I've been thinking about this tweet from John Carmack (who created Doom), and just generally about a future where AI does most of the coding and I'm managing AI - and... I hate it? It's not to say that I don't see this happening. There's a ton of capital flying around to make it happen, and I think end-users will inevitably gain the capability to build certain categories of simple, customised software.

But as someone who has deeply loved the practice of programming for more than a decade, the future that GitHub (who - in the midst of me writing this - launched Copilot Workspace), Devin, and the rest are painting seems to be leaving out so much of nuances that comes with building anything non-trivial. I absolutely love the idea of operating at a higher level of abstraction (just like how I enjoy writing modern programming languages compared to C or assembly) and getting my ideas/work out faster. But not spending my entire day just asking AI to do things or reviewing code all day - which makes me wonder: how do developers even get good at reviewing code if they aren't spending a ton of time writing code and problem solving? What's the right level of abstraction in this "promised future" that lets me get into the nuances of building software for anything non-trivial and continue problem solving in general? This is perhaps not even a question, and just a general thought I've been thinking a lot about.