From your first prompt to your own agent.
Learning paths in three levels, from the first usable brief to an agent in continuous operation. Every station says what you will be able to do afterwards and how long it takes. Free, no account – your progress stays in your browser and lights up the route as you go.
Starter
After this level
You work confidently with an assistant: good briefs, the right tools and a feel for when you need to check.
AI in your working day
From your first usable answer to the limit where thinking for yourself wins – seven stations, no prior knowledge.
- Prompts
- Context
- Privacy
- Model limits
Start path
Your AI toolkit
Model choice, projects, voice, files and research: the features that turn a chat window into a working tool.
- Model choice
- Projects
- Voice
- Research
- Images
Start path
Intermediate
After this level
You give the AI memory, access and a schedule – and you know where to keep it on a short leash.
Knowledge & context
A memory that belongs to you: notes in Markdown, Obsidian as a knowledge base, rules files and connectors an agent can actually read.
- Obsidian
- Markdown
- AGENTS.md
- MCP
- Memory upkeep
Start path
Agents that act
Computer use, browser agents and the always-on agents from Grok Bot to Duet: what they can do, where they run and which approvals you keep.
- Computer use
- Browser agents
- Grok Bot, dots, Duet
- Approvals
- Prompt injection
Start path
Automating with scheduled tasks
From the first weekly report that arrives by itself to the automation that reports failures instead of hiding them.
- Scheduled tasks
- Cron
- Webhooks
- Monitoring
Start path
AI in your store
How your store shows up in AI answers, what agents demand of it, and how to make both measurable.
- Agentic commerce
- Product data
- Agent-ready
- Search Console
Start path
Visible in AI answers
Getting cited rather than merely ranking: what a model lifts from a text, and how to measure whether it happens.
- GEO
- Quotable writing
- Web vitals
- Measurement
Start path
Expert
After this level
You build with coding agents, run models on your own hardware and operate agents so that a mistake surfaces before it gets expensive.
Building with coding agents
Git and GitHub as a safety net, briefs with tests as acceptance, pull requests from agents and several agents in parallel.
- Git
- GitHub
- Claude Code, Codex
- Pull requests
- Worktrees
Start path
Local AI
Models on your own machine: what they can do, which hardware is enough and how to mix local and cloud sensibly.
- Ollama
- LM Studio
- Quantisation
- Hardware
- Routing
Start path
Running agents
Evals, skills, multiple agents, cost and a runbook: what it takes for an agent to still be running after the pilot.
- Evals
- Skills
- Multi-agent
- Cost
- AI Act
Start path
How the academy works
- Three levels build on each other, but nothing is locked. Start wherever the first sentence describes you.
- Every station names one outcome, not a topic. If it does not change what you can do, it is not a station.
- Lessons open with what to take away and close with one self-test, reasoning included. Not points – understanding.
- Dated facts carry their date. Products in this field change monthly; the lessons say what was checked and when.
- Progress is optional and local. Nothing is hidden behind a login.