> What makes this argument particularly poisonous, with or without AI, is that it is kinda true.

"End-to-end does everything" also glosses over the specificity of results. It'll tell you something is wrong but with much less detail as to what or how.

Even if the LinkedIn-pilled CEO declares all fixes must occur via a narrative-chaos engine, I would at least prefer it fixed just the actual problem instead of re-architecting everything until big broad tests stop failing.

I've got rage baited a bit today on Twitter, so much so, that I've ended up writing this mini-essay on the matter.

I'd love to hear other people opinions on this piece, it's my first long form article so I'd appriciate some feedback!

Thanks!