I'm going to give this a "maybe"; the challenge with large scale software isn't in authoring new code or whatever, it's in managing complexity. I think AIs / a

I'm going to give this a "maybe"; the challenge with large scale software isn't in authoring new code or whatever, it's in managing complexity. I think AIs / agents (and more importantly how we are learning to use them effectively) may help in that regard, but only if they are able to manage that complexity. This'll depend on context window sizes, their ability to explore a codebase, and how well their operators can provide relevant information. But that's only what they can consume (so codebase, documentation, etc), on top of that are the people that work with / for these systems for decades and who know a lot about things outside of what's written down.