Gemini Deep Research - Slang Dialects in 2025
Gemini Deep Research - Slang Dialects in 2025
Gemini Deep Research - Slang Dialects in 2025
Gemini Deep Research - Slang Dialects in 2025
Supabase - structured Slang data
Supabase - structured Slang data
Supabase - structured Slang data

About the Project — SlangShot

Inspiration

Language is a living organism—it mutates every day on Reddit, the unique streets of various cities, TikTok, Discord, hospital floors, classrooms, sporting arenas, and factory break-rooms. Watching slang leap across generations and cultures feels like observing evolution in fast-forward.

We wanted to build something that celebrates that creativity and fosters empathy: if I can instantly understand and even speak your dialect, I’m already seeing the world a little more from your side. Deep research by Gemini, Claude, and ChatGPT's deep-dive on linguistic evolution and cultural empathy underscored that point for us, so we turned the idea into a hackathon experiment.

What it does

SlangShot is a real-time translator for 68 documented dialects (generational, regional, professional, sub-culture, and meme-driven). Paste any text and choose:

Slang → Plain English (decode a Gen-Alpha group chat)
Plain English → Slang (re-“rizz” a marketing headline)

Slang Dialect Tree (updating weekly)

Generational

Gen Alpha

Brain Rot Core
- Skibidi / NPC Brainrot
- Ohio memes
- Italian Brainrot (absurdist faux-Italian memes)
- Fanum Tax, Glaze
Gaming Speak
- Roblox lingo (Obby, Bloxxed, ABC)
- Fortnite lingo (W-Key, Cracked, Rez)
TikTok Native
- POV-style speak, “It’s giving…”
- Viral audio memes, captionspeak
Looksmax / Mewing
- Mewing, Looksmaxxing, Mog, Sigma

Gen Z

Mainstream Vernacular
- No cap, bet, slay, vibe, bussin’, mid
Stan & Fandom Culture
- K-pop Stan
- Anime/Weeb
- Franchise Nerd Core
Therapy Speak Z
- Trauma talk, Boundaries, Narcissist energy
Chronically Online
- Twitter Discourse
- TikTok commentary style (“The way I…”, “Not me…”)
- Meme-poisoned / cursed posting
Aesthetic Subcultures
- Cottagecore
- VSCO girl (e.g. “sksksk”, Hydro Flask)
- IG Baddie (influencer talk, glam vibes)

Millennial

AIM / MSN Era
- BRB, LOL, ttyl, away messages
Hipster Millennial
- Obscure references, craft everything
LinkedIn Millennial / Corporate
- Adulting, personal brand, hustle speak
Tumblr Proto-Irony
- Social justice talk, aesthetic blogging
Core-core
- The hyper-meta “-core” trend (e.g. corecore, weirdcore)

Gen X

Grunge / Alternative
Valley Girl
Skater / Surf

Boomer

Hippie / Flower Power
Disco/Yuppie Hybrid
Boomer Facebook Dialect

Regional & Cultural

African American Vernacular English (AAVE)

Black Twitter
Ballroom / Drag
Hip-Hop Lexicon

Chicano English / Spanglish

East LA Chicano
Urban Professional Spanglish
Tex-Mex Border

UK Regional

Scouse
Cockney Rhyming
Multicultural London English (MLE)
Irish Urban
Scottish Patter

Oceanic Slang

Strine (Full Ocker)
Eshay / Lad Culture
Kiwi Colloquialisms

Southern US

Texas Talk / Cowboy
Louisiana Flavor
Appalachian

North American

Midwest American Slang
Canadian English Slang

Asian Diaspora English

Desi English (Indian English, Hinglish)
East Asian-American Slang

Caribbean & African Diaspora

Jamaican Patois
Nigerian Pidgin
South African Township English

Internet & Digital Subcultures

Gamer Speak

FPS Tactical
MOBA Salt / Esports Ranting
Speedrun / Glitch-Tech Talk

Chronically Online

Twitter Discourse
TikTok Brain / Editspeak
Reddit Moment

Crypto / Hustle Culture

Bitcoin Maximalist
NFT Enthusiast
Sigma Grindset / HustleTok

Fandom & Stan Culture

K-pop Stan
Anime / Weeb
MCU / Franchise Fan

Professional & Specialist (Refined)

Corporate / Tech Culture

Startup Hustle & LinkedIn Speak
Silicon Valley / Hacker Talk

Medical / Healthcare

ER Slang
Nursing Station Talk
Surgical Suite Banter

Law / Military / Tactical

Infantry / Grunt
Aviator / Pilot Talk
Special Ops / Tactical Jargon
Law enforcement slang

Under the hood:

Streaming GPT-4o for latency-free paraphrasing.
Dialects knowledge-base (Supabase + pgvector) researched with multiple LLMs for definitional accuracy.
Swiss-grid / Dada UI so every result looks share-ready on social media.
Mobile-first for on-the-go translations, but also a fully responsive, beautiful desktop experience

How we built it

Layer	Tooling	Notes
Deep research	Gemini 2.5 Pro, Claude 4 Opus, ChatGPT o3	Triangulated term lists, sociolinguistic context, and ethical guidelines.
Scaffolding	Bolt.new	One-click auth, routing, and edge-functions let us code features, not boilerplate.
Database	Supabase Postgres + pgvector	Tables for dialects, sub-dialects, micro-dialects, terms, definitions, citations.
API / Functions	Supabase Edge Functions	CRON pipeline that scrapes social sources, ranks new slang, and upserts records.
LLM layer	OpenAI GPT-4o (stream), gpt-3.5-turbo for background jobs	4o handles user-visible translations; 3.5 does offline classification & embeddings.
Frontend	React + TypeScript + shadcn/ui + Tailwind + Framer-Motion	Hand-crafted Dada/Swiss ink aesthetic for maximum vibe.

Challenges we ran into

Token limits inside Bolt.new’s monorepo: early prompts blew past 128 K contexts—refactored into smaller, reusable chains.
Git sync pain when we duplicated the starter repo
Dialect nuance & safety: ensuring fun output without recycling harmful stereotypes meant heavy prompt-guardrails and a moderation pass that leans on Gemini’s bias-detection heuristics.

Accomplishments we’re proud of

A fully structured, machine-readable slang lexicon that updates in near real-time—far cleaner and more contextual than crowdsourced sites.
Sub-200 ms streaming round-trip for most translations.
Un-cheugifying my own slang vocab - deadass, fr fr #millenialdad #outoftouch

What we learned

“Just translate it” hides tons of complexity—register, tone, and cultural baggage all matter.
Context is king: a single word can swing meaning across dialects; we had to build sentence-level translators, not glossary look-ups.
A disproportionate share of mainstream U.S. slang traces back to AAVE—credit and respectful usage are non-negotiable.

What’s next for SlangShot.io

Knowledge Base enhancements, leveraging the existing vector embeddings for more accurate translations
Chat Mode to have full conversations with a certain dialect.
Gamification (Rizz Check, Cheugy-meter, guess the slangified quote, etc.)
Creator Mode to bulk-export text in various dialects to reach broader audiences with messaging and marketing efforts.
RevenueCat paywalls for premium subscriptions.
11 Labs voice skins for authentic accent playback.
Dialect-to-dialect mode (e.g., AAVE → Boomer).
Broader language support (Spanish, Hausa, Tagalog) and under-represented communities.
Perf & scale: background worker queue, smarter caching, LiteLLM router to keep costs predictable.
Open-sourcing the dialect DB so academics and creators can build on top.