Lovable in reverse. Nothing gets generated: you get the app already on the store. Your sentence is matched against features we saw screen by screen, not against listing copy.
I want a budgeting app with shared expense tracking, recurring reminders, and a simple UI.
Nobody writes "shared expense tracking" in a subtitle. The feature is three taps into the app, on a screen the index never opened.
Shared expenses, a reminder that repeats, nothing bloated. Three requirements on one line.
It indexes title, subtitle and description. Words a marketing team chose, months before your question existed.
Installs, ratings, ASO spend. The result set comes back ordered by reach, and reach is not fit.
Only one of them has a group ledger. You learn which by signing up four times.
The index is a list of things apps do. The closer your sentence gets to a requirement, the sharper the result set.
Every capability in the index came from a crawl that opened the screen and read what was on it. Nothing is inferred from a category or a listing.
Write "split a bill with friends", not "finance app". Your words are matched against what screens do, so the phrasing a user would pick works best.
Must-haves outrank vague wishes. One requirement returns half the corpus and six returns nothing, so three is where the list gets useful.
Every app in the set was run on a real device and opened screen by screen. Searching it is just reading back what the run saw.
Exploration opens every screen in an app and records what is on it. That run is the only source of a capability.
A group balance plus an add-expense sheet becomes shared expense tracking. The name comes from the screen, not the store copy.
Your line is split into the same vocabulary, and each capability is marked must-have or nice to have.
Apps are ordered by how many must-haves have a screen behind them. What is missing stays on the row.
Capabilities are filed under a domain, and your sentence is mapped into the same set. That is why "split a bill" and "shared ledger" land on the same screens.
Three examples each. The list grows every time a crawl finds a screen the vocabulary has no word for yet.
A match is a claim, so every row names the screen that backs it. The row with nothing behind it says so instead of guessing.
com.Splitwise.SplitwiseMobile
The result set answers one question: which apps already ship these capabilities. It deliberately refuses to answer any others.
An honest gap is more useful than a confident guess, so the row stays empty until a screen fills it.
One sentence, and the apps that already ship it come back with the screens that prove it. No store search, no four sign-ups.
