There is a lot of noise about AI writing code. Most of it is either hype or fear. What gets talked about less is the boring, practical middle, which is where I actually live.
I use an AI coding agent as part of my normal workflow. Not to think for me, and not to paste in code I do not understand. I use it to move faster on the parts I already know how to do, to get a second read on an approach, and to handle the tedious edges so I can spend my attention on the hard decisions.
The rule I hold to is simple. I do not ship anything I could not explain line by line. If I cannot say why a piece of code is there, it does not go in. The agent is a tool, and I am still the one accountable for what lands in the repo.
Used that way, it is a real multiplier. Used the lazy way, it quietly builds a codebase nobody understands, including you. The difference is not the tool. It is whether you stay in the driver seat.