About

I'm Alex Voloshin. I help engineering teams operationalize AI coding agents.

This blog is the long-form companion to ai-assets, a public repo of 26 agents, 53 skills, 45 eval rubrics, 270 calibration samples, 18 hooks, and 32 user-invocable workflows. It runs on Claude Code, Codex, and Windsurf.

What I write about

Tri-vendor parity. Almost every public voice in agentic dev is locked to one runtime. The cross-vendor view is rarer, and it tends to age better. Posts track concrete differences between Claude Code, Codex, and Windsurf, plus the patterns that survive on all three.

Eval methodology. Most public talk about agent evaluation is hand-wavy. I write about specific rubrics, calibration sample design, and the failure modes calibration data catches when models drift.

Formal methods. LTL, model checking, and formal requirements review apply to agent specs more directly than people assume. I write about it when I have something concrete to say.

Production war stories. What broke, how I noticed, what the fix turned out to be.

What I skip

Takes about "AI changing everything", tutorials for first-time agent users, and vendor-locked content dressed up as agnostic.

Background

15+ years building production software.