15 things every software person should know

A teaching guide for AI-assisted software development. AI can write code fast, but these ideas are how you judge whether that code is any good. Click any note to dig in.

Inspired by the list in “My Agentic Engineering Workflow (after 15,700 sessions)” by David Ondrej. Explanations and examples on this page are an expanded companion, not a transcript.

How to use this as a teaching guide
  • Cover one or two topics per session. Each note has a plain-English definition, an everyday analogy, a concrete example, and a section on how the idea shows up when you work with an AI assistant.
  • Run the "Try this prompt" live. Paste it into your AI tool against a real project and talk through what comes back, including what the AI missed.
  • Use the discussion questions as a code-review checklist. They work well when you review AI-generated pull requests.
  • Follow the "Related" links. These ideas cluster: idempotency, queues, and race conditions go together, and so do least privilege and blast radius.
  • Deep links work. Add the topic to the URL (for example #idempotency) to send someone straight to one note.

Use ← → keys to move, Esc to close