audit-context-building

Build deep context by applying First Principles, 5 Whys, and 5 Hows at a micro scale during line-by-line code analysis, linking insights from code blocks to functions, modules, and the entire system architecture.

Context loss causes hallucinations because the mental model becomes unstable. Maintaining an explicit, evolving mental model based on micro-level code evidence prevents missing critical assumptions and invariants.

The best way to analyze complex codebases for bugs is to examine code block-by-block, applying First Principles to meticulously analyze each function, its inputs, outputs, and internal logic, ensuring no critical assumptions are missed.

Apply First Principles and 5 Whys during code analysis by breaking down each line and block to its foundational logic, asking why and how operations occur, then linking these micro insights to the entire system's architecture.

Yes, ultra-granular code analysis works for identifying reasoning hazards in smart contracts by requiring a stable, explicit mental model that evolves with new evidence from line-by-line examination of inputs, outputs, and internal logic.

Avoid line-by-line code analysis when needing a quick, high-level overview rather than a deep, stable mental model, as this micro-scale approach requires meticulous examination of every function and code block to prevent context loss.