Generational Food Vibe Check

Features

Four Generational Voices

Gen Z, Millennial, Gen X and Boomer each get their own rewrite of the same dish — title, ingredients and steps.

Mood-Based Menus

Clean Girl, Gremlin Mode, Broke AF and Payday. Each mood carries its own palette and its own food list.

Let Fate Decide

A betting mechanic that picks dinner for you when you cannot, with a running "current winner".

Community Suggestions

Anyone can suggest a dish with a photo. Submissions land in a moderation queue rather than going live.

Admin Review Queue

Approve or reject suggestions, upload imagery, and trigger recipe generation from a protected panel.

Zero-Cost Reads

Visitors are served cached results from edge storage. No visitor action ever bills an AI call.

How it works

1. Choose your lane: Gen Z, Millennial, Gen X or Boomer. This decides the voice everything is written in from here on.
2. Pick a mood: Clean Girl, Gremlin Mode, Broke AF or Payday — each with its own colour scheme and menu.
3. Let fate decide: Cannot choose? Place a bet and the app picks for you, then crowns a current winner.
4. Cook it: The full recipe appears in your generation’s voice — served instantly from cache, no AI call, no waiting.

AI with a spending limit

The obvious way to build this — call the model whenever someone opens a page — is also the way to wake up to a bill you did not budget for. Recipes are instead generated once by an admin, reviewed, and written to edge storage. Traffic and AI cost are completely decoupled.

• Generation is admin-triggered and never visitor-triggered
• Results cached in Cloudflare KV and served from the edge
• Cost stays flat whether the app gets 10 visits or 100,000

Personality as a design constraint

Writing one recipe is easy. Writing the same recipe four times so each version sounds genuinely like the generation it is aimed at — without becoming a parody — is a prompt-design problem, and it is exactly the problem behind any branded AI feature.

• A distinct voice per generation, held consistent across every dish
• Structured output — title, ingredients, steps — not free text
• Every result reviewed before it reaches a visitor

Serverless from end to end

No servers, no containers, nothing to patch. The whole app is static files plus edge functions, with object storage for uploads — the same architecture we would use for a client product that has to stay cheap and available.

• Cloudflare Pages Functions for all API routes
• KV for the catalog, R2 for user-uploaded images
• Secrets held outside the codebase, never in config

In numbers

• 4 Generational voices per dish
• 0 AI calls triggered by visitors
• 100% Served from the edge

At a glance

Why we built it

This one is not enterprise software, and that is the point. It is a small consumer app that shows what a language model does when you give it a voice and a hard budget. The AI never runs for a visitor — every recipe is written once, reviewed, and cached at the edge. The result feels instant, costs almost nothing to serve, and does not get more expensive when it goes viral.