To all the “the rate of code generated by agents far exceeds the rate at which humans can review it, so if you want to understand the generated code, you become a bottleneck” takes out there. As if being called a bottleneck were an insult 🤷
When coding, I find AI especially valuable when trying to achieve consistency across the codebase once I have a directional change in design. The “look at this example of how I did it and apply it to all other cases in the codebase”, where AI does all the heavy lifting, drastically cuts the time to implement a change.
If you can’t change design cheaply because refactoring skills are lacking, you’re less likely to end up with a suitable design for a given problem that emerges from insights you get as a byproduct of refactoring.
I often ask myself if a technology, beside lowering the costs for which lowering the costs is beneficial, is also lowering the costs for which lowering the costs is detrimental.
The idea of managing (inherent) interdependence in systems is at least as important as the idea of decoupling (independence) that the software development industry is so predominantly obsessed with since the field’s inception.