What a 3–5 day finish actually costs, line by line

Every quote I send has the same four lines, and the number is always the same shape. Not because the work is identical, but because the decisions behind those lines are the same every time, and seeing them is what tells you whether the number is real.

I finish AI-built apps for a living from Seoul — one person, a Korean corporation, no agency, PM or sales layer between me and the client. Below is what the money actually buys, what makes the number move, and what I refuse.

The four lines

1. Scope. One paragraph of "done": the single job the app must do for its first paying user. One list of "not now". This is the line that decides everything else, and it is free — a scope call costs you nothing and usually removes more work than it adds.

2. The finish itself. Read the existing code as it is, verify the seams in the environment it will actually run in, fix what fails, deploy to your domain, hand over the document. For an app whose integrations already exist, this is 3–5 days of work. For a Korean client that starts at roughly ₩1,500,000 (about $1,000) for an app and ₩500,000 for a landing page — a fraction of an agency quote for the same scope, because there is no layer to pay for.

3. Revisions. Included until you are satisfied, and only sustainable because the scope is one paragraph. If revisions were unlimited against an unlimited scope, "free revisions" would be a lie in the contract, and every studio knows it.

4. Risk, priced honestly. This is the line that most quotes hide. A bank or payment-provider onboarding, a telecom integration, a hardware driver, an app-store review with a rejected build: those are weeks, not days, and they are named before the quote rather than discovered in week three. When the risk is large, the honest quote is larger and later. That is not a worse deal, it is the only one that survives contact with the platform.

Payment follows the same shape: nothing before the scope is written, a deposit when we start, and the balance after you have clicked the working build yourself, on your own domain, with one real card payment.

What makes the number go up

Not features. Features are the cheap part. The number moves on:

• Anything that needs someone else's approval. A company verification at a payment provider, a business account at a store, a domain mail record with a provider that has to be reached by ticket.
• Anything where being wrong costs money. A refund flow, a double-charge, a subscription that must fail closed. That work is tested twice and takes twice as long.
• Anything that does not exist yet. Content, art, level design, a pricing model. I can build a game's systems in days; I cannot invent a game's content in days.

Everything else — the screens, the CRUD, the third API — is the part an AI tool already gave you. You are not paying for that again.

What I refuse, and why it helps you

I turn down work when a client wants a store launch in five days on a brand-new corporation with no verified business profile, or a payment integration on a personal account that the provider has already flagged, or "no need for a contract, we trust each other". Each of those ends the same way: a build that is finished and cannot ship, and a client who concludes the developer was slow. Refusing at the quote stage is cheaper for both of us than discovering it at the store review.

I also refuse to bill for a diagnostic. If you built something with Cursor, Lovable, Bolt, v0, Replit or Claude and it has been "almost ready" for a month, reading the code and telling you what is actually blocking launch is free, in writing, whether or not you hire me. Half of those reports end with "this is a two-day fix, do it yourself" — and that is a fine outcome.

The questions that separate a real quote from a wish

If you are comparing studios, the four lines above tell you which ones are guessing:

1. What exactly is finished at the end? A paragraph, not a deck.
2. What is explicitly not in this price? If the answer is "nothing", the quote is a wish.
3. Who deploys, and on whose account? If the deploy lands on the studio's account, you are renting your own product.
4. What happens if the store rejects the build? The honest answer names the fee or the week.
5. When is the balance due? "After you see it working" is a sentence that costs the studio something to write.

None of this is exotic. It is just usually missing, because a quote written to win the job and a quote written to be executed are different documents.

If you want the free version of that report on your own repo: [email protected] — no form, no call required to get the answer. The studio's English page is at https://cpcai-studio.netlify.app/en/, and the same site in Korean and Japanese.