Dax Raad, co-founder of OpenCode, discusses the rapid growth of his open-source coding harness to nearly 8 million monthly active users in less than a year. He explores the paradox of AI speeding up coding while not necessarily improving software quality, highlighting challenges like shipping too many features, accumulating technical debt, and the importance of slowing down to maintain quality.
Summarized by Podsumo
OpenCode grew from 650,000 to 8 million monthly active users in just a few months, with a notable spike in January 2026 due to Anthropic's ban on Claude subscriptions, which backfired and boosted OpenCode's adoption.
Dax emphasizes that AI agents can lead to over-shipping features and accumulating 'hacks,' as the natural prickle engineers once felt when writing hacks is muted, skewing judgment.
Inference businesses are highly profitable, with margins up to 80% on open-source models, and Dax notes that GPU supply is tight even for a company like OpenCode, forcing upfront payments and reservations.
Dax advocates for a 'slow down to speed up' approach, focusing on cleaning up tech debt and maintaining quality, which he believes is a key differentiator for startups competing against larger companies.
The role of engineers is shifting from writing code to setting up guardrails for AI agents, using patterns like domain-driven design to ensure safe and modular code, as agents are 'idiots that work 24/7.'
"We're shipping features we shouldn't, absorbing too many hacks, and the worst part is, we're not even moving faster. We just feel like we are."
"The natural place for them to do is hit that button as much as possible, do the same amount of work, and just cash in that extra time."
"The landmines are still there, they just won't blow up on you today, and we don't even pay attention to a lot of these landmines being placed by AI agents."