How to think about performance
Vocabulary, percentiles, and the three questions every performance investigation reduces to.
Notes on software architecture, cloud systems, and engineering with AI.
Vocabulary, percentiles, and the three questions every performance investigation reduces to.
What auditing over a thousand of my own AI coding sessions taught me about where the tokens actually go.
Raw audit report behind the session-efficiency DOs and DON'Ts — 854 sessions, 916 findings, and the ranked fixes.
A workflow for the part of shipping that tests cannot reach. In the AI era, code production has multiplied but verification has not, and a skill is needed to capture what to verify.
A 55-slide deck on the engineering discipline for working with AI coding agents — Claude Code, skills, and the systems knowledge behind tokens and caching.
A 15-slide deck through the seven phases of building trustworthy software with AI agents, presented as a blog post.
Code generation is abundant; engineering judgment is not. A seven-phase lifecycle for building trustworthy software with AI agents.
Agentic software engineering-এর পুরো lifecycle — idea থেকে architecture, build, review এবং production feedback পর্যন্ত।