What You Built and What Comes Next

The Full Journey

Take a moment to look at where you started and where you are now.

Lift 1 — Build Something Real with AI. You sat down at a chat tool, not sure if you could build software. You learned that prompting is a skill — Scope, Intent, Structure. You wrote your first user stories. You built an Avalanche Field Guide in a chat window with made-up data, and it actually worked. That was the moment: I can do this.

Lift 2 — From Chat to Code. You moved into a real development environment. You learned that a codebase is just files in folders, and that version history means you can always undo any change. You migrated your chat prototype into a real project and connected it to live data — real danger ratings, real weather, real snowpack conditions. Your capable colleague moved from a chat window into your project, reading and writing real files on your behalf.

Lift 3 — Giving AI Context. You solved the biggest frustration — AI forgetting your project every conversation. You created a project context file, the house-sitter note that your AI coding assistant reads automatically. Every new conversation started with your assistant already knowing the project. Iteration got faster. You got more ambitious.

Lift 4 — From Your Workspace to the World. And now, you're about to deploy. Your Field Guide is going live — a real URL, accessible to anyone, anywhere.

That's the journey: chat → codebase → context → deployed application. You built real software — not a toy, not a demo, not a homework assignment. A real, data-driven application that someone heading into the backcountry could actually use.

Know the Risks

You built something real. Now here's the honest part.

AI-generated code — the code your AI coding assistant has been writing on your behalf — is powerful, but it has some limitations. Being aware of these doesn't diminish what you've built. It makes you a more responsible builder.

The Two-Week Cliff

Here's a pattern that happens with AI-built projects: things go incredibly well for the first week or two. You're adding features fast, everything works, you're on a roll. Then you make one change, and something that was working before suddenly breaks.

This happens because AI optimizes for "working right now," not "easy to change later." It builds fast by connecting things directly — like wiring every light in the house to one switch. Everything works great at first. But when you ask AI to change one light, it pulls on wiring that's connected to everything else. The kitchen renovation breaks the living room lights because they were tangled together from the start.

This isn't a flaw in AI — it's what happens when you build fast without structure. The faster you build, the more important it is to pause and verify that what was already working still works.

The solution? Testing. Other tracks at this conference are learning how to have AI write automated tests — checks that run every time you make a change and catch breaks before you do. That's a skill you can pick up next.

The Validation Gap

AI generates code faster than anyone can manually check it. This is wonderful for speed, but it means there's a gap between "AI says it's done" and "it actually meets requirements." That gap gets wider when your specs are vague — AI fills in the blanks with its best guess, and its guess might look right without being right. Every time you verified your acceptance criteria — "does this actually show today's danger rating?" — you were closing that gap manually. As projects get bigger, you'll want automated ways to do that.

Security and Accuracy

Like any code — whether written by a human or by AI — security matters. AI-generated code can introduce issues like not properly protecting user data or missing input validation. The good news: your deployment pipeline includes automated security scans that check for common vulnerabilities every time you deploy. Think of it like a spell-checker for security — it won't catch everything, but it catches the obvious issues before they go live.

For anything mission-critical or handling sensitive data, you'd pair these automated scans with a review from someone who understands security. That's true of all software, not just AI-built software.

Similarly, AI can generate information that looks correct but isn't. For your Field Guide, the live data comes from real APIs, so the danger ratings are real. But if you asked AI to explain avalanche science, it might get details wrong. Always ground critical safety information in authoritative sources — like the Utah Avalanche Center — not in AI-generated text.

This Is the Beginning

None of this should make you feel less proud of what you built. Every piece of software in the world has limitations and risks. What matters is knowing they exist and getting better at managing them over time.

You've taken the first — and hardest — step: you proved to yourself that you can build software. The skills you learned today — prompting with specificity, writing user stories, giving AI context, working as a team — those are real skills that transfer to whatever you build next.

Here's the choice you'll face from here: AI lets you produce things faster. The temptation is to use that speed to produce more. But today you invested that speed into understanding — learning how to prompt with specificity, how to write stories that define real outcomes, how to verify what AI builds. That's the difference between building a dependency on AI and building a skill set that makes AI more useful over time. Keep investing in understanding, and what you can build will keep growing.

Team Activity: Looking Back and Looking Forward

Format: Team Discussion Time: ~5 minutes

This is your last team discussion before the final sprint. Take a moment to reflect.

Discuss: - What surprised you most about today? What was easier or harder than you expected? - If you could go back to Run 1 with everything you know now, what would you do differently? - After today, what's one thing you might build with AI outside of this workshop? A tool for your job? Something for your family? A side project you've been thinking about?

Key Insight

You came in not knowing how to build software. You're leaving with a deployed, working, AI-built application — real data, real features, accessible to anyone with a link. That's not a small thing. Software creation is no longer gated behind specialization — and what you do with that is entirely up to you.