The Understanding Funeral

Inspiration

Every learning app assumes an empty vessel. Flashcards, quizzes, tutorials — they all try to pour correct information in. Nobody asks what's already in there that shouldn't be.

The enemy of learning isn't ignorance. It's being confidently wrong. You cannot teach someone who already believes they know. You have to undo the wrong belief before the correct one has anywhere to go.

No tool addressed this. So I built one.

What it does

Pick any topic. Explain it with full confidence — no hedging.

Then the confrontation begins. The AI doesn't correct you. It asks you to make a specific, testable prediction using your stated understanding. Then it shows you what actually happens.

If you were wrong — and you usually are in interesting ways — the AI says:

"Before we go further — we need to hold a funeral."

A memorial card appears. Four fields, pre-filled by AI:

• What I believed — your wrong understanding, stated clearly
• Where it came from — why it felt true
• What it cost me — real consequences of believing it
• What I now understand instead — the correct version in your own words

You click Lay it to rest. It's buried. Added to your Cemetery.

After 5 funerals, your Misconception Fingerprint unlocks — the pattern of how you get things wrong, not just what you got wrong.

How we built it

Built entirely on MeDo using the Large Language Model plugin for:

• The confrontation phase — generating specific, testable predictions based on the user's stated understanding
• Pre-filling all four funeral fields with honest, specific content
• Teaching the correct version on the Learning Page
• Analyzing patterns across all buried misconceptions to generate the Fingerprint

The dual visual aesthetic — clinical white for confrontation, warm memorial for the funeral — was specified in the requirements and built by MeDo in a single pass.

Cemetery data persists in browser local storage with shareable public URLs generated per user.

Challenges we ran into

The confrontation logic was the hardest prompt engineering challenge. The AI must not correct the user immediately. It must ask for a prediction, let the user commit to it, then reveal the reality. Getting this sequence to feel natural — not like a trap — required careful instruction about tone and pacing.

The Fingerprint feature required the AI to analyze patterns across multiple misconceptions rather than within a single one. Identifying that someone "consistently anthropomorphises abstract systems" from five different wrong beliefs about five different topics is a genuinely complex inference task.

Accomplishments that we're proud of

The funeral itself. It sounds gimmicky until you do it. Writing the obituary for something you believed — where it came from, what it cost you — creates a psychological closure that simple correction never does. Users report that misconceptions they've "corrected" before finally feel gone after the funeral.

Also proud of the Cemetery page. A visual graveyard of everything you used to believe, with dates and full memorial cards on hover. People screenshot it. People share it. It's the most personal artifact any learning app has ever produced.

What we learned

Ceremony matters in learning. The ritual of the funeral — the heading, the four fields, the animation, the button — creates emotional weight that makes the correct understanding stick differently than a correction ever would.

Also learned that the Misconception Fingerprint is the feature users most want to share. Knowing how you're wrong — not just that you're wrong — feels like genuine self-knowledge. That's rare.

What's next for The Understanding Funeral

• Classroom mode — teachers assign topics, view student cemeteries, identify shared misconceptions across a class
• Cross-device sync — server-side storage so the cemetery follows you across devices
• Export — download your full cemetery as a PDF or shareable image
• Subject-specific confrontation modes — tuned for STEM, history, economics, medicine
• Spaced repetition integration — resurface buried misconceptions at intervals to confirm they stayed buried

Built With

• browser
• css
• html
• javascript
• large-language-model-(llm)-api
• localstorage
• medo