A model can write a few hundred lines while you are still naming the function. It can sketch an API, patch a bug, and assemble a small app from a paragraph. If your value was the speed of your hands, that value just got cheaper. The interesting question is what did not.
Typing was never the whole jobLink to this heading
The hands were the visible part, the way coding speed was the visible part of a team that still released late. The job underneath is deciding what the system is allowed to do. What happens if this request arrives twice? What is a bug, and what is a product choice? Which shortcut will be expensive in six months? A model will answer those if you ask. It will also answer confidently when you forget to ask.
Judgment is the part that stays expensiveLink to this heading
A developer who can tell a generated diff "no" is worth more than one who can produce a longer diff. Taste sounds soft. In a codebase it is concrete: names that match the business, boundaries that do not leak, tests that fail for the reason you think, and the courage to delete a clever solution that the next person cannot safely change.
Use the speed on purposeLink to this heading
Let the tool draft. Then do the work it is bad at when nobody is watching: walk the failure path, name the invariant, and refuse the change you cannot explain. The people who stay valuable are not the ones who type faster than the model. They are the ones who still know what the program is for after the model has finished typing.