About the author: I'm Charles Sieg, a cloud architect and platform engineer who builds apps, services, and infrastructure for Fortune 1000 clients through Vantalect. If your organization is rethinking its software strategy in the age of AI-assisted engineering, let's talk.
Thirty tasks across seven projects. Infrastructure provisioning, multi-platform feature development, a complete site template redesign, PDF parsing improvements, MCP server conversions, and a full SEO overhaul. The tfadm MCP server conversion hit 133x, the highest single-task leverage factor I have recorded.
Task Log
| # | Task | Human Est. | Claude | Sup. | Factor | Sup. Factor |
|---|---|---|---|---|---|---|
| 1 | Convert infrastructure management tool to MCP server (19 tools, 2 resources, 3 prompts) | 40h | 18m | 5m | 133.3x | 480.0x |
| 2 | Consolidate compute architecture and write 10 serverless handlers with design documentation | 24h | 25m | 5m | 57.6x | 288.0x |
| 3 | Implement ELO-based scoring level system across shared UI library and 3 client applications | 24h | 25m | 8m | 57.6x | 180.0x |
| 4 | Build decision fatigue compensation model with service layer, schemas, router, and 8 MCP tools | 16h | 18m | 5m | 53.3x | 192.0x |
| 5 | Multi-platform mastery tracking view across 3 client applications with shared UI patterns | 12h | 15m | 3m | 48.0x | 240.0x |
| 6 | PDF parser: multi-column detection, font metrics, italic handling, skill category merging | 20h | 25m | 5m | 48.0x | 240.0x |
| 7 | Certification-to-domain mapper with 80+ static patterns across 10 providers and 42 tests | 12h | 15m | 3m | 48.0x | 240.0x |
| 8 | Modern site template design: 20+ template files with 900-line SCSS, dark mode, scroll animations | 16h | 22m | 5m | 43.6x | 192.0x |
| 9 | Multi-platform enrollment abandonment with confirmation modals across 3 clients | 6h | 10m | 3m | 36.0x | 120.0x |
| 10 | Feature view implementation for iOS (7 files: persistence, phase, state, view, routing, dashboard) | 4h | 8m | 3m | 30.0x | 80.0x |
| 11 | Feature view implementation for web with full page component and indicator | 4h | 8m | 3m | 30.0x | 80.0x |
| 12 | Feature view implementation for desktop (6 files with module CSS and wiring) | 4h | 8m | 3m | 30.0x | 80.0x |
| 13 | Mastery tracking screen for desktop (new screen, CSS, routing) | 6h | 12m | 3m | 30.0x | 120.0x |
| 14 | Mastery tracking page for web (persistence, component, CSS, wiring) | 6h | 12m | 5m | 30.0x | 72.0x |
| 15 | iOS onboarding flow with file upload and branching path selection | 4h | 8m | 3m | 30.0x | 80.0x |
| 16 | Research and compile hands-on labs master list across 13 certification exams | 16h | 35m | 5m | 27.4x | 192.0x |
| 17 | Full SEO overhaul: OG tags, Twitter Cards, JSON-LD, canonical URLs, sitemap, RSS auto-discovery | 8h | 20m | 5m | 24.0x | 96.0x |
| 18 | Container orchestration configs for 2 services (16 files: task defs, security groups, load balancer rules, DNS, IAM) | 6h | 15m | 3m | 24.0x | 120.0x |
| 19 | Infrastructure stack with 8 configurations, target, version, and deployment for workloads API | 6h | 15m | 3m | 24.0x | 120.0x |
| 20 | Analyze existing infrastructure-as-code and map to new configuration system with gap analysis | 4h | 10m | 2m | 24.0x | 120.0x |
| 21 | Desktop onboarding flow: file upload and review with branching path selection | 3h | 8m | 3m | 22.5x | 60.0x |
| 22 | Update backend MCP server to streaming HTTP transport with retry logic | 1.5h | 4m | 3m | 22.5x | 30.0x |
| 23 | Write comprehensive system documentation with data schemas, fatigue model, and query reference | 3h | 8m | 2m | 22.5x | 90.0x |
| 24 | Enrollment abandonment feature for iOS (auth client, state management, dashboard) | 1.5h | 4m | 3m | 22.5x | 30.0x |
| 25 | Mastery tracking view for iOS (7 files: persistence, phase, state, view, routing) | 4h | 12m | 5m | 20.0x | 48.0x |
| 26 | Enrollment abandonment feature for web application | 2h | 6m | 3m | 20.0x | 40.0x |
| 27 | Fix 24 test failures across 3 service test suites with test infrastructure repairs | 4h | 12m | 2m | 20.0x | 120.0x |
| 28 | Onboarding flow with file upload and path selection across 3 clients | 8h | 25m | 5m | 19.2x | 96.0x |
| 29 | Enrollment abandonment feature for desktop across 6 files | 1.5h | 5m | 3m | 18.0x | 30.0x |
| 30 | Convert 3 MCP servers from stdio to streaming HTTP transport with connection pooling | 4h | 18m | 5m | 13.3x | 48.0x |
Aggregate Stats
| Metric | Value |
|---|---|
| Total tasks | 30 |
| Human-equivalent hours | 270h (33.8 working days) |
| Claude wall-clock time | 426m (7.1h) |
| Supervisory time | 114m (1.9h) |
| Tokens consumed | ~2,443,000 |
| Weighted avg leverage factor | 38.0x |
| Weighted avg supervisory factor | 142.1x |
Analysis
The tfadm MCP server conversion stands out at 133.3x. That task involved wrapping an existing infrastructure management tool with 19 MCP tools, 2 resources, and 3 prompts. The high leverage came from the mechanical nature of the work: each tool followed the same pattern, and the AI could stamp them out without much variation. A human would have spent the better part of a week on the boilerplate alone.
Multi-platform feature rollouts continued to drive volume. Three separate features (release notes view, mastery tracking, enrollment abandonment) each got deployed across iOS, web, and desktop clients. The per-platform leverage ranged from 18x to 30x, but the aggregate tasks (all 3 clients at once) hit 36x to 48x because the AI could carry context across platforms and adapt patterns without re-learning the architecture.
Infrastructure work clustered around 24x: container orchestration configs, infrastructure-as-code analysis, and stack provisioning. These tasks involve a lot of cross-referencing documentation and generating repetitive configuration files. Lower leverage than pure code generation, but still significant given the precision required.
The lowest factor was 13.3x for the MCP transport conversion. That task required understanding three different server implementations and their connection models, then making architectural decisions about retry logic and connection pooling. More thinking, less typing.
The modern template redesign at 43.6x was notable: 20+ Jinja template files plus a 900-line SCSS stylesheet with full dark mode support and scroll animations, produced in 22 minutes. A front-end developer would spend two full days on that.
Let's Build Something!
I help teams ship cloud infrastructure that actually works at scale. Whether you're modernizing a legacy platform, designing a multi-region architecture from scratch, or figuring out how AI fits into your engineering workflow, I've seen your problem before. Let me help.
Currently taking on select consulting engagements through Vantalect.