How do you actually work with AI coding tools in production? From breaking down features into AI-friendly tasks? How to choose between agents vs. manual prompting? How to writing code that LLMs understand better? Where do spec-driven workflows fit in?
0:00 - Introduction 1:09 - What's the WORST thing you can do when adopting AI? 4:44 - Experimentation vs. Following Old Mental Models 7:06 - Working at Feature Level: Breaking Down AI Tasks 10:10 - Cheesy's Workflow: Brainstorming, Stride, and Task Management 13:45 - Phil's Approach: Staying in Flow State vs. Using Agents 16:01 - The Death of Prompts: Plugins and Tools Take Over 18:11 - Context Engineering vs. Prompt Engineering 21:00 - Context Window Size: Bigger Isn't Always Better 23:48 - Spec-Driven Development → Task Management Tools 25:22 - Model Wars: Anthropic vs. Open Source (Qwen, DeepSeek) 30:00 - Should You Short Anthropic Stock? (Philosophical Discussion) 33:00 - Why Claude Code Still Leads Despite Model Convergence 35:01 - Hardware Costs and the Future of AI Accessibility 38:11 - Does Boilerplate Death Change Architecture? 42:00 - When Should You Care About Code Organization with AI? 45:26 - Writing Code FOR LLMs: Semantic JavaScript and Context 47:49 - Wrap Up: Future Topics on LLM-Friendly Code
Podden och tillhörande omslagsbild på den här sidan tillhör
Zarar Siddiqi. Innehållet i podden är skapat av Zarar Siddiqi och inte av,
eller tillsammans med, Poddtoppen.
