Everyone talks about how AI makes us faster.

Very few people talk about how much faster it lets us burn out. Don't get me wrong: AI is one of the best things to happen to our industry since the compiler. I use it every day. My team is AI-first. I've completed proofs of concept, refactors, and reviews in days instead of weeks. The productivity gains are real. But I've started noticing a different problem. The bottleneck is no longer writing code.

The bottleneck is carrying context. As AI reduces the cost of execution, we naturally take on more. More projects. More conversations. More decisions. More parallel streams of work. At the same time, AI has dramatically increased the amount of information we produce. Design docs, RFCs, analyses, meeting notes, pull request descriptions, code—everything can now be generated faster than ever. The amount of text has increased by an order of magnitude.

The amount of human attention has not. I've already seen AI-generated documentation that wasn't fully proofread by its author. AI-generated code that received only superficial review. Not carelessness. Overload. We're creating information faster than we can consume it. For years, our industry optimized for production.

Now the scarce resource is attention. Even AI models eventually hit context limits and need a reset. Humans don't have that luxury. We simply keep accumulating context until the cost shows up as stress, fatigue, poor decisions, or burnout. And unlike a model, we can't just start a fresh session. Think of it like driving: AI should remain in the passenger seat.

The moment we stop reading, reviewing, and thinking critically about its output, we've moved it behind the wheel. The next challenge for our industry isn't how to generate more.

It's how to protect our attention while doing so.