Rethinking the Quiz: A Tool for Self-Directed Learning

What should you learn next? That’s a deceptively hard question to answer when you’re teaching yourself a technical topic. I’ve spent months experimenting with a tool to help people figure that out, and the latest iteration is a set of "questions" — not flashcards, not quizzes — available at questions.wizardzines.com. Right now it has two sets: questions about UDP and questions about sockets.

Two Kinds of Assessment

Traditional schooling relies on two types of evaluation. Summative assessment determines grades — pointless when you’re self-teaching. But formative assessment, evaluations meant to adjust learning activities and improve outcomes, is exactly what a self-directed learner needs. The goal wasn’t to test people on material I was teaching; it was to help them evaluate their own understanding and decide what to learn next on their own terms.

A Quick Reality Check

A Twitter poll asking whether people found diagnostic quizzes at the start of a class helpful drew about 90 responses. The themes were revealing. Quizzes trigger bad memories of school for many. People who liked them mostly appreciated being able to skip sections they already knew. But the most instructive takeaway: being told you don’t know something is an unpleasant experience for many people, even when it’s constructive.

First Try: Flashcards You Learn From

That insight led to a reframe: instead of a test that exposes gaps, build something that helps people learn. The first prototype was a set of flashcard decks, starting with one on containers. Each has roughly 14 questions; you click a card to reveal the answer and then tag it as "I knew that!", "I learned something", or "that’s confusing". The answers were written to be self-contained lessons, hopefully sparking enough interest people go deeper on their own.

The format had its upsides: it is fun, it finishes in 3 to 5 minutes, and people like reflecting on what they learned. But there were problems too. The word "flashcards" carries baggage tied to memorization, the card layout was too constrained, and the click-to-flip interface confused some people.

Model Mismatch

The biggest lesson had nothing to do with the UI. Writing good test questions is far harder than making good statements. A statement can be true and acceptable — for instance, that an HTTP request has four parts. But turn it into a question — "What are the four parts of an HTTP request?" — and the answer becomes contentious. Other folks naturally describe an HTTP request as having three parts (first line, headers, body). When a question doesn’t line up with how someone mentally models a topic, they push back. A better question aimed the same unsung insight at a choosable spot would be something like "Does every HTTP request have headers?", which has a clearer yes/no anchor, even if edge cases exist.

The rule that emerged: a question must match the reader’s mental model, unless the model itself is wrong. Questions get held to a stricter standard than statements.

That experience also reinforced a suspicion — multiple choice. Lots of people dislike it, and I wasn’t a fan either, so the format was dropped.

Drawing from The Little Schemer

A fresh inspiration came out of recalling The Little Schemer, the programming classic built around questions and answers presented side by side. The point of that structure isn’t to get everything right. It’s to make you consider whether you know the answer yet and to pick up new ideas as you go.

This pointed the project in two directions. First, the UI pivoted from card flips to a side-by-side question-and-answer layout on a single page. Showing the questions and answers simultaneously means you can see both at once, and you can scroll back to reread previous questions easily — giving you more control than the card format allowed.

Second, it got a new name, "questions." That matches how self-directed learning actually works: you ask yourself questions, find answers, and repeat the loop until your understanding feels right. Coming up with good questions from scratch is hard when you’re a novice in an area, so providing a curated set of (answered) questions can be a starting point.

The Last Button

Version one of the Linux-based question set omitted an "I learned something" button — and the web traffic told the story. People were tweeting scores like "I got 8/10" or "I got 10/10." That outcome focused attention on the questions you already knew, steering away from the core value: surfacing the two questions where you can learn something new.

So the button came back, along with a satisfying SVG and CSS animation for writing feedback. And the chatter started shifting from scores to comments like "I learned something." It’s a small sign, but it hints people are moving past judgment and finding the real value — a discovery of what’s worth learning next. After all, building something deceptively simple usually turns out more complicated than it looks, which is why this format still has room to grow.