Start here: from zero to confident with AI tools
You don't need maths or a machine-learning background. You need to understand a few ideas about how these models behave, and then practise using the tools well. This page sets you up: what the track covers, which tools to install, a 30-minute daily routine, and the rules that keep you safe from day one.
1. How this track teaches
- Twenty-one modules in six levels, from how a model writes (Level 1) to building and securing (Levels 4–5) and leading your team's adoption (Level 6), plus a case study. The module map shows them all.
- The same ten sections in every module: the idea, how it works, try it, why it matters, the five tools, a memory hook, mistakes, a lab, a quiz, and further reading.
- Press Play on every visualizer. Watching context fill up or an agent loop run is faster than reading about it. Inside a player, ← → step and space plays.
- Do every lab in two tools. Noticing what differs between tools is how you learn the concept rather than one product's buttons.
- Hover or tap a dotted word for a one-line definition; they're all in the glossary below.
2. What you need
- Some programming experience helps from Level 2 onwards; Level 1 needs none.
- A real project to practise on. Your own repository is best: every lab then produces something you can reuse.
- Accounts for at least two of the five tools. Each needs a sign-in; most offer a free tier, a trial, or inclusion in a paid plan you may already have through work. Check each vendor's current pricing before you commit.
- 30 minutes a day, five days a week, for about six months (the 26-week roadmap).
3. The five tools
All five are coding agents: an AI model in a loop that can read your code, run commands and make edits, with your approval. They differ in where they run and in their vendor, not in the core ideas. Checked October 2026.
| Tool | Made by | Where it runs | Reads project instructions from | Install |
|---|---|---|---|---|
| Claude Code | Anthropic | Terminal, plus IDE extensions, desktop and web | CLAUDE.md, or AGENTS.md when there's no CLAUDE.md | Quickstart |
| Antigravity CLI | Terminal (agy), alongside the Antigravity app and IDE | AGENTS.md and GEMINI.md | Install | |
| GitHub Copilot | GitHub (Microsoft) | VS Code and other IDEs, plus a cloud agent that works on GitHub issues | .github/copilot-instructions.md, AGENTS.md | Set up in VS Code |
| Cursor | Anysphere | Its own AI code editor (based on VS Code) | .cursor/rules/, AGENTS.md | Quickstart |
| Codex CLI | OpenAI (open source) | Terminal, plus IDE extension and cloud | AGENTS.md | Codex CLI |
Antigravity CLI replaced Gemini CLI for individual users on 18 June 2026. Gemini CLI still works with Gemini Code Assist Standard or Enterprise licences and paid API keys, and keeps the same ideas: GEMINI.md, MCP, skills and hooks.
4. Three open standards that make it portable
Learn these once and your setup moves between tools in a day instead of a quarter:
AGENTS.md
One Markdown file in your repository with build and test commands, layout and rules for coding agents. Read by all five tools (Module 09).
MCP
The Model Context Protocol: one standard way to connect tools and data (GitHub, Jira, databases, your APIs) to any AI app (Module 10).
Agent Skills
Task know-how packaged as a folder with a SKILL.md, loaded only when a task needs it (Module 11).
5. The daily routine (30 minutes)
A fixed theme per day means you never waste time deciding what to learn.
| Day | Theme | Output |
|---|---|---|
| Monday | Learn: read this week's module and run its visualizers | 3 bullet notes in your learning log |
| Tuesday | Build: do the module's lab in tool A | A commit, a config file or a saved prompt |
| Wednesday | Cross-check: the same lab in tool B | One line: what differed, and which was better |
| Thursday | Read: one release note or article about what changed this week | One thing that changed |
| Friday | Explain: write the week's idea in your own words, as if teaching someone | A short note you'd be happy to reread in a year |
Optional weekend: 1–2 hours on a longer course from a module's "Go deeper" list, or a mini-project.
6. Your learning log
Keep one file, for example LEARNING.md in your notebook repository, with one line per session. After six months it becomes your own reference guide.
2026-10-14 · AI 01 · Tokens: Telugu used 3.1× the tokens of English for the same sentence · ai/modules/01
2026-10-15 · AI 01 · Same prompt 3× in Claude Code and Codex: names varied every run; sampling · ai/modules/01
2026-10-16 · AI 02 · /context: 31K used before my first message (3 MCP servers = 14K) · ai/modules/027. Five rules from day one
8. Self-check quiz
Which open standard puts build commands and project rules in one file that all five tools read?
What does MCP standardise?
An AI tool wrote a change and the tests pass. Who is responsible for merging it?
Where are models most often confidently wrong?
You've corrected the agent on the same thing twice. What now?
9. Glossary
Every dotted word on the AI pages, in one place.