Antirez—the creator of Redis—just published something that cuts through the noise.
In “Don’t Fall Into the Anti-AI Hype,” he lays out what he’s actually doing with AI, not what he thinks about it philosophically. The results are hard to argue with.
The Evidence
Four recent projects:
-
UTF-8 support for linenoise - Complete implementation with automated testing framework. Hours, not weeks.
-
Debugging transient Redis test failures - The kind of bug that usually eats days of printf debugging. AI traced it quickly.
-
A pure C BERT embedding library - 700 lines of working code in 5 minutes.
-
Redis Streams internal modifications - 20 minutes.
This isn’t someone theorizing about AI’s potential. It’s someone shipping production code to millions of users, showing what’s actually possible now.
The Craft Argument Falls Apart
The common resistance goes: “I care about craftsmanship. I want to understand what I’m building. AI-generated code is sloppy.”
Antirez addresses this directly. He’s spent decades advocating for code clarity, simplicity, and deep understanding. He still does. But he’s also pragmatic enough to recognize when a tool multiplies his output without sacrificing quality.
The craft isn’t in typing the characters. It’s in knowing what to build and why.
What This Means for Teams
If one of the most respected systems programmers in the world is shipping production code with AI assistance, the question isn’t whether your team should adopt these tools.
The question is how fast they can get up to speed.
Antirez frames it like open source: smaller teams can now compete with larger ones. The 10x engineer multiplier becomes available to everyone willing to learn.
The Real Concern
He’s not naive about the downsides. Will companies use AI to expand what small teams can accomplish? Or to reduce headcount while maintaining output?
That’s a governance and leadership question, not a technology question. The technology is here either way.
The Takeaway
Resistance to AI tools isn’t preserving craftsmanship. It’s falling behind.
The programmers who engage deeply with these tools—learning their limits, understanding when to trust them, knowing how to verify their output—will multiply their effectiveness.
The ones waiting for permission, or hoping the trend reverses, will watch their peers pull away.
Source: antirez.com/news/158