2 Chat: Paste code, get answers
This chapter covers
- Follow-ups over "explain this code"
- Serializing a repo into one pasteable file
- Interrogating LLMs with constraints
- The forgetting curve and learning notes
- Chat, workflow, and agent: when to use which
Most developers' first AI interaction with code looks the same: paste a file, type "what does this do?", get a wall of generic text that could describe any codebase on earth. They try again. Same wall. They conclude the tool is overhyped and go back to reading code manually. The problem was never the model. It was the question.
One good follow-up question will teach you more than ten "explain this code" prompts. This chapter is your first real rep: you'll paste a file into a chat window, ask a bad question, watch it fail, then ask a good one and get a genuine design insight the code alone couldn't tell you. By the end, you'll have a one-page map of how a real codebase works. No installation required. Chat is free, works on any model, and it's the best tool for exploring a few files at a time. It breaks down past a handful of files, and it depends on you knowing what to ask. When you don't know what to ask, or the codebase is too big to paste, that's where workflow and agent take over (Chapters 3 and 4).
Figure 2.1 Paste code and get a generic wall of text, follow up on the one line that surprises you, constrain the output with numbers instead of adjectives, then write a learning note that fights forgetting.