Dax Raad

Dax Raad is a creator of OpenCode, a model-independent, open-source coding agent, and a founding-team member of SST, the developer infrastructure company he joined after its original founders had established it. He builds software around a demanding premise: developers should retain control of their infrastructure, identities, tools, and AI models without sacrificing products that are genuinely enjoyable to use.

Raad began working in software shortly after high school and started a company with a friend around 2012. When the business was acquihired, he led engineering at the acquiring company. Consulting across startups, a healthcare and transportation venture, and an engineering-director position followed, giving him firsthand exposure to the distance between competent engineering and effective product execution.

Frustration with Amazon Web Services tooling led him to SST, initially as a user and contributor. He invested in the company and joined its existing founding team around 2021. SST’s major 2023 rewrite streamlined configuration, command-line tooling, and application-development workflows, making infrastructure as code more approachable for working developers.

His associated projects also include OpenNext for AWS, which enables Next.js deployments on AWS, and OpenAuth, an open, standards-based authentication provider. His position on authentication is precise: outsourcing login mechanics can be sensible, but surrendering ownership of an application’s users to an outside vendor creates a much deeper dependency. In a public critique of authentication services, he distinguishes technical convenience from control over application identity.

Raad also co-founded Terminal, a developer-oriented coffee company with a command-line storefront. Its playful interface reflects his belief that technical products can be expressive without becoming inaccessible.

OpenCode and the limits of AI productivity

OpenCode emerged from Raad’s dissatisfaction with moving between Neovim, the terminal, and browser-based chatbots. Claude Code demonstrated the value of an agent that works directly with project files; OpenCode applies that terminal-native approach without binding developers to a single model provider. Its open-source architecture accommodates diverse editors, integrations, deployment environments, and workflows that one team cannot anticipate alone.

OpenCode Zen extends that principle to model access: Raad envisions aggregating demand, improving inference terms, and making a curated selection of proprietary and open models available across developer tools.

His approach to AI products rests on four concrete convictions:

• Developer choice is a product requirement. Coding agents should adapt to existing editors, terminals, models, and working habits instead of imposing a closed stack.
• Product discovery, onboarding, and retention remain difficult. Effective marketing must earn attention; onboarding must reach a clear first moment of value; enduring products must keep serving users as their needs grow.
• Powerful primitives make simplicity durable. Flexible underlying components can support an approachable default experience while exposing advanced configuration, self-hosting, and stronger data control to demanding teams.
• Productivity must be measured by useful outcomes. Raad questions elaborate multi-agent demonstrations and misleading benchmarks, emphasizing shipped software, understandable codebases, typed interfaces, and deterministic feedback.

His AI Engineer Code talk distills the larger point: AI accelerates implementation, but it cannot supply product judgment, original positioning, or the sustained execution required to make something people keep using.