Descriptions:
Gergely Orosz of The Pragmatic Engineer sits down with Dax Raad, co-founder of OpenCode, the most widely used open source coding harness, which has grown to roughly 10 million active users in under a year. The conversation is notably candid: despite building one of the most popular AI coding tools in existence, Raad says he’s working as hard as he ever has and that AI has not eliminated the hard problems of software engineering — it’s made many individual tasks easier while leaving the fundamentally difficult parts intact.
On the business side, Raad reveals that OpenCode Zen — the inference service the team built to smooth onboarding — hit a $50 million annualized run rate within five to six months of launch, largely driven by demand for aggregated access to open source models at reliable rate limits. A second revenue line, an enterprise control plane for organizations deploying OpenCode at scale, is in private deployment and approaching public availability. Raad also discusses how OpenCode is bottlenecked by GPU supply even as an open source project, and shares the internal memo he sent his team acknowledging they had been shipping too many features and accumulating architectural debt.
The episode covers competitive dynamics in the coding agent space, why Raad deliberately chose not to add a viral sharing prompt to OpenCode, and a broader discussion of what companies actually need from AI tooling versus what they believe they need. It is one of the more grounded first-person accounts from inside a fast-scaling AI developer tools company.
📺 Source: The Pragmatic Engineer · Published May 27, 2026
🏷️ Format: Interview
