Architecting exponential output.

About AI Leverage

AI leverage is the ratio between the human-equivalent hours a task would have required and the wall-clock minutes it actually took with Claude Code as the implementer. A senior engineer who would need a full week to scope, write, test, and ship a feature might see Claude finish the same work in twenty minutes. That is a leverage factor of roughly 120x. The point is not throughput; it is end-to-end engineering output measured against the work a human would do alone, at the same quality bar, in the same codebase, with the same tests passing.

I track every non-trivial task I complete with Claude Code. The numbers below come from that log. I publish a daily record of what I built, how long it took me to supervise, and what a human would have spent on the same work. The public record keeps me honest. Estimates are uncomfortable to defend in writing, and that discomfort is exactly the test the metric needs.

Articles on Leverage and AI-Assisted Engineering

• articleThe Leverage Factor: Measuring AI-Assisted Engineering OutputFebruary 24, 202610 min read
• articleAgentic Coding and Decision Fatigue: The Cognitive Cost of Supervising AIFebruary 28, 202621 min read
• articleThe Leverage Factor, Part 2: Defending the NumbersMarch 2, 202618 min read
• articleAgentic Coding, FOMO, and Flow State AddictionMarch 2, 202616 min read
• articleAutomated TDD with Claude Code: Testing Strategy for AI-Assisted EngineeringMarch 3, 202613 min read
• articleEleven Years in Forty-One DaysApril 4, 20269 min read
• articleOverlooked Productivity Boosts with Claude CodeOctober 9, 202513 min read