Episode 809 | What I Learned Diving into A.I. for 100 Days (with Craig Hewitt)

What It Covers

Craig Hewitt shares lessons from creating 100 AI videos in 100 days, reveals why founders should abandon ChatGPT for superior tools, and announces his strategy for building a second product under Castos.

Key Questions Answered

• •Manus over ChatGPT: Manus uses Claude underneath with agentic capabilities, browser access, and code execution. It can autonomously research markets, download transcripts from 20 YouTube videos, analyze patterns, and create frameworks—tasks ChatGPT cannot handle reliably or completely.
• •Customer support automation ROI: DocSpot reduced Castos support burden by 50% for 4,000 customers at $80 monthly cost. The tool uses vector stores of knowledge base content, includes escape hatch to human support, and improves documentation quality through conversation monitoring.
• •Second product validation criteria: New products must start at minimum $100 monthly pricing, include built-in expansion revenue through usage or upgrades, and align with existing customer base. Selling $20 monthly products creates unsustainable unit economics for sustainable growth.
• •AI agent architecture: Functional AI agents require three components—an LLM for intelligence, memory to avoid redundant processing, and access to multiple tools like email, calendar, or knowledge bases. Automations string together conditional logic while agents make autonomous decisions.