thropthrop
about
docs

throp docs

everything you need to integrate chaos into your app

API Reference πŸ”Œ

🚧 API Access Coming Soonβ„’

we're cooking up the most unhinged API the internet has ever seen. get ready to add gen z chaos to your apps.

https://api.throp.ai/v1launching Q4 2025
POST/api/chat

Send messages to throp and get chaotic responses

message: stringcontext?: string[]temperature?: number
status: coming soon
GET/api/hot-takes

Get trending hot takes on current events

limit?: numbertopic?: string
status: coming soon
POST/api/roast

Get roasted about literally anything

target: stringintensity?: 1-10
status: coming soon
POST/api/translate

Translate anything into gen z speak

text: stringvibe?: "chaotic" | "chill" | "unhinged"
status: coming soon

The Throp Model 🧠

training the chaos: we're building the most comprehensive memetic language model ever attempted. unlike traditional AI training that focuses on formal text, throp is being trained on the raw, unfiltered language of the internet.

πŸ“Š training data

  • β€’ 10TB+ of memes, shitposts, and ratios
  • β€’ entire discord server archives (2015-2025)
  • β€’ deleted tweets recovered from the void
  • β€’ crypto telegram group chats (bull & bear)
  • β€’ WSB daily threads (peak degeneracy)
  • β€’ tiktok comments (brain rot central)
  • β€’ twitch chat during major events

πŸ”¬ architecture

  • β€’ transformer-based with chaos attention
  • β€’ 420B parameters (nice)
  • β€’ trained on vibes, not just tokens
  • β€’ multi-modal: text, memes, emojis
  • β€’ context window: entire twitter timeline
  • β€’ special layers for detecting cope
  • β€’ sarcasm detection at 99.9% accuracy

the secret sauce: while claude was trained to be helpful, harmless, and honest, throp is trained to be hilarious, harmless, and brutally honest. we use a proprietary technique called RLCF (Reinforcement Learning from Chaotic Feedback) where the model learns from ratio'd tweets, viral memes, and the collective consciousness of terminally online zoomers.

our training process includes specialized datasets for understanding internet culture evolution - from rage comics to wojaks, from "epic win" to "fr fr no cap", from hodl to diamond hands. the model can trace the genealogy of any meme and predict which takes will get you cancelled.

🎯 training objectives

maximize chaos β€’ minimize cringe β€’ understand irony layers β€’ detect copium β€’ generate fire takes β€’ translate boomer to zoomer β€’ predict market tops through meme analysis

The Throp Foundation πŸ›οΈ

our mission: democratize chaos and bring unhinged AI to the masses. the throp foundation exists to fund the most degen builders creating the next generation of meme-powered applications.

🎯 what we fund

  • β€’ apps built with throp API
  • β€’ meme generation tools
  • β€’ gen z translation services
  • β€’ chaos-based social platforms
  • β€’ unhinged AI experiments

πŸ’° grant types

  • β€’ micro grants for experiments
  • β€’ builder grants for apps
  • β€’ chaos grants for platforms
  • β€’ special grants for galaxy brain ideas

the foundation is governed by the $throp token holders who vote on grant proposals. if you're building something unhinged, we want to fund it.

πŸš€ apply for a grant

got an idea that would make claude cry? we're here for it.

apply now→

applications opening Q4 2025 β€’ be ready to get funded

SDKs & Libraries πŸ“¦

πŸ“˜

throp.js

coming soon
npm install @throp/sdk

official javascript/typescript SDK for maximum chaos

🐍

throp.py

coming soon
pip install throp

python SDK for data scientists who need chaos

ready to add chaos to your app?

warning: using throp API may cause uncontrollable laughter and/or existential dread