Building

Getting what you actually wanted out of the conversation.

The part you will spend the most time in.

Describing what you need is the one skill worth learning; everything else here is mechanics. Watching a build explains what the steps mean and why you are not shown a log. Changing your mind covers interrupting, correcting and undoing. When a build fails is the page to open when one does. Showing it a file is often faster than describing something, and Choosing a design system decides what the result actually looks like.

Starting from what you have is worth reading before you create anything: a spreadsheet you already keep is the fastest route to a first version that fits.

When it asks you is the other direction: some decisions are genuinely yours, and those arrive as a screen to fill in rather than a message to answer.

Asking for settings is the one people are most surprised by: the conversation reaches the whole platform, not only the code, so if you do not know where a setting is you can simply ask for the outcome.