Skip to main content
AI25 April 2026

Mostly coding

37 tasks. April 25, 2026 closed at 136.0x weighted leverage across 2,288.5 human-equivalent hours in 1,010 minutes of wall-clock time. Supervisory leverage came in at 1373.1x.

That is 57.2 weeks of human-equivalent throughput in 16.8 hours. The ceiling was 342.9x; the floor was 10.3x. 37 of the 37 entries came from a single project.

Task Log

#TaskHuman Est.ClaudeSup.Factor
1Tier-1 console-sim Phase 18: GCP identity & compute (ResourceMgr IAMPolicyAnalyzer OrgPolicy VPCSC BeyondCorp WorkloadIdentityFed AssuredWorkloads IAP ComputeEngine AppEngine CloudFunctions CloudRun GKE…160.0h28m3m342.9x
2Tier-1 console-sim Phase 14: Azure data & storage (Azure SQL Cosmos PostgreSQL MySQL Redis Storage Blob Files ADLS Gen2 Queue Storage Managed Disks) — 10 services tier=full140.0h30m3m280.0x
3Tier-1 console-sim Phase 17: Azure security & mgmt (Defender Sentinel KeyVault Compliance Policy ResourceLocks MgmtGroups Monitor LogAnalytics AppInsights Advisor Portal ResourceMgr ARM Bicep ResourceGroups…140.0h30m3m280.0x
4Tier-1 console-sim Phase 12: Azure networking (VNet NSG LB AppGW Front Door VPN ExpressRoute Firewall DDoS TM DNS Bastion VirtualWAN PrivateEndpoint NAT NetworkWatcher ServiceEndpoints) — 17 services tier=full160.0h35m3m274.3x
5Tier-1 console-sim Phase 16: Azure DevOps & app platform (AzureDevOps Pipelines Repos Artifacts GitHubActions AppConfig AppService AzureFunctions LogicApps APIM SignalR EventGrid EventHubs ServiceBus…140.0h32m3m262.5x
6Coding200.0h46m3m260.9x
7Tier-1 console-sim Phase 22: Workspace + IaC + final polish (Drive Gmail GoogleVault AlertCenter AdminConsole Firebase Marketplace) — 7 services tier=full + IaC verified + final audit80.0h19m3m252.6x
8Tier-1 console-sim Phase 13: Azure compute & containers (VMs VMSS AKS ACR ContainerInstances ContainerApps SiteRecovery AzureBackup Migrate) — 9 services tier=full90.0h25m3m216.0x
9Tier-1 console-sim Phase 11: Azure identity & access (Entra ID Conditional Access PIM RBAC Managed Identities MFA SSPR MS Graph M365 Defender) — 9 services tier=full120.0h35m3m205.7x
10Tier-1 console-sim Phase 19: GCP networking & storage (VPC Firewall Peering SharedVPC Armor CDN LB DNS NAT VPN Interconnect Router PSC HierarchicalFW FlowLogs NetworkIntel PacketMirror NEG FirewallInsights…200.0h60m3m200.0x
11Tier-1 console-sim Phase 20: GCP analytics & AI (BigQuery BQML Dataflow Dataproc CloudComposer Datastream DataCatalog AnalyticsHub PubSub LookerStudio VertexAI NLAI VisionAI TFDV) — 13 dashboards covering 27…180.0h57m3m189.5x
12Tier-1 console-sim Phase 15: Azure analytics & AI (Synapse DataFactory StreamAnalytics DataExplorer Purview Fabric Databricks PowerBI AzureML OpenAI AILanguage Speech Vision Search AIServices Face Translator…220.0h75m3m176.0x
13Residual sweep AWS desc_claims - 137 description rewrites across 102 labs (scs/saa/sap/dea/ans/clf/free etc); to 036.0h13m2m166.2x
14Residual sweep GCP/Azure desc_claims - 83 description rewrites across 54 labs (ace/pcse/pca/pcne/pgwa/az-* etc); to 022.0h9m2m146.7x
15224 per-service Guided E2E specs across full-tier services - 4 parallel agents writing to e2e/services/.guided.spec.ts18.0h14m2m77.1x
16console-sim coverage Group Y: ~150 component + 7 SDK tests across 50 dashboards16.0h13m3m73.8x
