Tagged

Performance

3 posts

Diagnostic kill-switches: A/B in production with one CLI flag

A two-minute production experiment ruled out 31 detection rules as the cause of a CPU saturation, without redeploying or shipping new code. The pattern is one CLI flag per expensive subsystem. This post is about the discipline that pays for itself the first time you need to bisect under fire.

24 Apr 2026 · 7 min read

AI EngineeringEngineering Practices

Patterns of agent driven code development

Six lessons from one saturation incident, all about how to keep agent velocity without paying for it in production fires.

23 Apr 2026 · 11 min read

PerformanceProfiling

How to read a flamegraph: a 10-minute guide for production debugging

Flamegraphs visualize where a program actually spends its CPU. This guide walks through what the pixels mean, how to capture one without being a perf wizard, three worked examples, and the four patterns that show up over and over in real production code.

20 Apr 2026 · 9 min read