README.md

metadata

title: Bro WTF
emoji: ⚡
colorFrom: yellow
colorTo: red
sdk: gradio
app_file: app.py
license: mit
short_description: Same feeling, different generation.
tags:
  - build-small-hackathon
  - track:wood
  - achievement:offbrand
  - achievement:fieldnotes
  - achievement:sharing
  - gradio

Bro, WTF? -- Same Feeling, Different Generation

"That's cap" is the same sentence as "that's a load of malarkey." Your grandma's slang and your kid's slang are saying the exact same thing. Nobody invented the feeling. Every generation just puts new clothes on it.

Drop a saying from any era -- Gen Z, Boomer, 1920s, wherever -- and the app translates it across every generation. Side by side. Same feeling, different words. The parallels are obvious once you see them.

What it does

You type a saying, expression, or piece of slang from any generation. The app shows you 4-5 equivalent expressions from different eras and cultures, then explains the universal feeling underneath all of them. "Slay" is "killing it" is "nailed it" is "knocked it out of the park" is "the cat's meow." The word travels. The crown stays. Every generation thinks they invented their slang. This app shows them the receipt.

How it works

The input goes to Mistral-7B-Instruct-v0.3 via the HuggingFace Inference API with a system prompt that enforces a structured format: 4-5 translations labeled by generation or culture, followed by 1-2 sentences about the universal feeling. Temperature is set to 0.85, the highest in the suite, because this topic rewards creativity and unexpected connections. The prompt tells the model to be funny, accurate, and never condescending about any generation. If the API is unavailable, the app falls back to its handwritten response library. The structured format (labeled list plus summary) stays consistent whether the response comes from the model or the fallback set.

The fallback system

The app ships with 9 handwritten fallback responses covering "that's cap," "slay," "no cap," "it's giving," "groovy," "lowkey," "bussin," "sus," and "period." Each one maps the expression across 5-6 generations or cultures and ends with a short observation about why the feeling persists. The "no cap" fallback points out that we lie so often we need a special word for when we stop. The "period." fallback notes that every generation has found a way to put a verbal punctuation mark at the end of a sentence so hard it closes the paragraph. These handwritten entries aren't scaffolding for the model -- they are the quality target. The model generates fresh translations when available, but the handwritten set demonstrates the format, the voice, and the insight level that every response should hit.

Why this matters

Generational conflict is the cheapest fight on the internet. Every platform profits from "Boomers ruined everything" vs. "kids these days" because outrage drives engagement. But the actual linguistic evidence says the opposite: every generation is feeling the same things and reaching for the same words. "That's cap" and "tell it to Sweeney" are the same raised eyebrow separated by 80 years. This app doesn't argue about which generation is right. It shows them they're all saying the same thing. That's a harder point to make, and a more useful one.

Try it

• "that's cap" -- every generation has needed a way to say "I don't believe you and I want you to know I don't believe you"
• "slay" -- from ballroom culture to your aunt's Facebook, the word for "nobody could look away"
• "lowkey" -- the latest version of whispering an opinion so you can deny it later if it goes badly
• "bussin" -- food has always needed its own category of compliment beyond regular good

Built with

• Gradio
• Mistral-7B via HuggingFace Inference API
• Python
• Zero external dependencies beyond gradio and huggingface_hub

Process notes

The prompt engineering for this app required solving a structural problem, not just a tonal one. The output needs to be a formatted list (labeled by generation) followed by a summary -- two different writing modes in one response. The system prompt specifies "don't explain what each one means individually, just show them side by side" because the parallels should be self-evident. Readers see the pattern faster when you don't narrate it for them. The summary at the end is there to name the underlying feeling, not to explain the translations. Getting the model to consistently produce this two-part structure without breaking format took several rounds of prompt tuning.

The research behind the handwritten fallbacks was the most time-intensive part of this build. Mapping slang accurately across generations requires knowing not just what the words mean, but when they peaked, which community they came from, and whether the equivalence is real or superficial. "Slay" originates in ballroom culture -- including that origin isn't trivia, it's respect. "No cap" has roots in older AAVE that predate Gen Z by decades. The fallbacks had to be funny without being shallow, and accurate without being academic. The summary lines are where the real writing lives: "we lie so often we need a special word for when we stop" is the kind of observation that makes someone send the app to a friend.

The UI uses a warm yellow-to-red gradient that feels energetic and playful, matching the tone of the content. The input prompt says "Drop a saying from any generation" -- informal, inviting, no instructions needed. Clickable example buttons let users try the most recognizable entries immediately. The app doesn't require any knowledge of slang history to use. You can type what your kid said at dinner and find out your grandmother said the same thing in 1955. That moment of recognition is the entire point.

---

Heuremen -- Let's stay connected and keep information free.