Building OpenCode with Dax Raad: Conversation with Open Source Coding Agent Creator (Category AI4SDLC)
I saw it. fresh-out The Pragmatic Engineer, where Gergely Orosz talks to Dax Raad, the creator of OpenCode. It’s not just about the history of the product. This is a rare case where the author of one of the fastest growing AI coding tools consistently refuses to sell magic and speaks about AI with noticeable skepticism.
OpenCode is an open-source coding agent that started out as a terminal-based tool and grew to a GUI. According to Dax, the product has grown by 650 thousand 8 There are millions of MAUs in a few months, and about a million people use them every day. The numbers are impressive, but the story behind them is more interesting.
1The most revealing episode is the lockdown from Anthropic. When Anthropic shut down OpenCode from working through Claude subscriptions, it looked like an existential threat to the product. Dax describes exactly the opposite: the team turned to OpenAI and other model providers, and this episode became a growth accelerator, not a blow. The lesson Dax draws from this sounds almost like a product manifesto: proper positioning is more important than speed. OpenCode didn't win because its harness was the best - Dax honestly admits that the early months were just good enough. He won because he was the first to take the category of open-source coding agent, which no one had time to stake out. "Get positioning right and the world just keeps handing you wins" - its wording. First position and market share, then finish quality.
2He also talks about the business model quite directly. The company makes money through OpenCode Zen, and Dax calmly spreads the economy of inference: GPU is a capital asset that is depreciated, and tokens are sold, in his estimation, with a margin to 90% depending on the model. Against the background of conversations about "AI is unprofitable for all" useful optics, although it is worth remembering that this is the assessment of a person who earns on inference.
3The hottest part of the issue is about productivity. Dax says something like this: before AI he spent 95Percentage of energy to think about what to do 5Percent on the execution itself. Now-- 96per cent for thought and 4Percent on execution. Formally improving, but day by day work feels just as hard. AI removes doing, but does not remove thinking, and the bottleneck has always been thinking. He calls confident predictions about the future of AI a form of self-reassurance: people usually predict an advantage for their own group.
4I also liked the section on engineering culture. Dax describes a picture that is familiar to many: engineers who care about quality are drowned in slop PRs from colleagues who don’t care about it, and burn out in the rake. The second observation is subtler: AI relieves psychological friction - guilt over a cut corner, and tech debt begins to accumulate unnoticed. But there is a symmetrical plus: refactoring by agents has become cheap, and it is worth using it more aggressively than we are used to. It’s funny that this logic returns “enterprise” patterns like domain-driven design: they are again used as guardrails – for both junes and agents.
5Dax’s career advice is simple: a future-proof combination is solid knowledge in software engineering plus deep domain expertise. Engineers systematically underestimate the second part.
For me, the main conclusion is that when the creator of a tool that grew up on the wave of AI coding says that AI changes less than it seems, it is worth listening more carefully than the presentations of vendors. The bottleneck shifts not to code generation, but to thinking, quality, and domain ownership. It's worth investing in.
#AI #AI4SDLC #Engineering #Management #Product #Software #Architecture