31/08/2026
AI has made it ridiculously easy to generate code.
And I’m starting to think our next engineering problem won’t be writing enough code.
It’ll be dealing with all the code we can now produce.
More PRs to review. More dependencies to keep track of. More abstractions. More surface area. More code that works but nobody on the team fully understands.
The old bottleneck was often getting something built.
Now, getting something built might be the easy part.
The harder part is deciding what deserves to exist in the first place, reviewing what was generated, and keeping the codebase understandable six months later.
I’m not against AI writing code. The productivity gains are real.
But generating code faster doesn’t automatically mean we’re engineering faster.
If our ability to produce code grows much faster than our ability to understand and maintain it, we haven’t removed the bottleneck.
We’ve just moved it.
I’m curious how other engineering teams are dealing with this. Are you seeing the same thing in your codebase?