A few months ago, OpenCode had 650,000 monthly active users. Now that number is 6.5 million and heading toward 8 million—almost the fastest growth in the AI coding tool wave.
But more notable than the growth is what co-founder Dax Raad said in a recent podcast interview. While every AI tool founder touts smarter models and 10x productivity, Raad—someone who is directly enjoying the AI dividend—poured cold water on the entire industry.
He says AI coding tools create three fatal illusions in real engineering teams.
1. Exploding output does not mean a better product
AI has made writing code 10x faster. That's a fact. But Raad's point is simpler: the engineer's instinctive response becomes "I used to ship one feature; now I can ship ten."
Products increasingly resemble Frankenstein—every part is attached, but the whole is out of control. Every feature carries a long-term maintenance cost, and AI will not help you repay technical debt. With 10x more code, architectural consistency, maintainability, and core user experience can actually deteriorate.
Raad says AI makes "doing" cheap, but "figuring out what to do" remains expensive. His own time allocation barely changed: pre-AI, 95% thinking and 5% execution; now, 96% thinking and 4% execution. A smaller denominator, slightly different ratio.
Speed is a property of the tool. Direction is a property of the person. Conflating the two is the mistake many engineering teams are making.
2. AI is quietly erasing your sense of guilt
This is the most hidden one.
Before AI, writing bad code made you uncomfortable. The second time, more uncomfortable. The third time, you'd stop and refactor. That sting was an engineer's self-correction mechanism—a quality sensor.
Now? Agents do the dirty work for you. The landmines are still there, but they no longer hurt. Your judgement is being gradually stolen.
Raad has seen it firsthand on his team: some engineers use AI to pump out output, caring only about completing tasks; other, more principled engineers drown in garbage PRs, exhausted by cleaning up technical debt—until burnout. This isn't a tool problem; it's an engineering leadership problem. Most companies don't realize their code quality is silently collapsing.
When errors no longer hurt, you aren't far from the errors themselves.
3. Engineers use AI to save time, not to produce more
This is the sharpest point.
The story AI vendors sell to CFOs: use our tools and team output doubles. What actually happens? Engineers finish the same work in the same time and leave early.
Raad says this is perfectly rational for individuals—who doesn't want to go home earlier? But for companies, the ROI of AI tools is severely overstated. With incentives unchanged, most teams' output stays flat; engineers just clock out earlier.
You bought "10x productivity" and got "same output + less overtime." The information asymmetry between CFOs and AI vendors leaves engineers in the middle as silent beneficiaries.
Going deeper: why Raad dares to say this
He himself is riding the AI dividend—OpenCode is growing faster than anyone. But his underlying judgment is: positioning beats intelligence.
OpenCode wins not because it's the smartest, but because it occupies the "open source" position. In developer tools, open source ultimately becomes the default. Model competition is background noise—Claude is strong, but OpenAI and open models will catch up. In a market where model capabilities rapidly converge, neutrality and ecosystem position matter more than model IQ.
Raad also hates benchmarks. Academic benchmarks test "weird programming puzzles in three-file repos," which has nothing to do with real engineering. Optimizing for benchmarks makes you mistake that for what matters. Instead, OpenCode tests agents on real PRs from open-source projects: give an agent a real task and see whether it can produce code that actually gets merged.
Real work, not academic puzzles.
Conclusion
Dax Raad's interview offers a rare perspective: someone enjoying the AI dividend tearing down the AI hype. This isn't contrarianism—it's honesty.
For engineers, the reminder: don't be fooled by "output volume" or by the "feeling of speed." What truly matters—judgement, direction, quality perception—is something AI not only can't help with, but may quietly steal.
For teams, the reminder: once you buy AI tools, incentive structures and management practices must change too. Otherwise you won't get a "productivity revolution"—just "engineers who leave work earlier."
OpenCode went from 650,000 to 8 million MAU, yet Raad's message is: growth is fast, but don't trust its meaning too quickly.
References:
- The Pragmatic Engineer Podcast: Building OpenCode with Dax Raad
- 36Kr: Popular AI Coding Tool: Co-founder Says Engineers Still Have Hope
- Codacy AI Giants Podcast: The Creator of OpenCode Thinks You're Fooling Yourself About AI Productivity