Summary of Building OpenCode with Dax Raad

Overview of Building OpenCode with Dax Raad

This episode features Gergely Orosz talking with Dax Raad, co-founder of OpenCode, one of the fastest-growing open-source AI coding tools. The conversation centers on OpenCode’s explosive growth, the realities of building AI-powered dev tools, and why Dax believes AI has made some parts of software development easier while leaving the hardest parts essentially unchanged. A major theme is that AI can speed up code generation, but it does not automatically improve product judgment, team motivation, or software quality.

Key Themes and Takeaways

• OpenCode grew extremely fast

  • Launched in mid-2025 and scaled from roughly 650k monthly active users in December to 2.5M in January, then to around 6.5M–8M shortly after.
  • The company started with just 3 co-founders, later growing to about 20+ people.

• AI makes shipping easier, but not necessarily better

  • Dax says AI helps with execution, but the biggest challenges in product work remain:
    • deciding what to build,
    • avoiding unnecessary features,
    • maintaining quality,
    • and not accumulating technical debt.
  • His core feeling: “I’m working as hard as ever” despite faster tooling.

• The team had to slow down intentionally

  • Dax sent a memo to the team warning that they were:
    1. shipping too many features,
    2. accepting too many hacks,
    3. and not spending enough time cleaning things up.
  • His point: AI can make teams feel faster without actually improving output or direction.

• Positioning matters more than raw speed

  • OpenCode succeeded partly because it claimed the open-source coding agent territory before anyone else.
  • Dax views open source as a strong long-term wedge in dev tools, especially when major model vendors are competing aggressively.

• Inference and GPU access are major businesses and bottlenecks

  • Dax argues inference can be highly profitable because the marginal cost is largely electricity plus hardware amortization.
  • But the industry is constrained by GPU supply, and even fast-growing startups can hit capacity limits.
  • Large cloud and AI firms are consuming enormous amounts of supply, tightening the market.

• AI tools do not automatically increase organizational productivity

  • In many companies, engineers use AI to do the same amount of work with less effort, not to produce 10x more output.
  • Motivation, incentives, and organizational culture still determine whether productivity gains translate into business gains.

OpenCode’s Business Model

Dax described two main revenue lines:

1. OpenCode Zen (Inference / model access)

• Built initially to reduce onboarding friction.
• Lets users access models without setting up their own accounts and rate limits.
• Has grown into a major business for:
  • hosted inference,
  • open-source model access,
  • and aggregated model support.
• Dax said this side of the business hit a $50M run rate surprisingly quickly.

2. Enterprise control plane

• For companies with many engineers, OpenCode needs admin tooling:
  • permissions,
  • SSO,
  • budgets,
  • rate limits,
  • and provider management.
• This is the “boring” but necessary enterprise layer.
• Dax suggested they may eventually charge primarily for inference rather than the control plane itself.

Why OpenCode Won Early

Dax credits several product choices:

• Fast path to value

  • The first experience had to feel great immediately.
  • OpenCode focused on removing friction from installation and first use.

• Terminal experience mattered

  • The team built their own terminal rendering framework instead of using an off-the-shelf one.
  • This was an “irrational” quality decision, but it helped distinguish OpenCode from competitors.

• Open source as a strategic wedge

  • Dax believes dev tools often become standardized through open source adoption from the bottom up.
  • He sees OpenCode as benefiting from being the neutral, open alternative in a competitive model ecosystem.

• Quality as differentiation

  • He argues that quality is one of the last meaningful advantages small companies still have.
  • A product a year old can already start to “rot” if quality is neglected.

Dax’s View on AI, Teams, and Engineering

AI changes workflows, not fundamentals

• AI may reduce coding effort, but it does not remove the need for:
  • thinking,
  • product judgment,
  • architecture,
  • and careful system design.
• He said his time breakdown has barely changed:
  • before AI: 95% thinking, 5% doing
  • after AI: 96% thinking, 4% doing

Agents can hide bad judgment

• A major concern is that AI makes it too easy to ship:
  • hacks,
  • brittle fixes,
  • and short-term patches.
• The “muted prickle” idea: before AI, engineers felt the pain of writing hacky code. Now, agents absorb that pain for them, which can weaken judgment.

Engineering leadership is becoming more about guardrails

• Dax thinks the role of engineers is shifting toward:
  • building safe patterns,
  • conventions,
  • tests,
  • and systems that let less experienced humans or AI agents ship without breaking things.
• He sees this as less a new idea and more a return to old software engineering principles.

Product Philosophy

Dax emphasized a few durable principles:

• Solve one thing well
  • A strong product usually has one core idea that users can reach quickly.
• Feedback loops matter
  • Everyone on the team should feel the consequences of poor decisions.
• Clean-up is worth it
  • Even if refactoring doesn’t directly increase revenue, it preserves long-term product health.
• Taste still matters
  • He believes “taste” is real, but only if you genuinely care about quality.
  • If you say quality matters but don’t act like it does, it shows.

Notable Opinions and Observations

• On viral AI predictions

  • Dax is skeptical of social-media narratives like “24–29-year-old engineers will dominate because they have pre-AI principles and post-AI speed.”
  • He sees most of these predictions as people trying to reassure themselves during a time of rapid change.

• On startup culture

  • He’s more skeptical now of very young teams because he believes emotional maturity matters a lot in intense, intimate startup environments.

• On product quality vs. business logic

  • He pushes back on the idea that rational business decisions alone make great products.
  • Some of the best decisions are “irrational” quality investments that compound over time.

Dax’s Personal Background

• Grew up programming, with a software engineer father.
• Started building projects as a kid, including Minecraft servers and mods.
• Worked on early startups, then moved into open source full-time.
• Previously worked on:
  • Serverless Stack (SST)
  • OpenNext
• His experience across dev tools and open source clearly informs OpenCode’s strategy.

Final Takeaway

Dax’s message is that AI is real, powerful, and useful — but it does not remove the hardest parts of software building. OpenCode’s success came not just from using AI well, but from strong positioning, quality-focused product design, and a willingness to say no to superficial speed. In his view, the best teams will still need judgment, restraint, and craftsmanship, even in an AI-native world.