From Chat to Code¶
You Built Something Real¶
In Run 1, you and your team built an Avalanche Field Guide — a real, interactive backcountry safety resource — by talking to AI in a chat tool. That's not nothing. You went from a blank screen to a working field guide in under an hour.
But you probably bumped into some walls:
- The data is made up. Your danger ratings are hardcoded text, not today's actual conditions. The weather is whatever AI invented. There's no way to connect to live data sources from inside a chat.
- Every change is a re-roll. Want to update one thing? You re-prompt, and AI regenerates the whole output — but remember, AI is probabilistic. The new version might change things you didn't ask it to change. There's no way to edit just one piece.
- It's one big output. Want to add a second page? A weather section? Navigation between pages? Chat gives you one artifact at a time. A real application is many files working together.
- It's hard to undo mistakes. If AI regenerates something and breaks what was already working, there's no easy way to go back to the version that was right.
- Your team can't build together. One conversation, one person prompting. There's no way for two teammates to work on different parts of the field guide at the same time.
None of this is a problem with your prompting skills — those are solid. The limitation is the tool. Chat tools are designed for conversations, not for building things that persist.
Think of it this way: working in chat is like a meeting where nobody took notes. You had a great conversation, maybe even built something on a whiteboard — but when you come back tomorrow, all you have is a fuzzy memory. Want to change one thing? You have to reconstruct the whole conversation. Working with files is like walking out of that meeting with a shared document. You can point to a specific page, change one line, and hand it to a teammate — because the work lives in the artifact, not in anyone's memory of the conversation.
What Changes Now¶
During Setup, you got access to something new: a development environment — a place where your work lives as files on disk, not text in a chat window.
Your AI coding assistant works the same way the chat tool did — you talk to it in plain English, you use the same Three Pillars and user stories, you follow the same Explore → Plan → Implement → Verify workflow. The difference is what it can do:
| Chat Tool | AI Coding Assistant |
|---|---|
| Data is hardcoded — whatever AI makes up | Connects to real, live data sources |
| Every change re-generates the whole thing | Edits the specific file that needs to change |
| One output at a time | Creates and manages multiple files together |
| Hard to undo specific changes | Every change is tracked and reversible |
| One person, one conversation | Your whole team can work on the project at once |
Same skills you already have. New capabilities that remove the walls you just hit.
Key Insight¶
The transition from chat to code isn't about learning something completely new — it's about unlocking what your existing skills can do when AI has access to real files instead of just a conversation window. Everything you learned in Lift 1 still applies. You're just giving AI more to work with.