AI Writing 100% of Code at Anthropic and OpenAI: What Does This Mean for Engineers?

Hey there. Ever heard the line, ‘AI writes all our code now’? I thought it was a joke until I saw a Reddit post from Anthropic and OpenAI engineers confirming it. Yep — they’re saying AI handles 100% of their code these days.

This came up in a Fortune article (linked in the Reddit thread) where Boris Cherny from Roon talks about how AI tools like GitHub Copilot or Anthropic’s own systems are now doing the heavy lifting for code. But here’s the catch: the humans aren’t just sitting around sipping coffee. They’re still handling the ‘big picture’ stuff — like figuring out what the software should do, testing for edge cases, and making sure the code plays nicely with the rest of the system.

It’s kind of wild to think about. If AI can write code, what do engineers do all day? The answer? They’re still problem-solvers, but the problems look different now. Instead of debugging line by line, they’re debugging the *plan* — like, ‘Why isn’t this AI-generated code working with the old database?’ or ‘How do we make this AI code secure?’

I’m curious how this affects the job market. Will we need fewer engineers? Probably not. But the skills they need might shift. Maybe we’ll see more demand for people who can train AI tools or interpret what they’re doing. Or maybe it’ll free up time for engineers to focus on creative, high-level work.

What’s fascinating is how quickly this evolved. A few years ago, AI code tools were a novelty. Now, they’re a core part of the workflow. If you’re into tech, this is a big deal. If you’re a developer reading this, you might want to start thinking about how to work *with* these tools instead of against them.

So, what do you think? Are we heading toward a future where engineers are just ‘AI supervisors’? Let me know in the comments — I’d love to hear your take.