBar/ExamScoreReport/Sidebar progress+log roles,…12.0h55m3m13.1x
28Deployment7.5h36m4m12.5x
29Build real SettingsView for iOS app mirroring web settings page (appearance, language, study prefs, voice, accessibility, privacy, about sections)3.0h15m5m12.0x
30Migrate WelcomeView.swift to AVIAN iOS design system (AvianTokens, AvianFont, AvianButton, AvianCard, AvianEmptyState)1.5h8m3m11.2x
31Rebuild DashboardView.swift to multi-course portfolio matching web Dashboard.tsx4.0h22m5m10.9x
32Migrate DashboardView.swift (1521 lines) to AVIAN iOS design system tokens, typography, and components3.0h18m3m10.0x
33accelastudy.ai canon-swap propagation: replace hardcoded counts with [value] placeholders across press, about, how-it-works, faq, accessibility, pricing, courses, free, 5 feature pages and shared pricing-card…5.0h35m2m8.6x
34Testing4.0h30m1m8.0x
35Activities catalog reorg (default+5 addons across 62 categories) + 4 web bug fixes (data-driven Service Match applicability, Privacy footer link, Bio Profile→Resume, unenroll→autopilot cascade) + Settings…14.0h110m18m7.6x
36Infrastructure5.0h40m2m7.5x
37Design and frontend3.0h25m3m7.2x
38Restructure SettingsView.swift to 4-tab layout (General/Autopilot/Accessibility/Privacy) with Audio section, extended-time toggle, Privacy Policy link, and SoundManager integration2.0h18m5m6.7x

Aggregate Statistics

MetricValue
Total tasks38
Total human-equivalent hours632.5
Total Claude minutes1,410
Total supervisory minutes170
Total tokens6,445,000
Weighted average leverage factor26.9x
Weighted average supervisory leverage factor223.2x
Human-equivalent weeks15.8

Analysis

The highest factor of the day came in at 85.7x and the lowest at 6.7x, a spread of 12.9 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 50.0 of the 632.5 human-equivalent hours, or 8 percent of the day. No single task dominated the total, so the weighted average is representative.

Supervisory time was 170 minutes against 1,410 minutes of execution, a ratio of about 1 to 8. Supervisory leverage of 223.2x 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.