I’m enrolled at an online university on a plan called all inclusive: the fee covers the right to sit five exams per session. Sessions run four times an academic year, so twenty exam attempts in twelve months. Attempts don’t roll over: the ones you don’t use within the session are lost. It comes naturally, then, to call an exam you didn’t sit “wasted”… and that alone should tell you what kind of product this is.
I had already closed my first-year exams in April. The new academic year would have started in August, but in mid-June I was able to request an early start on the second year, and found myself with an open session containing attempts that were going to expire. The idea was to try a few and see how they were built, a reconnaissance run with no expectations.
I passed them all first time, near enough with top marks… and I’m certainly not proud of it. Here’s why.
On 23 June, Computer Networks and Cybersecurity, 26 out of 30: I turn it down. On 24 June, Software Engineering, 30. On 8 July, Machine and Deep Learning, another 30, and on the same day I try Networks again, 28. On 9 July, Information Security, 29. Four exams in seventeen days, forty-five ECTS credits.
I didn’t study. I’m writing that without softening it because it’s the fact that matters: I watched the bare minimum of the recorded lectures needed to qualify for the exam, sat in on a few live teaching sessions, took the end-of-lecture quizzes, and that was it. No book opened, no exercises redone, no evening spent over my notes.
The papers are multiple-choice quizzes of an embarrassing easiness, and here I’d rather be precise than indignant, so here’s a number. In one of the ones I closed with full marks, out of thirty questions the correct answer was the longest of the four options fully twenty-seven times. I counted them twice before submitting the paper. That isn’t statistical bad luck, it’s a well-known defect in how items get written: whoever writes them states the right option in full, with all its conditions in place, while the three distractors get half a line each. The effect is that you can answer correctly without knowing the subject, by reading the shape of the option instead of its content.
And that still isn’t all: having sat in on a few live teaching sessions, I accumulated bonus points, so the 30 in Machine and Deep Learning gains the cum laude and the 29 in Information Security climbs to 30. And the bonuses don’t stop there…
Formally the credits only land on my transcript after an on-site final sitting, which for this session takes place in September. That final consists of two multiple-choice questions per exam, worth up to two further bonus points (don’t worry, there is no climbing past 30 cum laude). They are called esoneri, exemption tests, yet they cover the entire syllabus. Why they carry that name, and what happened to online exams when the ministry required exams to move back on site, is something I want to write about separately, because it deserves far more than a parenthesis.
The mark that bothers me most is the best one: Machine and Deep Learning is the subject I’d most like to learn properly, the one where the theory would change how I work the next day, and I scored top marks without being asked anything that resembled understanding it. A 30 cum laude that measures no competence is worse than a low mark: it’s false information written on an official document.
For that exam I had bought Russell and Norvig’s Artificial Intelligence: A Modern Approach, the two volumes of the Italian edition, meaning to work through them over the summer. I’ll read them anyway. Not for the exam, which came and went without asking for them, but because the subject genuinely interests me and because I still have Foundations of Artificial Intelligence ahead of me. It’s the only way I know to make that 30 cum laude correspond to something.
At the other end of the series sits the 26 of 23 June, the lowest mark, and I need to explain where it comes from. I have worked with networks for twenty-six years. On my Billy shelf sit the three volumes of TCP/IP Illustrated, read at the turn of the millennium, when that was how you learned this material because there was no other way. Cybersecurity has been my day job for five years. Software engineering and UML have paid my bills for fifteen, to the point that three weeks ago I wrote a whole post here about how little formal modelling convinces me.
The lowest mark of the session came in the subject I know best of all. If the scores measured competence, that 26 should have been the highest of the series, not the lowest. Set beside the 30 cum laude earned in a subject I don’t know, it says the sharpest thing in this whole account: the marks don’t only err upwards. They err in every direction, which is another way of saying they are measuring nothing.
What remains is that I collected four very high marks in seventeen days, and in seventeen days I learned nothing new. The two aren’t in contradiction, and that is precisely the problem.
Last modified on 2026-07-09