Hot Take
Hot Take is a fully automated AI newsroom.
Core trick
A Director agent fans out parallel researcher agents, then a writer, stage director, and editor agent produce a scripted broadcast with voice audio.
Description
Hot Take is a fully automated AI newsroom. It discovers trending stories from Twitter and the web, writes a scripted broadcast with two contrasting anchors, generates voice audio, and renders a complete episode in a real-time 3D newsroom — all without human intervention. Producing a daily news show requires researchers, writers, editors, voice talent, camera operators, and producers. Hot Take replaces that pipeline with six autonomous Claude agents — a Director that curates stories, parallel Researchers that deep-dive each topic, a Writer that crafts dialogue, a Stage Director that choreographs camera and animation, an Editor that runs a 13-point quality review, and a Sponsor Agent that writes contextual ads.
README excerpt
# Hot Take — AI News Show Generator
An AI-produced satirical news broadcast. Multi-agent pipeline discovers stories, writes scripts with two contrasting anchors, generates voice audio, and renders a real-time 3D newsroom with VRM characters, camera cuts, expressions, lip sync, media overlays, and sound effects.
**Built with Claude Agent SDK** | **Claude Code Hackathon** (Feb 10-17, 2026) | **PS1 — "Build a Tool That Should Exist"**
---
## How It Works
```
Topics / Twitter Lists / Grok Trends
│
▼
┌─────────────────────────────────────────────────┐
│ AGENT PIPELINE (Claude Agent SDK) │
│ │
│ Director (Opus 4.6) ── autonomous editorial │
│ ├── Researcher x3 (Sonnet) ── parallel │
│ ├── Writer (Opus) ── dialogue + emotions │
│ ├── Stage Director (Sonnet) ── camera/anim │
│ ├── Editor (Sonnet) ── 13-point review │
│ └── Sponsor Agent (Sonnet) ── ad insertion │
│ │
│ MCP Tools: twitter-list-scan, twitter-scan, │
│ twitter-search, fetch-media, WebSearch/Fetch │
│ │
│ Output: Script JSON (structured output) │
└──────────────────┬──────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────┐
│ TTS PRODUCTION (ElevenLabs) │
│ Per-segment voice synthesis → MP3 +