Clone project
A curriculum whose unit of instruction is one application, reproduced end to end — usually a recognisable commercial product, rebuilt feature by feature until the learner has something that behaves like the original. The Uber clone, the Airbnb clone, the Twitter clone. The teaching is normally a long video; the repository is the answer key.
Why it is its own form
The spoke’s other forms are distinguished by what a lesson is: a pointer (meta-curriculum) or a runnable file (executable-curriculum). A clone project has no lessons in that sense. It has a build order. Progress is measured by how much of the app works, and the curriculum ends when the app does.
Two consequences follow, and both separate it from everything else here.
The scope comes from the product, not from a discipline. A CS study plan chooses topics; a clone inherits them. Whatever the original does — auth, maps, payments, history — is what gets taught, in whatever proportion the original happens to need. Nobody decided that a beginner should meet payment integration before data modelling; Uber has payments, so the tutorial does.
The finish line is an artifact, not a readiness. interview-preparation material aims at being able to answer. A clone aims at having built. Where no assessment regime exists to shape a curriculum (synthesis), a portfolio object is the thing on offer instead.
The scratch-versus-services split
Build-it curricula divide on a line the “build it” label hides, and the corpus now has one of each:
- From scratch — ai-engineering-from-scratch: implement it by hand, then use the framework. The learner ends up understanding a mechanism.
- From services — adrianhajdin-uber-clone: implement nothing, integrate five vendor SDKs. The learner ends up having shipped an application.
Neither is a lesser version of the other, and the difference is not difficulty. They answer different questions — how does this work against how do I assemble a working product — and a curriculum that does one is poor preparation for being asked the other.
Decay
The fastest-rotting form in the spoke. A clone’s dependencies are commercial services with their own release cycles: free tiers change, SDK majors land, consoles get redesigned, and any of these can break a step in a video that cannot be edited. Link rot takes years; a changed onboarding flow at one of five vendors takes a weekend.
Evidence here is a single source. Whether clones typically get maintained after the video ships is unmeasured, and so is completion — the standing hole for every curriculum in this spoke.