Why we ship three explanation tiers for every problem

The single-explanation trap

The prep market pattern is one explanation per problem, usually pitched somewhere between "expert" and "condensed." That pitch is optimized for search indexing and shallow satisfaction, not for the learner. A senior candidate already knows the pattern and skims for the trick. A beginner sees the clever solution and has no scaffolding to internalize it.

What three tiers do

ELI5 removes the jargon and replaces it with a concrete object-world walkthrough. Intermediate names the pattern, gives the minimal algorithm, and calls out two common mistakes. Expert opens with complexity, compares alternatives, and ties to related problems.

That sequence maps to the three questions a learner actually asks, in order.

1. What is this problem really about, underneath the variable names?

2. What is the canonical recipe for this class of problem?

3. When would I pick a different approach?

Why this matters more in 2026

AI inference is cheap. Serving three explanations per problem is no longer a cost decision. The opportunity cost is on the learner side, where a bad first explanation burns hours. We optimize for the learning curve, not the margin.

How we generate them

We author the prompt tree once per tier and keep it versioned. We cache generated explanations per problem so the second learner to ask a specific question does not trigger a new inference. Cache eviction follows a monthly recompute on a rolling schedule.

What it changes for you

You can pick the tier that matches where you are on this specific problem. On a topic you know cold, skip to Expert. On a topic that still feels foreign, drop to ELI5 and work your way up.