Build with AI, properly
Four hours in five blocks, for capable people who already use Claude but lack the layer one level below the tool. Read it in order, or jump to the block you need. Each lesson has a concrete example and a thing to try.
Three threads run through all of it
My own way of working, the official docs repackaged, and the programming fundamentals nobody explains to non-developers.
The course is being written. This is the outline, and lessons appear here as they are finished.
The map, and how to talk to Claude
1.1Why this course exists
1.2The map
1.3There is no art to prompting
1.4Things already switched on that you did not know about
1.5Which model, how much thinking
Fundamentals that stop your thing from breaking
2.1Git without fear
2.2Where your code runs: local environment, packages, servers
2.3Secrets and environments
2.4Will it crash when real users arrive?
Connecting things: APIs, MCP, integrations
3.1What an API is
3.2Connect a tool in one line
3.3Where the connection actually lives
3.4What it costs and what can go wrong
3.5The 80% report, end to end
The official Anthropic toolbox
4.1The shipping box model
4.2CLAUDE.md: your style, and your project
4.3Your first skill
4.4Request vs enforcement: your first hook
4.5Plugins and subagents in five minutes
4.6Giving Claude your knowledge: docs, frameworks, your own wiki
4.7Claude outside the terminal
Money, models and the wider world
5.1Token literacy
5.2Two ways to pay, and the cheaper model myth
5.3The wider world: models, routers, harnesses
5.4VPS vs your own computer
5.5Decision sheet