About

I’m George Diab. I’ve been building software since 1998, starting as a web and database developer at Bank of America while I finished my degree at UNLV. Most recently I was an Engineering Manager and Architect at Codingscape, and before that I led technology teams at RevUnit and Credit One Bank. I live in Las Vegas.

I write here about AI-assisted engineering, which for me means the unglamorous version: agents working test-first against real code, verified with tests and linting, then reviewed by a different model than the one that wrote it so the blind spots don’t line up. Most posts come out of something that annoyed me enough to fix.

What I’ve shipped lately

Two iOS apps are live on the App Store, AnxiLog+ and Arbatash, both built and maintained solo and almost entirely with agents. Arbatash is getting multiplayer now, still agent-built.

The tooling is public at github.com/gdiab. linkledger-cli is a local-first memory layer for agents. agent-status-ledger reads my own Claude Code and Codex session transcripts, correlates them against git commits, and produces a daily standup for the agents I run, with every claim tagged by evidence level so an agent’s self-report never gets confused for verified work. There are a handful of Claude Code skills in there too, including the 12-Factor assessment skill I use in client work.

My notes live in a plain-Markdown second brain in a git repo, operated by custom skills and MCP tooling. I wrote about that one here.

The consulting practice

I run a solo AI consulting practice at georgediab.net: diagnosis first for small businesses, and a fractional AI architect tier for teams that want someone who lives in the tools. It’s a separate site on purpose. The tools I recommend there are the tools I use every day, which is also why I’m comfortable telling someone they don’t need them.

Talking to me

The writing here has been going since 2011, though the current run started in 2026. If a post is useful, or wrong, I’d like to hear about it: me@georgediab.com.