This course teaches through one finished machine. Before you copy any part of it, spend a session deciding how you would solve the problem — because the reference build is only one answer among many good ones.
Here is the entire brief, before a single part gets chosen: sort a mixed bag of Skittles by colour, automatically, with no hand-sorting once it starts. Everything else in this course — a rotating wheel, a spectral sensor, a Raspberry Pi, a laser-cut frame — is a choice that was made to solve that brief, not a requirement baked into the problem itself. A good engineer separates the two: the problem is fixed, the solution is not. Every workable solution, however different it looks, ends up doing the same five jobs — it just does each one a different way:
Any sorting machine, however it's built, has to solve five separate jobs. Write these five headings on a page, with space under each:
Don't judge ideas yet, just list them. To get started, here are three real options: a gravity hopper with an agitator dropping Skittles into an indexed wheel (this course's approach); a conveyor belt carrying them past a sensor one at a time; a vibratory bowl feeder, like an industrial parts feeder, lining them up single-file. Add a fourth of your own — even “a person places them by hand” is a legitimate answer for a first prototype, it just removes automation from that one job.
This course uses a dedicated spectral sensor (the AS7341) with its own onboard illumination LED. Other real options: a camera with image-processing colour detection (more flexible, more compute needed); three simple photodiodes behind red/green/blue filters, compared directly (cheaper, cruder); or, at the extreme, a person calling out colours by eye. List at least one you hadn't considered before reading this.
This course uses a rotating slide that tips each Skittle into one of several fixed positions as it turns. Other real options: moving cups or bins on a carousel that come to the Skittle instead of the other way round; a row of pusher solenoids or servos, one per colour, each nudging a match sideways off a shared belt; or a small robotic arm that picks up and places each Skittle individually — slower, but trivial to reprogram for a different number of colours. Sketch one of these as a rough diagram, not just a sentence.
The specific combination this course builds — wheel feeder, spectral sensor, Raspberry Pi Zero, rotating slide — wasn't picked because it's the fastest, cheapest, or most elegant solution to the brief. It was picked because, in this order, it touches the widest range of transferable skills: parametric CAD, 3D printing and laser cutting, stepper motors and homing, I2C sensing, basic classification, mechanical actuation, and system integration. A conveyor-and-camera version might be a better product; it would teach a narrower, more software-heavy set of skills. Keep this distinction in mind for the rest of the course: you're being shown a good teaching path, not being told the only right answer.
Decide, in writing, one of two things: (a) you'll follow this course's mechanical design broadly but you're free to swap in a different approach to one or two of the five jobs as you go, or (b) you're deliberately building something structurally different from the start. Either is a completely valid answer — write one paragraph justifying whichever you pick. There is no wrong choice here, only an unexamined one.