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

Key Takeaways Copied to clipboard!

• For founders exploring AI tools, Craig Hewitt recommends Claude Code (for bare-metal coding/agentic work) and Manus (an agentic AI tool using Claude with a web interface) over ChatGPT for serious productivity.  Copied to clipboard!
• An AI agent is defined by three components: an LLM, memory, and access to tools (like the web, calendar, or knowledge base), distinguishing it from simple chatbots or standard automations.  Copied to clipboard!
• Craig Hewitt is pivoting to build a second, high-value product (targeting at least $100/month with expansion revenue) because Castos, while stable, has hit a natural growth plateau after eight years.  Copied to clipboard!

Segments Expand All Collapse All

Sponsor Message and Episode Intro

Copied to clipboard!

(00:00:00)

• Key Takeaway: Ahrefs Brand Radar tracks brand visibility across AI platforms like ChatGPT and Perplexity.
• Summary: Ahrefs has launched Brand Radar to track brand visibility across search, social, and new AI platforms. Host Rob Walling introduces Craig Hewitt to discuss his ‘100 Days of AI’ series. Hewitt reveals a strong preference for tools other than ChatGPT for SaaS founders.

Craig’s 100 Days of AI Learnings

Copied to clipboard!

(00:04:33)

• Key Takeaway: AI tools currently function reliably about 80% of the time, leaving a necessary 20% human-in-the-loop augmentation.
• Summary: The ‘100 Days of AI’ series was inspired by Brendan Hufford’s ‘100 Days of SEO’ challenge. Hewitt found that while complex AI workflows exist, they often fail 20% of the time, requiring human oversight. This dichotomy means AI is currently amplifying human work rather than entirely replacing jobs.

Best-in-Class AI Tools Review

Copied to clipboard!

(00:08:20)

• Key Takeaway: Manus, an agentic AI tool using Claude with a web interface, is recommended as the single most valuable tool over ChatGPT for complex tasks.
• Summary: Claude Code is highlighted as the superior tool for coding and agentic tasks, acting like bare-metal AI access without a chat interface. ChatGPT is relegated to a consumer product, with its main utility for founders being reusable, trained GPTs for specific tasks. Manus is described as Claude Code with a web interface, capable of complex actions like researching and downloading data from multiple web sources.

Defining AI Agents and Use Cases

Copied to clipboard!

(00:15:35)

• Key Takeaway: A successful AI agent requires an LLM, memory, and access to multiple tools, differentiating it from simple chatbots or automations.
• Summary: An AI agent possesses three components: an LLM (like Claude or ChatGPT), memory to recall past interactions, and access to tools (web, calendar, vector stores). An agentic customer support tool, DocSpot, has successfully cut Castos’s support burden in half by referencing their knowledge base. Monitoring chatbot logs improves documentation quality because the bot only references existing knowledge.

Burnout and YouTube Production Secrets

Copied to clipboard!

(00:21:42)

• Key Takeaway: Producing 100 videos in 100 days relied heavily on an editor handling 85 videos and leveraging existing content ideas from the space.
• Summary: Hewitt experienced an identity crisis around episode 70, realizing much of the experimental AI building was worthless, pivoting to product building instead. Ideation was managed by borrowing concepts from other channels and news-jacking timely AI releases. An editor handled the bulk of the production, and Creator Hooks was used for title and thumbnail suggestions.

Strategic Need for Second Product

Copied to clipboard!

(00:25:56)

• Key Takeaway: Castos has reached a natural growth plateau, necessitating a new product layer to achieve the significant growth required for a $100M valuation.
• Summary: Castos is an eight-year-old, stable, profitable business, but organic growth efforts have plateaued despite hiring talented marketers. Hewitt is evaluating three new product ideas that must start at $100/month and leverage the existing 40,000-person audience. LinkBerry.ai, an AI tool for generating LinkedIn content, is one compelling option designed to solve the pain point of thought leadership creation.

Emotions of Launching Second Product

Copied to clipboard!

(00:33:09)

• Key Takeaway: Founders often feel a mix of excitement for new growth potential and anxiety about repeating the initial startup struggle when launching a second product.
• Summary: Building a second product is necessary because the primary business has hit its S-curve limit, making significant growth unlikely otherwise. Hewitt is excited by LinkBerry’s compelling value proposition of creating a month of LinkedIn content for $100. The anxiety stems from starting from scratch after eight years of stability and questioning if the initial success was repeatable.