When the Quiz Needed Rules

THE_SITUATION: THE_SCORING_WALL

The curriculum map gave Tutorroo a basic shape. It told me what the learning journey should cover, how the concepts should build, and where the obvious gaps lived.

Then I hit the less glamorous problem: what happens after a quiz session ends.

Not the simple version where each answer gets a tick or a cross. That part is easy. The hard part is deciding what the whole session means.

If my son struggles through a quiz, should the next set repeat the same skill, drop the difficulty, or explain the idea another way? If he flies through it, should the app reduce repetition, lift the difficulty, or rotate to a nearby skill before he gets bored?

This is where the product stopped being a cute quiz and started needing a bit of judgment between sessions.

THE_COMPLICATION: GOOD_INTENT_BAD_RULES

Gemini had been brilliant for the big picture. It helped me sort the messy pile of dad-hunches, education theory, and product notes without dropping the thread.

But scoring logic needed a different kind of help.

I did not need more roadmap. I needed someone to sit inside the mechanism with me and ask annoying questions like: how many questions is enough evidence? What counts as a wobble versus a real gap? How much should one bad session affect the next quiz?

My own JavaScript was not the only bottleneck. The bigger problem was turning a parent instinct into rules that did not make the app feel like a worksheet with a nicer font.

So I brought in Claude as the logic buddy.

The conversation felt different from asking for code. I was not saying, "build me a scoring engine." I was saying, "after a rough quiz session, what should the next quiz actually look like?"

That turned out to be the better question.

THE_RESOLUTION: TALKING_THROUGH_THE_LOGIC

We pulled the scoring idea back to something simple: look at the session, then decide the next set.

Poor run? Repeat the skill and make it easier. Strong run? Reduce the repeats and lift the difficulty. Messy run? Hold the level and get more signal.

That was the useful part of working with Claude. It helped turn a vague dad instinct into a few rules I could actually test.

Next_Action: Turn the session scoring chat into a small rules baseline before the next AI touches the quiz loop.

Vibe: Collaborative / Slightly Less Clueless