Kent talks with Dax Raad about building OpenCode in a crowded coding-agent market: why dev tools are still a consumer-style product, how fast shipping can make good products feel worse, and what "product skill" actually looks like when agents remove friction from implementation. They dig into onboarding, progressive disclosure, listening across many user requests for the real pattern, and why slowing down can be the right move—even when competitors ship faster.
Dax has spent years building tools developers actually use; on OpenCode he's thinking hard about product process while the space moves at breakneck speed. This episode is a practical look at product deterioration (not just code rot), bottom-up adoption for dev tools, and how coding agents change who decides what gets built—without replacing the need for taste, restraint, and clarity about what problem you're solving.
You'll hear concrete examples from OpenCode's terminal UI and onboarding, parallels to Kent's Epic Workshop app, and a grounded take on inference pricing, hype, and when "ship messy and fix later" does and doesn't hold up.
