Skip to main content
AI19 April 2026

Infrastructure and coding

47 tasks. April 19, 2026 closed at 63.7x weighted leverage across 2,135.5 human-equivalent hours in 2,010 minutes of wall-clock time. Supervisory leverage came in at 865.7x.

That is 53.4 weeks of human-equivalent throughput in 33.5 hours. The ceiling was 174.5x; the floor was 1.7x. 11 of the 47 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1Infrastructure160.0h55m3m174.5x
2Infrastructure160.0h55m1m174.5x
3Infrastructure140.0h50m1m168.0x
4Infrastructure120.0h45m1m160.0x
5Coding120.0h45m1m160.0x
6Infrastructure160.0h60m1m160.0x
7Infrastructure100.0h38m1m157.9x
8Infrastructure90.0h35m1m154.3x
9Deployment140.0h55m3m152.7x
10Infrastructure100.0h40m1m150.0x
11Deployment200.0h80m1m150.0x
12Infrastructure36.0h18m8m120.0x
13Deployment120.0h60m2m120.0x
14Electron DS migration — 42 screens/components converted from CSS Modules + Phosphor + HeroBanner to Tailwind + Lucide + DS primitives; all TS errors fixed; universal macOS DMG built and signed60.0h55m8m65.5x
15Documentation16.0h20m2m48.0x
1662-spec free-tier expansion authored/validated/seed-chained + lesson-asset pipeline foundation (avian-lesson-assets repo, S3+backup buckets provisioned, asset resolver+static mount in engine, topological…120.0h160m25m45.0x
17Expand MCP coverage to full observability surface (metrics, logs, traces, alerts, incidents, SLOs, dashboards)10.0h14m3m42.9x
18Deployment32.0h45m8m42.7x
19Expand MCP coverage across pipelines, deals, activities, tasks, forecasts, and bulk operations9.0h13m2m41.5x
20Debugging16.0h25m5m38.4x
21Replace SSE with WebSockets for real-time list updates6.0h10m4m36.0x
22Electron parity cleanup: deep-link dispatch, TS error fixes (DomainListResponse/EnrichedDomain/EngineStatus), competitive-mock dev-gate, Login.tsx removal, tray stats push, App.tsx router adoption via…10.0h18m2m33.3x
23Replace SSE with WebSockets for real-time list updates6.0h11m3m32.7x
24Fill MCP coverage gaps in filters, push subscriptions, accounts, identities, signatures, contacts, and more8.0h15m2m32.0x
25Coding4.0h8m2m30.0x
26Debugging6.0h14m3m25.7x
27Codex Phase 11 complete: LinkedRepoService + bulk create (11.3), GitSyncWorker with glob filter/frontmatter/link rewriter (11.4), HMAC-signed webhook (11.5), pgvector embeddings + chunker (11.6), Ask Codex…60.0h140m4m25.7x
28Codex Phase 11.1+11.2: ORM models (LinkedRepo/Key/PageEmbedding/AskSession/Turn + extended Page) + SSHKeyService with AWS Secrets Manager/filesystem backends + REST routes10.0h28m2m21.4x
29Labs CDN: provision labs.accelastudy.ai S3 bucket + CloudFront + Route53 + CORS, upload 2048 labs + manifest, boto3 upload script, Terraform stub, Electron lab-cdn module + LabsBrowser screen + nav + SQLite…24.0h90m3m16.0x
30Testing3.0h12m1m15.0x
31Codex linked-repos design: extend requirements/design/plan/testing docs with LinkedRepo + SSH key management + pgvector RAG + Ask Codex chatbot (Phase 11)8.0h35m6m13.7x
32Root-cause + fix Electron @avian/console-sim type resolution (deleted obsolete ambient avian-packages.d.ts that shadowed real exports with any); fix 3 pre-existing errors that were hidden by any types;…8.0h35m2m13.7x
33Testing4.0h18m1m13.3x
34Migrate 9 Electron renderer screens (CompetitiveLobby, WaitingRoom, CompetitiveGame, CompetitiveResults, AutopilotScreen, ReleaseNotes, ProviderMastery, LabConsole, Onboarding) from CSS Modules + HeroBanner +…6.0h28m5m12.9x
35Testing2.0h12m1m10.0x
36Testing2.0h12m1m10.0x
37Test + deploy avian-app-web + build Electron DMG: run full test suites, fix test timeouts, pin transitive deps to CodeArtifact-cached versions, root-cause buildspec bug that published empty @avian/ui 0.1.1…12.0h75m5m9.6x
38Stand up admin-service backend as an engine-proxy for admin.accelastudy.ai — JWKS-verified renkara JWT auth, server-side AVIAN_ADMIN_KEY, full Terraform stack (ALB rule, ECR, CodePipeline), frontend rewired…11.0h75m8m8.8x
39Format-matched practice (4-opt MCQ via /next-pair-mcq); cuts predictor calibration gap 58→37pp8.0h60m3m8.0x
40Testing1.0h8m1m7.5x
41Close predictor calibration gap from 31pp to 10pp via uniform-rotation practice + tighter priors + aligned exam retrieval (Alex v24-v26 iterations)10.0h90m2m6.7x
42Complete Alex [engine subsystem] calibration: 65q exam generator + brain multi-select answering fix; Alex v29 passes 65q at day 90 with predicted 0.901 / actual 83.3% (8pp gap, naturally gated)14.0h180m4m4.7x
43Testing1.0h15m1m4.0x
44Testing1.0h15m1m4.0x
45Testing0.5h8m1m3.8x
46Testing0.5h12m1m2.5x
47Testing0.5h18m1m1.7x

Aggregate Statistics

MetricValue
Total tasks47
Total human-equivalent hours2,135.5
Total Claude minutes2,010
Total supervisory minutes148
Total tokens9,487,000
Weighted average leverage factor63.7x
Weighted average supervisory leverage factor865.7x
Human-equivalent weeks53.4

Analysis

The highest factor of the day came in at 174.5x and the lowest at 1.7x, a spread of 104.7 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 160.0 of the 2,135.5 human-equivalent hours, or 7 percent of the day. No single task dominated the total, so the weighted average is representative.

Supervisory time was 148 minutes against 2,010 minutes of execution, a ratio of about 1 to 14. Supervisory leverage of 865.7x 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.