June 16, 2026
177: "Is Typing Becoming Old-Fashioned in the AI Age?" ft. Justin Coats
Erik and Justin take a practical tour through AI “tools that actually ship.” They start with Lovable to build a real landing page fast, then move to NotebookLM for source-grounded research and repackaging, and finish with Spinach AI for meeting intelligence that turns conversations into executable next steps.
🧭 Conversation Highlights
- Erik’s Lovable experience: a subsite built in minutes by feeding a prompt and letting the tool pull branding, structure the story, and generate interactive components like email capture and a database
- Voice vs typing in AI tools: speaking helps you move at the speed of thought, surfaces gaps in what you can articulate, and makes iteration easier
- Lovable’s workflow options: Build mode for speed versus Plan mode for a more production-ready blueprint you can edit before publishing
- A “human first” approach to using tools: have the conversation with the interested person, then use AI to turn that learning into assets like websites, decks, and meeting summaries
💡 Key Takeaways
- If you want results quickly, talk to the tool (dictate) instead of trying to perfectly write your thoughts first. It reduces breaks in your chain of thought and speeds iteration.
- Use conversation as your forcing function. Real questions from a real person help you figure out what you actually need before you ask AI to build it.
- When accuracy matters, ground AI output in your sources. NotebookLM’s citation behavior helps you verify without doing a ton of manual reading.
- Meeting intelligence tools can eliminate the “regurgitate the notes” overhead. The value is turning transcripts into tasks your team can execute.
❓ Questions That Mattered
- How do you get past the awkwardness of speaking to AI so you can describe what you want faster?
- When should you choose Lovable’s Build mode versus Plan mode for production readiness?
- What kinds of research or content work should be source-grounded to avoid hallucinations?
- How can meeting transcripts be turned into actionable tasks without relying on scattered human memory?
🗣️ Notable Quotes
- “Get comfortable talking to your computer.”
- “Start with a conversation with somebody who’s interested in what you’re doing, and then go to the AI tool.”
- “Stop limiting yourself. Jump in, play around, have fun. You’re not gonna break it.”
🔗 Links & Resources




