Descriptions:
Five-time startup founder Ryan Carson walks through how he runs his company nearly solo using a suite of AI agents, with his OpenClaw instance “R2” functioning as a full executive assistant. The core system, which Carson calls Claw Chief, is built from markdown files and cron jobs: R2 sweeps his inbox every 15 minutes, parses Calendly links to book meetings, sends proactive follow-up emails, and runs a nightly prospecting job that searches for potential business development contacts, adds them to a Google Sheets CRM, and sends cold outreach on Carson’s behalf. He reports booking 10–20 cold meetings in just a few weeks through this fully automated pipeline.
Carson manages R2 primarily via OpenAI Codex on a $200/month ChatGPT Pro subscription — which he estimates delivers $2,000–3,000 in token value at current pricing — while also running Claude Code and Devin in parallel sessions. The MacBook Pro running R2 lives in his closet and is accessed remotely over SSH via Tailscale. He prefers Firecrawl over the Brave Search API for agent web crawling, and uses Slack (not Telegram) for all agent communication due to its threading model. A nightly Git backup ensures the entire agent configuration is recoverable.
Carson’s most pointed insight is that startup doctrine has inverted: where founders once minimized documentation to ship faster, comprehensive CLAUDE.md-style reference files and well-structured cron jobs are now the prerequisite for unlocking 10x agent leverage. He says he ships at least 10 pull requests per day and finds onboarding agents easier than onboarding human employees — because the agent always retains its training.
📺 Source: Peter Yang · Published May 24, 2026
🏷️ Format: Workflow Case Study
