Watch us on YouTube: https://www.youtube.com/watch?v=CuPIxJjKOY8
Andrew and Nathan are back to discuss how they actually use AI for practical work, specifically for software development and coding.
Nathan kicks things off by explaining his preference for CLI (Command Line Interface) based agents, like Claude Code. He details his evolution from treating AI like a “junior software engineer” to now pushing the boundaries with advanced “hive mind” or “swarm” frameworks. This approach involves spinning up multiple AI agents with specific roles—like a coder, an architect, a QA tester, and even a documentation agent—to tackle complex features simultaneously.
Andrew shares his completely different approach, which he calls vibe coding?! He primarily uses the chat window in Gemini, leveraging customized “Gems” that are pre-instructed for specific tasks. He has different Gems for architecture, reasoning, and coding. A key instruction he bakes in is “don’t assume anything, ask for clarification,” which helps avoid the common AI “looping problem.” By grounding his AI Gem with his project’s code repository, he can iteratively challenge the AI, test his own assumptions, and rapidly build features.
The conversation explores how these AI-assisted methods have dramatically accelerated development time, cutting app prototyping down from six months to just a few weeks. They conclude by sharing practical, non-technical tips that anyone can use, whether you’re coding or analyzing a spreadsheet, emphasizing the importance of breaking down large problems and clearly defining the AI’s context.
Support the pod: https://3reate.com, https://ko-fi.com/3reate or https://patreon.com/3reate
Chapters:
(00:00) Hello world!
(01:22) Nathan’s approach: Using AI for coding.
(03:15) Advanced AI: Using “hive minds” (swarms).
(06:56) Vibe coding? with Gemini Gems.
(09:26) Tell AI “don’t assume anything.”
(13:50) AI speeds up app development significantly.
(16:51) Best skill: Breaking down massive problems.
(21:33) The importance of bounding the AI’s context.
(24:07) Pro-Tip: Ask AI to write better instructions.
