Plenty of people reach this page with something already half-built: an app they vibe-coded over a weekend, a template they downloaded, a project someone else abandoned. Everything so far has assumed an empty folder. It does not have to. This chapter aims the same path at code that already exists.
1.4.1Skip this if your folder is empty
If you have nothing yet, you are on the normal path. Go to the next chapter.
If you do have something, nothing about this book changes except where you start. The stages run in the same order, and a few of them are simply already finished. Your job in this chapter is to find out which ones.
1.4.2Survey it before you touch it
The instinct is to hand the whole thing to your agent and say "clean this up". Do not. You would be trading the only working version you have for a version nobody has run.
Read first. Have the agent go through what is there and write down what it finds: what the project is, how to start it, what it is built with, and what the data looks like. Nothing else. No fixes, no tidying, no improvements.
That written survey is worth more than it sounds. It is the front-door document your project never had, and it is the first honest description of what you actually own.
1.4.3Your gaps are your reading list
Now line up what you have against the fifteen stages. A typical weekend project scores something like this:
| Stage | Usually true of an existing project |
|---|---|
| Set Up | Done. It runs, and something is installed |
| Plan | Missing. Nobody ever wrote down what it should do |
| Architect | Partly. There is a layout, but nobody chose it |
| Build | Partly. Features exist, in whatever shape they landed |
| Test | Missing, almost always |
| Ship, Operate | Missing. It runs on your laptop and nowhere else |
The empty rows are your reading list, and they are usually the same rows for everyone. That is the actual reason a weekend app dies in a month: the parts that were skipped are the parts that keep software alive.
1.4.4Keep, redo, or delete
You will be tempted to throw it all away and start clean. That is almost always the expensive choice. Sort it into three piles instead:
- Keep anything that runs and that you can describe. Working code you do not understand is still working code.
- Redo only what you cannot explain and that touches money, passwords, or personal data. Those two conditions together, not either one alone.
- Delete whatever nothing calls: the abandoned file, the half-finished page, the feature you talked yourself out of.
Rule of thumb: rewriting feels like progress and usually is not. Every hour spent replacing code that already worked is an hour not spent on the empty rows above.
1.4.5Have your agent survey what you own
Do this now: paste the prompt and read the survey before changing one line. Then carry on through the book, moving faster through the stages it says you already finished.
