Don't push past confusion

The biggest mistake learners make is nodding along when something doesn't make sense, then losing the thread three steps later. Claude is right there in the conversation. Talk to it.

Phrases that work well

You don't need a special command or a precise question. Plain English is the interface. A few that consistently get useful answers:

• "break down that code" — get a line-by-line walkthrough.
• "explain that like I haven't seen it before"
• "why does this work this way?"
• "what would happen if I changed X to Y?"
• "show me what's in <file>" — pull a companion file inline.
• "walk me through renaming X to Y" — predict the result before you edit.
• "I don't understand this part — <quote it>"
• "what should I say next?" — when you're not sure how to continue.

When you hit an error

Paste the error back to Claude with one of:

• "what does this mean?"
• "how do I fix this?"
• "is this a bug in my code or in the workshop?"

Claude has the full lesson context loaded — the README, the source files, the previous output. It almost always has more to say about an error than a search engine will.

When you want to go deeper

Workshops walk you through a single happy path. If you're curious past that, ask:

• "what would the next-level version of this look like?"
• "what tradeoffs am I making with this approach?"
• "how would you do this in production?"
• "what would break if I scaled this up?"

Slow down, then resume

If a step ran and you didn't actually read the output, say so — "hold on, walk me back through what that just did" or "don't run anything yet — let me read that". The agent will pause and explain before continuing. There's no penalty for slowing down, and the workshop is faster overall when you don't have to backtrack.