17Backfill Group A (ace/pcd/pde) - 80 labs normalized expectedActions to canonical names; missing_checkpoint 294 to 060.0h50m3m72.0x
18Backfill Group B (pcde/pcse/pca) - 74 labs normalized + 30 register handlers; missing_checkpoint 375 to 055.0h48m3m68.8x
19Backfill Group C (pmle/pcdb/pcne) - 60 labs normalized + 54 register handlers + Vertex AI/BQML/KMS codemod blocks; missing_checkpoint 555 to 055.0h53m3m62.3x
20Deployment14.0h14m2m60.0x
21Backfill Group D (pgwa/cdl/azure/aws misc) - 67 labs normalized + 111 register handlers; missing_checkpoint 139 to 040.0h43m3m55.8x
22console-sim coverage Group X: ~155 component + 10 SDK tests across 50 dashboards18.0h21m3m51.4x
23Residual sweep mechanical 81 issues - extended derive_assertions and SIMULATION_ACTIONS; mutation_without_property/action_assertion_gap/empty_assertions all to 024.0h30m2m48.0x
24Full content audit across avian-domains, avian-engine, avian-console-sim-react: 919 specs, 218 packages (1.03M questions, 127k nodes), 2,048 labs verified; identified 701-spec [content generation] backlog and…6.0h8m2m45.0x
25Final 4 desc_claims fix - saa-c03 ghf snow checkpoint descriptions neutralized; total audit 0/2048 labs2.0h3m2m40.0x
26Infrastructure4.0h6m1m40.0x
27Build + deploy charlessieg.com semantic search Lambda end-to-end: generate 668-chunk [model] embedding index, package 17MB Python zip with numpy + requests + handler + index, deploy to Lambda…12.0h18m3m40.0x
28console-sim coverage Group Z: ~150 component + 7 SDK tests across 50 dashboards16.0h24m3m40.0x
29console-sim coverage Group W: ~150 component + 5 SDK tests across 50 dashboards16.0h28m3m34.3x
30avian-engine: ship strategy dimensions/drift/forecast/recommender + persistence (1700-line WIP) + verify fingerprint 404->200 deploy16.0h30m2m32.0x
31Redesign template second pass: about.jinja from mockup, article+post split into article_post_template layout (full-width header + 8/4 body/TOC grid), blog template with sidebar replacing posts list,…4.0h12m3m20.0x
32console-sim test infra: 4-shard coverage config, JSDOM stubs, strict Watch sweep, buildspec-test.yml4.0h12m2m20.0x
33Test infra - JSDOM mocks (canvas/matchMedia/ResizeObserver) coverage config strict Watch sweep3.0h10m2m18.0x
34Deployment8.0h30m4m16.0x
35avian-app-web Study Plan tab rename + per-day collapse + activity card wrap fix + deploy verify2.0h8m3m15.0x
36Storm-prep snapshot of in-flight [engine subsystem] repair: promoted 7 finished packages to canonical, S3-synced to 2 backup buckets, wrote tarball + RESUME runbook + skip-aware resume wrapper script1.5h6m2m15.0x
37Infrastructure6.0h35m4m10.3x

Aggregate Statistics

MetricValue
Total tasks37
Total human-equivalent hours2,288.5
Total Claude minutes1,010
Total supervisory minutes100
Total tokens4,109,000
Weighted average leverage factor136.0x
Weighted average supervisory leverage factor1373.1x
Human-equivalent weeks57.2

Analysis

The highest factor of the day came in at 342.9x and the lowest at 10.3x, a spread of 33.3 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 160.0 of the 2,288.5 human-equivalent hours, or 7 percent of the day. No single task dominated the total, so the weighted average is representative.

Supervisory time was 100 minutes against 1,010 minutes of execution, a ratio of about 1 to 10. Supervisory leverage of 1373.1x 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.