Frequently Asked Questions

ARTFunny investigates the question, "Can a machine have a sense of humor?" by first asking what humor requires from anyone. The site combines comedy theory, joke structure, history, performer studies, mechanism guides, and interactive experiments.

The goal is not to declare humans or machines the winner. It is to help visitors notice setup, expectation, timing, misdirection, context, wordplay, persona, and audience response. AI is both a subject of that investigation and a tool used in several clearly labeled experiments.

ARTFunny is built with the Next.js App Router, React, TypeScript, and custom responsive CSS. Next.js renders the public pages, metadata, sitemap, server components, and secure server endpoints used by the live tools.

Editorial pages are stored as Markdown with validated frontmatter and rendered with Gray Matter and Remark. Structured experiment content and fallback examples use JSON. Zod validates live tool inputs and structured AI responses. Images are served through the Next.js image system, and JSON-LD supplies structured metadata for search engines.

The site does not require a visitor account or a community-results database. Its published articles, curated datasets, and prebuilt demonstrations remain useful independently of visitor traffic or model availability.

The browser sends the form to an ARTFunny server endpoint. The server checks that the request came from the site, applies input-size and field validation, enforces a request limit, and sends the submitted text through OpenAI moderation before generation or analysis.

If the input passes those checks, the server calls the OpenAI Responses API and requests a structured result matching the tool's schema. The validated result is returned to the browser and labeled as live AI output. The OpenAI API key remains on the server and is never included in browser code.

Every AI-assisted experiment includes a complete prebuilt demonstration. The page therefore remains understandable when an API key is missing, the provider is unavailable, a request times out, or a visitor chooses not to submit text.

The interface distinguishes a "Live AI result" from a "Prebuilt fallback demonstration." A fallback is not presented as though it was generated from the visitor's current submission.

Live endpoints use same-origin checks, input limits, schema validation, moderation, request throttling, timeouts, and handled error states. The current application limit is ten AI requests per network address during a ten-minute window, although deployment architecture can affect how that limit is applied.

These controls reduce risk; they do not make generated content perfectly safe or correct. AI jokes and explanations can be dull, inaccurate, culturally insensitive, overconfident, or simply unfunny. Live responses are not reviewed before they appear, and analysis is presented as interpretation rather than fact.

No. The homepage Joke of the Day is retrieved from API Ninjas through a separate server-side request and has its own curated fallback. It is a third-party joke feed, not one of the site's OpenAI-powered experiments.