Exams are the mark of a lazy instructor who doesn’t want to design appropriately challenging assignments, as the author says: “I hadn’t fully prepped the class”.
As soon as those questions leave the classroom, they will be passed around in group chats. That’s why exams remain relevant. If a student averages 95% on their assignments, but shows up to the exam and can’t solve ∫sin x dx like they had supposedly done a hundred times on their own, then they’re probably cheating.
The math exam which I wrote was also very demanding and had a very tight time limit. There wasn’t any time to compare results with a group chat or use tools like wolfram alpha. Either you knew how to solve the exercises or you didn’t and failed.
You repeat this process every year for 10 years and now you have 200 questions. The next professor comes in and does the same but now also has 200 questions + answer guides.
With some basic programming you could make it automatically change the numbers (to "friendly" numbers that break up into round numbers) and have an infinite pool of challenging quiz, midterm, and final questions. If everyone gets different questions they cannot cheat by sharing answers. Only by sharing approach.
Apart from some high-tech cheating, I'd say it was a reasonably cheat-proof class. I don't know how applicable that model is to calculus though.
What you are describing in an incredibly labor-intensive process. If you want to restructure education so no teacher ever has more than 8 students and 2 classes a semester (and tuition is $100k/yr)... maybe you can make this work.
In the real world, where a teacher has 3 sections of 30 students, "just make exciting and engaging material!" is honestly an asinine statement. You can't spend 2 hours evaluating each student's homework assignment.
Putting a multiple choice test online and expecting students to not take advantage of this examination method is very naive.