The Slice That Got Cheap

AI collapsed the cost of the demo slice — the part that was always cheapest to show — while everything the screens rely on stayed the same size.

The Most Visible, Least Representative Part

If you came here from Vibecoding, you watched something remarkable: a working prototype appeared in a weekend, most of it written while you made tea. That experience teaches a lesson, and the lesson is subtly wrong. The demo — the screen, the button, the confirmation — is the most visible part of a product, and also the least representative. It is the part every product shows off, and the part that says least about how much work remains. AI collapsed the cost of building it, and it happens that this was exactly the slice that was always cheapest to demonstrate, even before AI arrived. The expensive parts were never the screens. They were everything the screens quietly rely on, and nothing about a fast weekend tells you how large that everything is. This lesson is about keeping the speed while correcting the conclusion.

The demo is the most visible part of a product and the least representative
AI collapsed the cost of exactly the slice that was always cheapest to show
The screens were never the expensive part, even before AI
A fast weekend says nothing about the size of what the screens rely on

What Stayed Exactly As Expensive

Here is what the fast weekend did not touch. Where the bookings actually live, and what happens when one of them is wrong. Who can sign in, and who is allowed to see what. How a change reaches customers without breaking the thing they are in the middle of using. How you find out it broke. Who answers the email when a customer is locked out on a Saturday. None of these got cheaper, because none of them were in the demo. So the sentence "it took a weekend" is true and misleading at the same time: true about the slice, false about the product. This matters because the collapsed slice changes how the whole job feels. The gap between prototype and product looks smaller than it used to, while remaining exactly as wide. Your judgement recalibrated on the wrong evidence — which is nobody's fault, and still your problem.

Data, accounts, shipping, watching and support were never part of the demo
"It took a weekend" is true about the slice and false about the product
The gap feels smaller than it used to while staying exactly as wide
Judgement recalibrated on the demo is judgement trained on the wrong evidence

The Speed Is Real — Keep Using It

None of this means the speed was an illusion. Building the rest of the screens for the groomer is genuinely fast with AI, and you should let it be. The admin view, where she sees the whole week at a glance instead of scrolling through a customer page that was never meant for her. The cancellation flow, so a customer can free a slot without ringing mid-groom. Each of these arrives in hours, not days, the same way the first screen did. Use that. The mistake was never believing AI is fast — it is — but concluding that the product is therefore mostly done. The honest position holds both ideas at once: the visible work has become cheap, and the visible work was always the small share. This module builds the rest of the visible part, then spends its remaining lessons on what "finished" actually has to mean.

The remaining screens — the admin view, the cancellation flow — still arrive fast
The speed is real; the conclusion drawn from it is what goes wrong
Visible work became cheap, and visible work was always the small share
Hold both ideas at once, and let AI do what it is genuinely good at

Prefer slides, quizzes, and saved progress? Read this lesson in the library — free, no sign-up.