Engineering Principles

I define the problem and completion criteria before optimizing implementation speed, then keep code, documentation, tests, and review aligned to those criteria. I use AI tools for problem decomposition and implementation support while retaining direct responsibility for requirements, technical decisions, code review, test criteria, and release decisions.

Core Principles

1. Separate Symptoms from Causes and Set the Scope of the Fix

I do not assume that a slow or failing path has one cause. As in the ClumL outbound LLM API rate-limiting fix, which separated a concurrency race from fixed-window waiting, I isolate causes before deciding what to change and what to leave unchanged.

2. Make Completion Criteria Testable

A successful API response alone is not enough. I check response shapes, database writes and reads, screen routing, and error handling. The TmaxCloud pre-deployment testing for generated APIs, GlueSQL regression tests, and Coupler release checks apply this principle.

3. Capture Existing Behavior before Replacing an Implementation

Before replacing a library or internal implementation, I record current behavior in tests. In the Chrono-to-Jiff migration, I captured timestamp behavior before separating the implementation change from dependency cleanup, then checked affected screens and server compatibility.