
Agent Observability: Debugging Reasoning, Not Code
Agent observability == debugging reasoning (≠ code). A practical guide to runs, traces, and threads, and how the same traces power agent evaluation.
Writing about software engineering, AI systems, performance, product tradeoffs, and the practical side of shipping modern software.
Filter the notes by the kind of engineering problem they cover. Tags are editorial labels, not raw CMS categories.

Agent observability == debugging reasoning (≠ code). A practical guide to runs, traces, and threads, and how the same traces power agent evaluation.

A candid build note on CIBI: the product constraint, the Expo React Native stack, OCR, local-first storage, and why the app tries very hard not to sound like a bank.

Agent skills are supposed to make you faster. Mine had become a maintenance problem, so I built a small reversible workflow before deleting anything.

As developers, we have a profound impact on UX, not just through collaboration with designers but in key areas where we have full control: performance, accessibility, and usability.

WaterLator is a web application that calculates your daily and workout optimal fluid intake. In this post I go into the story, steps and behind the scenes of making WaterLator

It is really challenging to keep up with creating content and the amount o work load I usually have. In this blog post I go through my thought process and why I am giving more priority to blogs.