What Is This Duck For?

An absurd web app that asks AI to analyze a plastic duck with far too much seriousness.

By Héric Libong
What Is This Duck For?

Project origin

What Is This Duck For? started from a simple challenge: build an intentionally useless application, but make it clear, coherent, and polished enough for that uselessness to become the actual subject of the project.

The starting point is a trivial object: a yellow plastic duck. The application treats it as a serious, almost administrative matter, asking an AI to produce solemn explanations about its supposed purpose.

The point of the project is not to solve a practical problem. It explores a contrast instead: take an object with no stakes and apply the language of an official investigation, a ministry report, or an overconfident symbolic analysis.

The result works as a small experiment in interface, writing, and tone. The duck stays at the center, but every answer turns that banal object into a pseudo-scientific, mystical, bureaucratic, or openly suspicious diagnosis.

🔗 Live demo: ricolajazz-whatisduckfor.hf.space

Project overview

The application offers a short, theatrical web interface: a central duck, a report panel, very formal labels, and a visual mood inspired by official documents.

The user is not looking for a useful answer. They submit the same fundamental question: what is this duck for? The AI then replies according to the selected interpretation mode.

Five analysis modes

  • Analyze the Duck: a pseudo-scientific reading of the duck’s function
  • Request Deeper Interpretation: an over-intellectualized and symbolic interpretation
  • Escalate to Ministry: a bureaucratic response where the duck gains ridiculous administrative importance
  • Trust the Duck: a mystical reading that treats the duck as a silent authority
  • Do Not Trust the Duck: a paranoid interpretation that refuses to believe in the duck’s innocence

Each mode produces a short, structured, absurdly serious report. The writing constraint matters: the AI must stay readable, controlled, and specific to the selected mode, without drifting into generic noise.

Escalate Ministry Duck Persona

Technical stack

Backend and AI generation

  • Python and FastAPI to serve the web application
  • Google Gemini API to generate the absurd reports
  • Separate prompts by persona to keep a distinct tone for each selected mode
  • No user accounts, no database, no dashboard

Frontend

  • Jinja2 for HTML templates
  • Vanilla HTML, CSS, and JavaScript for the interface
  • Visual staging centered on the duck and the report panel
  • A deliberately lightweight architecture, suited to a short and demonstrable project

Project status

The project is a finished concept: the FastAPI application exists, the main interface is in place, the five analysis modes are implemented, and report generation via Gemini is operational.

It is not designed as a product meant to evolve forever. Its value comes from its tight scope: an absurd idea, a coherent interface, a simple AI mechanism, and a tone serious enough for the joke to work.