Your First Real Project: You're Ready to Start

Here's a quiet little secret about everyone who ever got good at this: at some point they stopped doing the exercises and started a thing that was theirs. A messy, half-finished, nobody-asked-for-it thing they wanted to exist. That's the moment it stops being a course and starts being a skill.

Today is that day. This lesson is short on purpose — because the real work isn't more reading, it's opening a folder. You already know everything you need. What's left is the last mile: turning I did the exercises into I started something. Let's cross it.

Picking your first real project

The single biggest mistake at this stage isn't technical — it's picking a project so big it never gets past day one. So before you build anything, choose well. A good first project has four marks:

• Small. Finishable across a weekend of short sessions, not a saga you're still untangling in a month. You want the win.
• Yours. It scratches a real, personal itch — because motivation is what survives the first bit of friction, and your own annoyance is the most reliable fuel there is.
• File-shaped. It lives in a folder and is mostly made of documents, pages, or data — the exact home turf this whole course gave you.
• Forgiving. Nothing breaks if the first version is rough. No live business data, no thing you'll be embarrassed by if it's imperfect.

Stuck for an idea? Here's a starter menu — one line each, pick the one that makes you lean in:

• A little website for a hobby — the Recipe Box pattern, re-skinned for your bikes, your books, your band.
• A family archive — years of loose photos and documents, sorted into something you could actually hand to a relative (the lesson-nine move).
• A monthly report routine built from your own spreadsheets, so next month is one sentence instead of an afternoon.
• A reading or watch tracker with a tidy summary page.
• A small info page for a club or team you're part of.
• A "tidy my digital life" project — one messy folder, made calm.

And an honest anti-list, because knowing what not to pick saves week one: not your company's real, live data, not "an app with logins" (not yet), and not something you actually need finished by Friday. First projects are for learning the rhythm, not hitting a deadline.

The three habits that keep it safe

Everything mechanical you'll do, you already learned. What you haven't been handed yet is how the people who do this well actually work day-to-day. It comes down to three small habits — and this is the one new thing this lesson teaches.

Habit one — start in plan mode. On anything real, your first move is to turn the dial (lesson four) to Plan: Claude reads, thinks, and writes you a plan — but changes nothing until you say go. Read the plan. Then let it work. That's not the timid way to start; it's how professionals start too. Seeing the plan first is how you stay the driver.

Habit two — the build loop. Here's the reassuring part nobody tells beginners: Claude Code keeps a safety net for you automatically. Before it edits a file, it quietly snapshots what was there. So the loop is simple — let it build, look at what changed, then keep it or undo it. If a change isn't what you wanted, press Esc twice and the files snap back to how they were. You don't set anything up first; the net is always there. (One honest edge: this only rewinds files on your computer — it can't un-send an email or un-change a live system, which is exactly why Claude still asks before doing things like that.)

Habit three — short sessions, resumed. Remember the limited desk from lesson seven: everything in a session takes up room on it, and the longer a single session runs, the more the desk fills up and the more the conversation drifts. (If you did our AI Basics course, you've met this desk before.) So don't marathon. Work in focused sittings, stop while it's still going well, and come back tomorrow — claude -c in the terminal, or resume in the app (lesson twelve). Your CLAUDE.md and your files are what carry across, not the length of the chat (lesson five). Short and resumed beats long and foggy, every time.

When you get stuck (the 60-second triage)

You will hit friction in week one. Everyone does. Here's the whole map for getting unstuck — every line is something you already know:

• Confused by what Claude did? Ask it to explain, in plain words, what it just changed. It will.
• Wrong direction entirely? Press Esc twice, back to before, and steer again.
• Session feels foggy? End it and resume fresh — the desk was full (habit three).
• Explaining the same thing every time? That's a CLAUDE.md line (lesson five), or a skill (lesson six).
• One big, messy sub-task swallowing the session? Send a helper to do it (lesson seven).

That's it. Five moves cover almost every wall you'll meet — and not one of them is new.

Want to put it online?

One small door, since people always ask: if you build a website — the Recipe Box, or your new project — that kind of static site is about the easiest thing there is to put on the internet. You don't need a lesson for it. When you're ready, just ask Claude: "help me put this site online — walk me through the choices." It'll be honest that this means using an outside service and making a couple of small decisions, and it'll guide you through each one. A door, open whenever you want it.

Where you go from here

You've reached the edge of the beginner's map — so let's name the country beyond it honestly, not as a syllabus but as territory you're now equipped to explore. Skills can carry real scripts and templates, not just instructions. You can wire Claude to systems no directory has a ready-made connection for. You can even package what you build into plugins and share them. That's mechanic territory — building the tools, not just driving them — and the foundation you'd stand on to learn it is the one you just spent twelve lessons pouring.

A deeper course on exactly this territory is coming soon. When it lands, you'll be more than ready for it. In the meantime, Anthropic's own documentation is a fine map for the curious — wander it whenever a "could Claude also…?" thought shows up.

You're ready to start

Cast your mind back to lesson one. You arrived wondering whether you were even technical enough to be in the room — staring at a black screen, half-expecting to be told you'd walked in the wrong door.

Look at you now. You can brief a project before it starts, teach Claude a skill, hand a big job to a helper, plug it into live tools, and sit down at any cockpit — app, terminal, editor — and feel at home. That person from lesson one didn't get replaced. They got equipped.

So there's nothing left to study. The only thing between you and being someone who builds with Claude Code is a folder that doesn't exist yet.

Go make the folder.