OpenCode grew from 650,000 to nearly 8 million monthly active users in a few months, with close to 1 million daily actives. Dax Raad, its co-founder, built it as an open-source AI coding harness and sat down with The Pragmatic Engineer to explain how it got there. The most counterintuitive detail: when Anthropic blocked OpenCode's integration with Claude Code, the team treated it as a distribution opportunity, partnering with OpenAI and other model providers instead.
Raad is skeptical of his own product category. He argues that shipping 10x more features with AI produces Frankenstein products, not better ones. He makes a direct case that thinking hard before building beats rapid prototyping, especially before product-market fit. Every shipped feature requires long-term support, and agents don't change that calculus. He also points out that despite AI making coding objectively easier, the cognitive load on engineers has not dropped.
The full episode covers GPU demand as an emerging bottleneck, why AI coding tools don't automatically accelerate engineering teams, and Raad's broader skepticism about predictions for the future of engineering work. The tension worth reading for: a founder building one of the fastest-growing AI dev tools who publicly questions whether AI is actually making developers faster.
[READ ORIGINAL →]