We can now generate code faster and cheaper than ever with AI. That feels like progress. But this week I started wondering if we're solving one problem while creating another.
I'm genuinely worried that engineers will become review machines. AI generates, humans approve. It feels efficient but over time we might lose the people who actually understand what's in the codebase. That scares me a little.
A colleague said something this week that I can't stop thinking about. "We're already generating legacy code inside our existing legacy. AI will do exactly the same."
I don't have the answers yet. But ignoring this feels wrong. Every team adopting AI needs to ask: how do we move faster without losing the people who keep things from falling apart?
How is your team handling this? Are your engineers still writing code themselves, or has the review queue already become the new bottleneck?
