Shuttle Content Machine
The Problem
Consistent, high-quality LinkedIn content is a full-time job. As a CRO, I needed to stay visible and relevant across payments, fintech, voice AI, travel, and insurance — without spending hours each day writing.
What I Built
An AI content engine that runs every weekday, completely autonomously:
- Scans 20+ industry news sources across payments, fintech, voice AI, travel, and insurance verticals
- Selects a content pillar based on a weekly calendar and live news signals
- Writes a LinkedIn post in my voice — calm, strategic, diagnostic (Don Draper-inspired)
- Generates branded quote card images automatically
- Presents posts for approval via an interactive terminal UI
- Publishes approved content via Zapier to Radaar scheduling platform
Every post runs through a quality checklist before output. No AI vocabulary, no hype, no em dashes.
Video Pipeline
The system also produces short-form video content through a 6-phase workflow:
- Brief → Script → Storyboard → Audio → Test Clips → Final
- Text-to-speech via ElevenLabs with my voice clone
- AI video generation using Sora 2, Kling, Veo, and Hailuo models
- Automated cropping, syncing, and stitching with FFmpeg
Tech Stack
- Claude Code CLI with subagents for parallel processing
- Node.js with Canvas, Sharp, and FFmpeg
- ElevenLabs API for voice cloning
- Kie.ai API for AI video generation
- Zapier webhooks → Radaar for publishing
Content Framework
- 5 Content Pillars: Industry News (30%), Prospect Signals (20%), Shuttle Updates (20%), Thought Leadership (20%), Blog Promotion (10%)
- Writing voice: Don Draper structure — Observation → Structural Shift → Hidden Tension → Category Framing → Light Shuttle Placement
- Post length: 300–500 words
Results
100+ days of daily content generated and published. Consistent LinkedIn presence without the daily time investment.