Claude Code

Although I went to a secondary school where we were taught programming and my graduation project was in Assembly (x51), I wouldn't call myself a programmer. At university I didn't put much into it either and did the bare minimum. Though I did really enjoy the subjects on algorithms and their design. And then the moment arrived… I knew Claude, but hadn't been using it. I was focused on other things like generating images and fine-tuning a custom model (one that would create presentations the way I make them). Meanwhile things moved on and agents arrived. At that point my refined custom GPT for making presentations was driving me crazy — every new GPT version broke something. Around that time the US Government put a ban on Anthropic — which must have meant something — so I tried Claude Code.

Wow! It solved what I needed, then something else, and then more… up until they broke it before Mythos it was an incredibly capable assistant. You have to approach it responsibly of course — when it's writing code you need to write tests first. That month I attended Bodík's lecture and if I took one thing away from it, it was "Write tests." So I write them, with as high coverage as possible.

Since then I've been creating like mad. Containers, infrastructure, small applications, helpers for attribution or forensic analysis. I converted almost all my Czechitas presentations to a template a colleague created. Improving Docker containers, optimising Actions. Once it even saved me from a Supply Chain attack (Trivy).

I combine Code, Codex, and Copilot. Each has its advantages.