Everyone is using AI to write code now. But I think the real future of programming isn't AI generating full scripts or full products it's AI predicting the next token, and us deciding whether that token is right or not.
Here's why. When you write code manually, you're already doing something similar in your head predicting the next line based on patterns you've seen before, then checking if it's correct as you write it. Every line passes through your judgment.
When AI generates a full script or product, that judgment gets skipped. You review the code, sure, but not the way you'd review your own. You skim "this line does X, this line does Y" and move on. You're checking outcomes, not making decisions. That's a completely different mental process, and it's shallower.
But when AI suggests the next token, or the next line, instead of the next thousand lines, you're forced back into the same loop you'd use writing manually: is this right, or not? You stay the one making the call.
That's why I think the best use of AI in programming isn't generate my whole app. It's predict what comes next, and let me decide. Smaller suggestions keep engineers in the driver's seat. Full generations quietly take the wheel.
Top comments (0)