Build software with AIplanned
Fast is the easy part.
A good setup makes you quick at producing things that look finished. This course is the other half. Knowing what you actually asked for, and spotting when what came back is wrong. Nobody teaches this part, and it decides whether any of the rest holds up in production.
You'll need a paid Claude plan, which is $20 a month, and a machine you can install things on. The workflow course is not required, but everything here assumes you have a workflow to apply it to.
What It Covers
Knowing when it's actually done.
Including the parts that go wrong, because those are the ones worth watching.
Defining done before you start
What the feature is worth and what success looks like, then how finished it actually has to be. Demo, production and throwaway are three different targets, and the trouble starts when a throwaway quietly becomes production because nobody said which one it was.
Tickets as acceptance criteria
Write the criteria before any code exists, then read them. That's where you find out it misunderstood the feature, at the point where it costs you nothing. Say what's out of scope too, because agents are relentlessly helpful.
Pause before you build
Talk the thing through first and make it restate the problem back to you. Nearly every bad output traces back to a misunderstanding you could have caught in the first thirty seconds if you'd asked.
Reviewing what comes back
The most important lesson here. How to read a diff properly, and how to recognize the answer that sounds confident and is wrong. There is an order to check things in, and it saves you a lot of time.
The pull request is the boundary
Branches, PRs, and having Claude review the change before you do. Then wiring that review into GitHub Actions so it runs on every push whether you remember or not.
Your docs repo
A written source of truth for the architecture and the infrastructure, so nobody is guessing at what the system is supposed to be. Keeping it current as part of the definition of done, and keeping it from contradicting your CLAUDE.md, which is the failure that quietly makes both of them useless.
When it goes wrong
Spotting the loop where it circles the same wrong answer, and knowing when to stop steering and start over. Git is the safety net, and everything in this course assumes you can undo things.
Know when it opens.
Leave your email and we'll tell you when the first lessons are up. You'll hear from us once, at the point where there's actually something to watch.
Would rather we just built it for you? See what that costs.