50% Compliance, Not 0%: How a Logging Spike Almost Triggered the Wrong Architecture Rewrite

Dev.to / 5/6/2026

💬 OpinionDeveloper Stack & InfrastructureIdeas & Deep AnalysisTools & Practical UsageModels & Research

Key Points

  • The author initially believed a prompt augmentation pipeline was fully non-compliant (compliance=0%), which nearly led to a 27-hour architecture rewrite.
  • A logging-driven spike revealed the “0%” result was caused by a regex bug, and the system’s real compliance was around 50%.
  • The post documents a structured, chaptered investigation approach (with figures such as a decision-matrix quadrant chart and a 7-version timeline) to pinpoint where measurements and assumptions went wrong.
  • It also shares related engineering lessons, including false-negative debugging and broader LLM engineering practices, with references to a reproducible spike and project resources.
  • The article frames these findings as actionable guidance to help others avoid similar wasted effort when debugging LLM systems and agent architectures.

TL;DR: I thought my prompt augmentation track was completely broken (compliance=0%) and almost spent 27 hours rewriting plan-and-execute. Turns out the 0% was a regex bug — real value was 50%.

7 chapters + 4 figures (hero / decision matrix quadrantChart / 7-version timeline / data delta tables).

Full version (with mermaid quadrantChart + reproducible spike commands):
https://github.com/franciseliang99-dot/web-agent/blob/main/docs/blog-drafts/2026-05-w5c2-spike-story-final-en.md

Project repo (MIT, ⭐ welcome):
https://github.com/franciseliang99-dot/web-agent

LLM engineering + spike decision methodology + Chrome GPU deadlock diagnosis + regex false-negative lessons. Might save you 27h.